Studies in Systems, Decision and Control 25

Gerasimos G. Rigatos

Nonlinear Control
and Filtering Using
Differential Flatness
Approaches
Applications to Electromechanical
Systems
Studies in Systems, Decision and Control

Volume 25

Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
e-mail: kacprzyk@ibspan.waw.pl
About this Series

The series “Studies in Systems, Decision and Control” (SSDC) covers both new
developments and advances, as well as the state of the art, in the various areas of
broadly perceived systems, decision making and control- quickly, up to date and
with a high quality. The intent is to cover the theory, applications, and perspectives
on the state of the art and future developments relevant to systems, decision
making, control, complex processes and related areas, as embedded in the fields of
engineering, computer science, physics, economics, social and life sciences, as well
as the paradigms and methodologies behind them. The series contains monographs,
textbooks, lecture notes and edited volumes in systems, decision making and
control spanning the areas of Cyber-Physical Systems, Autonomous Systems,
Sensor Networks, Control Systems, Energy Systems, Automotive Systems, Bio-
logical Systems, Vehicular Networking and Connected Vehicles, Aerospace Sys-
tems, Automation, Manufacturing, Smart Grids, Nonlinear Systems, Power
Systems, Robotics, Social Systems, Economic Systems and other. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution and exposure which enable both a wide and rapid
dissemination of research output.

More information about this series at http://www.springer.com/series/13304

Gerasimos G. Rigatos

Nonlinear Control
and Filtering Using
Differential Flatness
Approaches
Applications to Electromechanical Systems

123
Gerasimos G. Rigatos
Unit of Industrial Automation
Industrial Systems Institute
26504 Rion Patras
Greece

ISSN 2198-4182 ISSN 2198-4190 (electronic)

Studies in Systems, Decision and Control
ISBN 978-3-319-16419-9 ISBN 978-3-319-16420-5 (eBook)
DOI 10.1007/978-3-319-16420-5

Library of Congress Control Number: 2015935609

Springer Cham Heidelberg New York Dordrecht London

© Springer International Publishing Switzerland 2015
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt
from the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained
herein or for any errors or omissions that may have been made.

Printed on acid-free paper

Springer International Publishing AG Switzerland is part of Springer Science+Business Media

(www.springer.com)
To Elektra
Foreword

This book analyzes the design of nonlinear filters and nonlinear adaptive controllers,
using exact linearization which is based on differential flatness theory and differ-
ential geometry methods. The obtained filters exhibit specific advantages as they
outperform in terms of accuracy of estimation and computation speed of other
nonlinear filters. The obtained adaptive controllers can be applied to a wider class of
nonlinear systems with unknown dynamics and can assure reliable functioning
of the control loop under uncertainty and under varying operating conditions.
Moreover, the book analyzes differential flatness theory-based control and filtering
methods for distributed parameter systems. The book presents a series of application
examples to confirm the efficiency of the proposed control and filtering schemes for
various electromechanical systems. These include:
(i) Industrial Robotics: neuro-fuzzy adaptive control of multi-DOF robots and
underactuated robotic manipulators, observer-based neuro-fuzzy control of
multi-DOF robotic manipulators, state estimation-based control of multi-
DOF robots and underactuated robotic manipulators, state estimation-based
control of robots and mechatronic systems under disturbances and model
uncertainties.
(ii) Mobile robotics and vehicles: state estimation-based control of autonomous
vehicles, control of cooperating vehicles with the use of nonlinear filtering,
distributed fault diagnosis for autonomous vehicles, velocity control of
four-wheel vehicles, active vehicle suspension control, control of various
types of unmanned vehicles, such as AGVs, UAVs, USVs and AUVs.
(iii) Electric Power Generation: state estimation-based control of the PMSG
(Permanent Magnet Synchronous Generator), state estimation-based control
of the DFIG (Doubly-fed Induction Generator), state estimation-based
control and synchronization of distributed PMSGs.
(iv) Electric Motors and Actuators: neuro-fuzzy adaptive control of the DC
motor, neuro-fuzzy adaptive control of the Induction motor, state estimation-
based control of the DC motor, state estimation-based control of the
Induction Motor.

vii
viii Foreword

(v) Power Electronics: state estimation-based control of power converters, state

estimation-based control of photovoltaic systems, state estimation-based
control and synchronization of distributed inverters.
(vi) Internal combustion engines: neuro-fuzzy adaptive control of diesel engines,
neuro-fuzzy adaptive control of spark-ignited (SI) engines, state estimation-
based control of valves in ship diesel engines, state estimation-based control
of turbocharged diesel engines, state estimation-based control of spark
ignited engines, state estimation-based control of the air-fuel ratio system in
combustion engines under various perturbations.
(vii) Distributed Parameter Systems: pointwise flatness-based control of distrib-
uted parameter systems, boundary flatness-based control of distributed
parameter systems, state estimation of distributed parameter systems, fault
diagnosis for sensor networks which monitor distributed parameter systems,
condition monitoring for multi-DOF buildings.
(viii) Communication Systems: state estimation for synchronization and channel
equalization in chaotic communication systems, filtering for compensation of
communication delays and packet drops in networked robotic control,
feedback control and stabilization of chaotic dynamics.
The book is primarily addressed to the academic community. The content of the
book can be the basis for teaching undergraduate or postgraduate courses on
nonlinear control systems. Therefore, it can be used by both academic tutors and
students as a reference book for such a course. The book is suitable for departments
of electrical, industrial and mechanical engineering, which can include in their
curriculum nonlinear control courses on the topic of the present monograph.
Moreover, the book is addressed to the engineering community. Engineers
working in industrial production, in electric power generation, in the design of
transportation systems, in the development of automation and electromechanical
equipment, or in several other application fields frequently come against nonlinear
control problems which have to be solved, at low cost and within hard time con-
straints. To cope efficiently with such control problems, engineers should be
acquainted with control methods of generic use, improved reliability, and clear
implementation stages. The nonlinear control and estimation methods which are
analyzed in this book fulfill the aforementioned requirements and can be a powerful
tool and a useful companion for engineers working on practical electromechanical
problems.

Athens Gerasimos G. Rigatos

March 2015
Preface

Differential flatness theory is currently a main direction in nonlinear control

systems. Differential flatness theory enables to develop global linearizing methods
for nonlinear dynamical systems, thus also facilitating the solution of complicated
nonlinear control and filtering problems. The present book aims at presenting recent
advances in differential flatness theory for nonlinear control and estimation.
Actually, it shows that through differential flatness theory it is possible to perform
filtering and state estimation for a wide class of nonlinear dynamical systems,
including single input—single output, multi input—multi output dynamical models
or even distributed parameter systems.
The book analyzes the design of nonlinear adaptive controllers and nonlinear
filters, using exact linearization which is based on differential flatness theory. The
obtained adaptive controllers can be applied to a wide class of nonlinear systems
with unknown dynamics and can assure reliable functioning of the control loop
under uncertainty and varying operating conditions. The obtained filters exhibit
specific advantages as they outperform in terms of accuracy of estimation and
computation speed other nonlinear filters. The book presents a series of application
examples to confirm the efficiency of the proposed nonlinear filtering and adaptive
control schemes for various electromechanical systems. These include: (i) Industrial
Robotics, (ii) Mobile Robotics and Autonomous Vehicles, (iii) Electric Power
Generation, (iv) Electric Motors and Actuators, (v) Power Electronics, (vi) Internal
Combustion Engines, (vii) Distributed Parameter Systems, (viii) Communication
Systems.
The book aims at providing an informative overview of results on flatness-based
control for single and multi-input dynamical systems which are described by
ordinary differential equations. The monograph analyzes the stages of design of
nonlinear adaptive controllers and nonlinear Kalman Filters, using differential
flatness theory. The application of differential flatness theory enables transformation
of the system dynamics to the linear canonical (Brunovsky) form. This is feasible
for all single input nonlinear dynamical systems and for MIMO dynamical systems,
which can be linearized through static state feedback. Moreover, for MIMO

ix
x Preface

dynamical systems which accept only dynamic feedback linearization, it is also

possible to succeed transformation to the canonical Brunovsky form.
In particular, the book comes up with new adaptive neuro-fuzzy control methods
that are based on differential flatness theory and which are suitable for both SISO
and MIMO dynamical systems. The differential flatness theory-based approach
extends the class of nonlinear systems to which adaptive neuro-fuzzy control can be
applied. By proving that a dynamical system satisfies differential flatness properties,
its transformation to the linear canonical (Brunovsky) form is possible. After such a
transformation, a modified control input is applied to the system, which contains not
only the initial control signal but also unknown terms associated with the system’s
nonlinear dynamics. These terms are identified online by neuro-fuzzy approxima-
tors. Thus, a nonlinear adaptive control scheme is formulated in which identifica-
tion of the unknown system dynamics is first performed and subsequently this
information is used for the computation of the control inputs. The stability of the
adaptive control scheme is proven through Lyapunov methods. Additionally,
adaptive neuro-fuzzy control schemes are developed which succeed simultaneously
the identification of the unknown system dynamics and estimation of the non-
measurable elements of the system’s state vector. The feedback loop of these
adaptive fuzzy control schemes contains neuro-fuzzy approximators of the system’s
nonlinear model and also state observers which provide estimates of the system’s
state vector. The stability of such control schemes is proven again with the use of
Lyapunov methods.
Furthermore, the book comes up with a new nonlinear Kalman Filtering
approach under the name “Derivative-free nonlinear Kalman Filter” that is based on
differential flatness theory. The Derivative-free nonlinear Kalman Filter consists
of the Kalman Filter recursion applied on the linearized equivalent model of the
treated system, together with an inverse transformation based on differential flatness
theory that enables to retrieve estimates for the state variables of the initial nonlinear
system. This is a nonlinear filtering algorithm which does not need to compute
partial derivatives and Jacobian matrices. In terms of accuracy of the provided state
estimation the algorithm’s performance is equivalent to that of the Unscented
Kalman Filter and is significantly improved compared to the Extended Kalman
Filter. In terms of speed of computation, the Derivative-free nonlinear Kalman
Filter outperforms other nonlinear estimation algorithms. The generalization of the
Derivative-free nonlinear Kalman Filter to the case of a distributed computing
environment results in the Derivative-free distributed nonlinear Kalman Filter. For
the latter distributed filtering method, it has been proven that it has better perfor-
mance than the widely used Extended Information Filter. Through a series of
examples the book shows that the proposed nonlinear filtering method can be part
of control schemes for nonlinear dynamical systems.
Moreover, the book aims at presenting flatness-based control methods for
systems with spatiotemporal dynamics. Such systems are described by partial
differential equations together with the associated boundary conditions and play a
critical role in several engineering problems, such as vibrating structures, flexible-
link robots, waveguides and optical fibers, heat conduction, etc. Additionally, the
Preface xi

book aims at presenting differential flatness approaches for state estimation/filtering

and fault diagnosis for distributed parameter systems. Being in position to recon-
struct the dynamics of such systems out of a limited number of sensor measure-
ments is important for monitoring their condition. Results on filtering and fault
diagnosis for nonlinear PDE models are shown to be applicable to systems of wave-
type and diffusion-type dynamics.
The control and filtering methods analyzed in the book are generic and suitable
for classes of systems, therefore one can anticipate the use of the book’s methods to
various engineering and science problems. The structure of the book is as follows:
In Chap. 1, an analysis is given about the basics of systems theory which can be
used in the modeling of nonlinear dynamics. To understand the oscillatory behavior
of nonlinear dynamical systems, benchmark examples of oscillators are given.
Moreover, the following properties are introduced: phase diagram, isoclines,
attractors, local stability, bifurcations of fixed points, and chaos properties. Next,
differential geometry and Lie algebra-based control are analyzed as a predecessor to
differential flatness theory-based control. First, the differential geometric approach
and the Frobenius theorem are analyzed. Next, the concept of input–output line-
arization is introduced and its association to transformation to normal forms is
explained. Furthermore, the concept of input-state linearization is presented and the
stages of its implementation are explained. Necessary and sufficient conditions for
applying input-state linearization and input–output linearization are provided.
In Chap. 2, flatness-based control for lumped parameter systems is first analyzed.
Such systems are described by ordinary differential equations. The chapter over-
views the definition and properties of differentially flat systems and presents basic
classes of differentially flat systems. First the equivalence property is explained,
which signifies that it is possible to transform differentially flat systems through a
change of variables into the linear canonical form. The chapter presents examples of
single input dynamical systems which are written into the linear canonical form by
using the differential flatness theory diffeomorphism and explains the design of the
associated feedback control loop. The case of MIMO differentially flat dynamical
system is also examined. It is shown that differentially flat systems which admit
static feedback linearization can be transformed into the linear canonical form. It is
also shown that for MIMO differentially flat systems that admit only dynamic
feedback linearization, it is again possible to succeed transformation to the linear
canonical form and subsequently to design state feedback controllers. Moreover,
elaborated criteria for checking differential flatness properties of dynamical systems
are given. Finally, the chapter studies differential flatness properties for distributed
parameter systems and methods for their transformation to an equivalent linear
canonical form.
In Chap. 3, differential flatness theory-based adaptive fuzzy control is proposed
for complex nonlinear dynamical systems. First, single-input single-output
dynamical systems are studied and it is shown how flatness-based adaptive fuzzy
controllers can be designed for such systems. Moreover, it is shown that multi-input
multi-output dynamical systems which admit static feedback linearization can be
transformed to a decoupled and linear canonical form for which the design of the
xii Preface

flatness-based adaptive fuzzy controller is a straightforward procedure. The latter

results can be also extended to the case of MIMO systems that admit exclusively
dynamical feedback linearization.
In Chap. 4, a new filtering method for nonlinear dynamical systems is analyzed.
The filtering method is based on differential flatness theory and is known as
Derivative-free nonlinear Kalman Filter. First the filtering method is applied to
lumped dynamical systems, that is, systems that are described by ordinary differ-
ential equations. Moreover, the problem of distributed nonlinear filtering is solved,
that is, the problem of fusion of the outcome of distributed local filtering procedures
(local nonlinear Kalman Filters) into one global estimate that approximates the
system’s state vector with improved accuracy.
In Chap. 5, differential flatness theory is used for the solution of industrial
robotics problems. These comprise among others adaptive control of MIMO robotic
manipulators without prior knowledge of the robot’s dynamical model, adaptive
control of underactuated robotic manipulators (that is, robots having less actuators
than their degrees of freedom), observer-based adaptive control of MIMO robotic
manipulators in which uncertainty is not related only to the unknown dynamic
model of the robot but also comes from the inability to measure all elements of the
robot’s state vector, and Kalman Filter-based control of MIMO robotic manipula-
tors. Finally, differential flatness theory is proposed for developing a robot control
scheme over a communication network that is characterized by transmission delays
or losses in the transmitted information.
In Chap. 6, it is proposed to use nonlinear filtering and control methods based on
differential flatness theory for autonomous vehicles control. In particular the chapter
analyzes steering control, localization and autonomous navigation of land vehicles,
unmanned surface vessels and unmanned aerial vehicles. It is shown that through the
application of differential flatness theory, one can obtain solution for the following
nontrivial problems: state estimation-based control of autonomous vehicles, state
estimation-based control of cooperating vehicles, distributed fault diagnosis for
autonomous vehicles, velocity control of ground vehicles under model uncertainties
and external disturbances, active control of vehicle suspensions, state estimation-
based control of unmanned aerial vehicles of the quadrotor type, and finally state
estimation-based control of unmanned surface vessels of the hovercraft type.
In Chap. 7, it is proposed to use differential flatness theory for nonlinear filtering
and nonlinear control problems met in electric power generation. Power generators
of various types are considered such as DFIGs and PMSGs, while the mode of
operation of these generators can be either the stand-alone one (single machine
infinite bus model), or the generators can function as part of the power grid (multi-
area multi machine power generation models). The chapter shows how differential
flatness theory can provide efficient solutions to the following problems: (i) state
estimation-based control of the PMSG, (ii) state estimation-based control of the
DFIG, (iii) state estimation-based control and synchronization of distributed power
generators of the PMSG type.
In Chap. 8, it is proposed to apply differential flatness theory-based nonlinear
filtering and control methods, to electric motors and actuators and to motion
Preface xiii

transmission systems. To this end, differential flatness theory is used for adaptive
control of the DC motor, for adaptive control of the induction motor, for state
estimation-based control of the DC motor, for state estimation-based control of
asynchronous electric motors and finally for observer-based adaptive fuzzy control
of microactuators (MEMS).
In Chap. 9, differential flatness theory is used to solve nonlinear filtering and
control problems which appear in power electronics such as voltage source con-
verters (VSC) and inverters, when these devices are used for connecting various
types of power generation units (AC and DC) to the grid. In particular, the chapter
proposes differential flatness theory for state estimation-based control of three-
phase voltage source converters, for state estimation-based control of voltage
inverters (finding application to the connection of photovoltaics to the electricity
grid), and finally for decentralized control and synchronization of distributed
inverters which are used to connect distributed DC power units with the electricity
network.
In Chap. 10, differential flatness theory is applied to nonlinear filtering and
control methods for internal combustion engines. The presented methods are con-
cerned with robust control and filtering for valves of diesel engines, with filtering
flatness-based of turbocharged diesel engines, with embedded adaptive control of
turbocharged Diesel engines, with embedded control and filtering of spark-ignited
engines, with embedded adaptive control of spark ignited engines and finally with
embedded control and filtering of the air-fuel ratio in combustion engines.
In Chap. 11, differential flatness theory-based methods for nonlinear filtering and
nonlinear control are applied to chaotic dynamical systems. Flatness-based adaptive
fuzzy control is proposed first for chaotic dynamical systems and manages to
modify the behavior of such systems without any knowledge of their dynamic
model. Next, differential flatness theory is proposed for developing a chaotic
communication system. A differential flatness theory-based Kalman Filtering
approach is proposed for performing blind equalization of the chaotic communi-
cation channel and for the synchronization between the transmitter and the receiver.
In Chap. 12, it is proposed to use differential flatness theory in nonlinear filtering
and control problems of distributed parameter systems, that is, systems which are
described by partial differential equations (of the parabolic, hyperbolic or elliptic
type). Methods of pointwise control of nonlinear PDE dynamics are introduced,
while methods for state estimation in nonlinear PDE models are developed. Such
results are applicable to communication systems (transmission lines, optical fibers,
and electromagnetic waves propagation), to electronics (Josephson junctions), to
manufacturing (heat diffusion control in the gas-metal arc welding proces), in
structural engineering (dynamic analysis of buildings under seismic waves,
mechanical structures subjected to vibrations, pendulum chains), in sensor networks
(fault diagnosis in distributed sensors monitoring systems with PDE dynamics), etc.
Finally, in Chap. 13, it is demonstrated that differential flatness theory is in the
background of other control methods, such as backstepping control and optimal
control. First, the chapter shows that differential flatness theory is in the background
of backstepping control. It is shown that to implement flatness-based control, it is
xiv Preface

not always necessary to transform the system into the trivial (linear canonical) form,
but it may suffice to decompose it into a set of flat subsystems associated with the
rows of its state-space description. The proposed flatness-based control method can
solve efficiently several nonlinear control problems, met for instance in robotic and
power generation systems. Moreover, the chapter shows how differential flatness
theory can be used for implementing boundary control of distributed parameter
systems. As a case study the control of a nonlinear wave PDE is considered.
The manuscript is suitable for teaching nonlinear control and nonlinear filtering
methods at late undergraduate and at postgraduate level. Such a course can be part
of the curriculum of several university departments, such as Electrical Engineering,
Mechanical Engineering, Computer Science, Physics, etc. The proposed book
contains teaching material which can be also used in a supplementary manner to the
content of undergraduate nonlinear control courses. The book can also serve per-
fectly the needs of a postgraduate course on nonlinear control and nonlinear esti-
mation methods.
Moreover, the book can be a useful reference for researchers, academic tutors,
and engineers that come against complex nonlinear control and nonlinear dynamics
problems. The book has both theoretical and practical value. It provides efficient
solutions to control problems associated with several electromechanical systems
met in industrial production, in transportation systems, in electric power generation,
and in many other applications. The control and filtering methods analyzed in the
book are generic and applicable to classes of systems, therefore one can anticipate
the application of the book’s methods to various types of electromechanical sys-
tems. I hope the book to become a useful reference not only for the industrial
systems, and the robotics and control community, but also for researchers and
engineers in the fields of signal processing, computational intelligence, and
mechatronics.

Athens, Greece Gerasimos G. Rigatos

March 2015
Acknowledgments

The author of this monograph would like to thank researchers and university staff
who through their constructive review and comments have contributed to the
improvement of the structure and content of this manuscript.

xv
Contents

1 Nonlinear Dynamical Systems and Global Linearizing

Control Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Characteristics of the Dynamics of Nonlinear Systems. . . . . . . 1
1.3 Computation of Isoclines . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4 Basic Features in the Study of Nonlinear Dynamics . . . . . . . . 8
1.4.1 The Phase Diagram . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.2 Stability Analysis of Nonlinear Systems. . . . . . . . . . . 9
1.4.3 Stability Analysis of Nonlinear Models . . . . . . . . . . . 11
1.5 Phase Diagrams and Equilibria of Nonlinear Models . . . . . . . . 12
1.5.1 Phase Diagrams for Linear Dynamical Systems . . . . . 12
1.5.2 Multiple Equilibria for Nonlinear Dynamical
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.5.3 Limit Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
1.6 Bifurcations in Nonlinear Dynamics . . . . . . . . . . . . . . . . . . . 21
1.6.1 Bifurcations of Fixed Points of Nonlinear Models . . . . 21
1.6.2 Saddle-Node Bifurcations of Fixed Points
in a One-Dimensional System . . . . . . . . . . . . . . . . . 21
1.6.3 Pitchfork Bifurcation of Fixed Points. . . . . . . . . . . . . 22
1.6.4 The Hopf Bifurcation . . . . . . . . . . . . . . . . . . . . . . . 24
1.7 Predecessors of Differential Flatness Theory. . . . . . . . . . . . . . 26
1.7.1 The Differential Geometric Approach . . . . . . . . . . . . 26
1.7.2 Elaboration on the Frobenius Theorem . . . . . . . . . . . 29
1.7.3 Input–Output Linearization. . . . . . . . . . . . . . . . . . . . 30
1.7.4 Elaborating on Input–Output Linearization . . . . . . . . . 33
1.7.5 Input-State Linearization . . . . . . . . . . . . . . . . . . . . . 37
1.7.6 Stages in the Implementation of Input-State
Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . ... 43
1.7.7 Input–Output and Input-State Linearization
for MIMO Systems . . . . . . . . . . . . . . . . . . . . . . ... 44
1.7.8 Dynamic Extension . . . . . . . . . . . . . . . . . . . . . . ... 45

xvii
xviii Contents

2 Differential Flatness Theory and Flatness-Based Control . . . . . . . 47

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Definition of Differentially Flat Systems . . . . . . . . . . . . . . . . 48
2.2.1 The Background of Differential Flatness Theory . . . . . 48
2.2.2 Differential Flatness for Finite Dimensional
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.3 Properties of Differentially Flat Systems . . . . . . . . . . . . . . . . 57
2.3.1 Equivalence and Differential Flatness . . . . . . . . . . . . 57
2.3.2 Differential Flatness and Trajectory Planning . . . . . . . 72
2.3.3 Differential Flatness, Feedback Control
and Equivalence . . . . . . . . . . . . . . . . . . . . . . . . . .. 75
2.4 Flatness-Based Control and State Feedback for Systems
with Model Uncertainties . . . . . . . . . . . . . . . . . . . . . . . . . .. 79
2.5 Classification of Types of Differentially Flat Systems . . . . . .. 82
2.5.1 Criteria About the Differential Flatness of a System . .. 82
2.5.2 A Sufficient Condition for Showing that a System
Is Not Differentially Flat . . . . . . . . . . . . . . . . . . . . . 85
2.5.3 Liouvillian and Nondifferentially Flat Systems . . . . . . 86
2.6 Elaborated Criteria for Checking Differential Flatness . . . . . . . 87
2.6.1 Implicit Control Systems on Manifolds of Jets . . . . . . 87
2.6.2 The Lie-Backlünd Equivalence for Implicit
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 89
2.6.3 Conditions for Differential Flatness of Implicit
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 90
2.6.4 Example of Elaborated Differential Flatness
Criteria to Nonlinear Systems . . . . . . . . . . . . . . . . .. 93
2.7 Distributed Parameter Systems and Their Transformation
into the Canonical Form . . . . . . . . . . . . . . . . . . . . . . . . . .. 96
2.7.1 State-Space Description of a Heat Diffusion
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 96
2.7.2 Differential Flatness of the Nonlinear Heat
Diffusion PDE . . . . . . . . . . . . . . . . . . . . . . . . . . .. 99

3 Nonlinear Adaptive Control Based on Differential

Flatness Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 103
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ....... 103
3.2 Flatness-Based Adaptive Neuro-Fuzzy Control
for SISO Systems . . . . . . . . . . . . . . . . . . . . . . . . . ....... 104
3.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . ....... 104
3.3 Flatness-Based Adaptive Fuzzy Control for SISO
Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . . ....... 105
3.3.1 Transformation of SISO Nonlinear Systems
into a Canonical Form . . . . . . . . . . . . . . . . ....... 105
Contents xix

3.3.2 Adaptive Control Law for SISO Nonlinear

Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 106
3.3.3 Approximators of SISO System Unknown
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 107
3.3.4 Lyapunov Stability Analysis for SISO Dynamical
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 109
3.3.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . .. 111
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 116
3.4.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 116
3.4.2 Differential Flatness for MIMO Nonlinear
Dynamical Systems . . . . . . . . . . . . . . . . . . . . . . . .. 117
3.4.3 Flatness-Based Adaptive Fuzzy Control for MIMO
Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . .. 120
3.4.4 Flatness-Based Control for a MIMO Robotic
Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 122
3.4.5 Lyapunov Stability Analysis for MIMO Nonlinear
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 127
3.4.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . .. 133

4 Nonlinear Kalman Filtering Based on Differential

Flatness Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
4.2 The Derivative-Free Nonlinear Kalman Filter . . . . . . . . . . . . . 142
4.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
4.2.2 Extended Kalman Filtering for Nonlinear
Dynamical Systems . . . . . . . . . . . . . . . . . . . . . .... 143
4.2.3 Derivative-Free Kalman Filtering to SISO
Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . .... 149
4.2.4 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . .... 152
4.2.5 Derivative-Free Kalman Filtering for MIMO
Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . 163
4.2.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter . . . . . 172
4.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
4.3.2 Overview of the Extended Information Filter . . . . . . . 173
4.3.3 Distributed Filtering for Sensorless Control . . . . . . . . 177
4.3.4 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

5 Differential Flatness Theory and Industrial Robotics . . . . . . . . . . 183

5.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots . . . . 185
5.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
xx Contents

5.2.2 Dynamic Model of the Closed-Chain 2-DOF

Robotic System . . . . . . . . . . . . . . . . . . . . . . . . ... 186
5.2.3 Linearization of the Closed-Chain 2-DOF
Robotic System Using Lie Algebra Theory . . . . . ... 192
5.2.4 Differential Flatness of the Underactuated
Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 195
5.2.5 Flatness-Based Adaptive Fuzzy Control
for the Underactuated Robot. . . . . . . . . . . . . . . . . . . 198
5.2.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots . . . 199
5.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
5.3.2 Estimation of the Robot’s State Vector . . . . . . . . . . . 201
5.3.3 Application of Flatness-Based Adaptive Fuzzy
Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 203
5.3.4 Dynamics of the Observation Error . . . . . . . . . . . ... 204
5.3.5 Approximation of the System’s Unknown
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 205
5.3.6 Lyapunov Stability Analysis. . . . . . . . . . . . . . . . ... 206
5.3.7 The Role of Riccati Equation Coefficients
in Observer-Based Adaptive Fuzzy Control . . . . . . . . 212
5.3.8 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
5.4 State Estimation-Based Control of Underactuated Robots . . . . . 218
5.4.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
5.4.2 Derivative-Free Nonlinear Kalman Filter
for the Closed-Chain 2-DOF Robotic System . . . . . . . 219
5.4.3 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 222
5.5 Distributed Filtering Under External Disturbances . . . . . . . . . . 223
5.5.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
5.5.2 Dynamics and Control of the Robot . . . . . . . . . . . . . 225
5.5.3 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
5.6 Distributed Nonlinear Filtering Under Measurement Delays . . . 230
5.6.1 Networked Control Under Communication
Disturbances . . . . . . . . . . . . . . . . . . . . . . . . . . ... 230
5.6.2 Networked Kalman Filtering for an Autonomous
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 231
5.6.3 Smoothing Estimation in Case of Delayed
Measurements . . . . . . . . . . . . . . . . . . . . . . . . . ... 232
5.6.4 Distributed Filtering-Based Fusion of the Robot’s
State Estimates . . . . . . . . . . . . . . . . . . . . . . . . . ... 235
5.6.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 236
Contents xxi

6 Differential Flatness Theory in Mobile Robotics

and Autonomous Vehicles . . . . . . . . . . . . . . . . . . . . . . . . . . .... 239
6.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 239
6.2 State Estimation-Based Control of Autonomous Vehicles . .... 241
6.2.1 Localization and Autonomous Navigation
of Ground Vehicles . . . . . . . . . . . . . . . . . . . . . .... 241
6.2.2 Application of Derivative-Free Kalman Filtering
to MIMO UGV Models . . . . . . . . . . . . . . . . . . . . . . 242
6.2.3 Controller Design for UGVs. . . . . . . . . . . . . . . . . . . 244
6.2.4 Derivative-Free Kalman Filtering for UGVs . . . . . . . . 247
6.2.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
6.2.6 Derivative-Free Kalman Filter-Based Navigation
of the Autonomous Vehicle . . . . . . . . . . . . . . . .... 252
6.3 State Estimation-Based Control and Synchronization
of Cooperating Vehicles . . . . . . . . . . . . . . . . . . . . . . . .... 261
6.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . .... 261
6.3.2 Distributed Kalman Filtering for Unmanned
Ground Vehicles. . . . . . . . . . . . . . . . . . . . . . . . . . . 263
6.3.3 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
6.4 Distributed Fault Diagnosis for Autonomous Vehicles . . . . . . . 265
6.4.1 Integrity Testing in Navigation Sensors of AGVs . . . . 265
6.4.2 Sensor Fusion for AGV Navigation. . . . . . . . . . . . . . 267
6.4.3 Canonical Form for the AGV Model . . . . . . . . . . . . . 270
6.4.4 Derivative-Free Extended Information Filtering
for UGVs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270
6.4.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
6.5 Velocity Control of 4-Wheel Vehicles . . . . . . . . . . . . . . . . . . 273
6.5.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
6.5.2 Dynamic Model of the Vehicle. . . . . . . . . . . . . . . . . 276
6.5.3 Flatness-Based Controller for the 3-DOF
Vehicle Model . . . . . . . . . . . . . . . . . . . . . . . . .... 280
6.5.4 Estimation of Vehicle Disturbance Forces
with Kalman Filtering . . . . . . . . . . . . . . . . . . . . . . . 283
6.5.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
6.6 Active Vehicle Suspension Control . . . . . . . . . . . . . . . . . . . . 288
6.6.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288
6.6.2 Dynamic Model of Vehicle Suspension . . . . . . . . . . . 292
6.6.3 Flatness-Based Control for a Suspension Model . . . . . 296
6.6.4 Compensating for Model Uncertainty
with the Use of the H1 Kalman Filter. . . . . . . . .... 297
6.6.5 Robust State Estimation with the Use
of Disturbance Observers . . . . . . . . . . . . . . . . . .... 300
6.6.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . .... 302
xxii Contents

6.7 State Estimation-Based Control of Quadrotors . . . . . . . . . . . . 304

6.7.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304
6.7.2 Kinematic Model of the Quadropter . . . . . . . . . . . . . 310
6.7.3 Euler-Lagrange Equations for the Quadropter . . . . . . . 311
6.7.4 Design of Flatness-Based Controller
for the Quadrotor’s Model . . . . . . . . . . . . . . . . . ... 313
6.7.5 Estimation of the Quadrotor’s Disturbance Forces
and Torques with Kalman Filtering . . . . . . . . . . . ... 315
6.7.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 318
6.8 State Estimation-Based Control of the Underactuated
Hovercraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 320
6.8.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 320
6.8.2 Lie Algebra-Based Control of the Underactuated
Hovercraft . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 323
6.8.3 Flatness-Based Control of the Underactuated
Vessel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 329
6.8.4 Disturbances’ Compensation with the Use
of the Derivative-Free Nonlinear Kalman Filter . . ... 330
6.8.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 332

7 Differential Flatness Theory and Electric Power Generation . . . . . 337

7.1 Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
7.2 State Estimation-Based Control of PMSGs . . . . . . . . . . . . . . . 338
7.2.1 The PMSG Control Problem . . . . . . . . . . . . . . . . . . 338
7.2.2 Dynamic Model of the Permanent Magnet
Synchronous Generator . . . . . . . . . . . . . . . . . . . ... 340
7.2.3 Lie Algebra-Based Design of State Estimators
for the PMSG. . . . . . . . . . . . . . . . . . . . . . . . . . ... 342
7.2.4 Differential Flatness of the PMSG. . . . . . . . . . . . ... 347
7.2.5 Estimation of PMSG Disturbance Input
with Kalman Filtering . . . . . . . . . . . . . . . . . . . . . . . 349
7.2.6 Simulation Experiments . . . . . . . . . . . . . . . . . . . . . . 352
7.3 State Estimation-Based Control of DFIGs . . . . . . . . . . . . . . . 358
7.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
7.3.2 The Complete Sixth-Order Model of the Induction
Generator. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 359
7.3.3 Input–Output Linearization of the DFIG Using
Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 363
7.3.4 Input–Output Linearization of the DFIG Using
Differential Flatness Theory . . . . . . . . . . . . . . . . ... 367
7.3.5 Kalman Filter-Based Disturbance Observer
for the DFIG Model . . . . . . . . . . . . . . . . . . . . . ... 371
7.3.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 373
Contents xxiii

7.4 Flatness-Based Control of DFIG in Cascading Loops . ...... 377

7.4.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . ...... 377
7.4.2 A New Proof of the Differential Flatness
of the DFIG . . . . . . . . . . . . . . . . . . . . . . . . ...... 378
7.4.3 Control of the DFIG in Cascading Loops. . . . ...... 380
7.4.4 EKF Implementation for Sensorless Control
of the DFIG . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
7.4.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
7.5 State Estimation-Based Control of Distributed PMSGs. . . . . . . 388
7.5.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
7.5.2 Dynamic Model of the Distributed Power
Generation Units. . . . . . . . . . . . . . . . . . . . . ...... 390
7.5.3 Lie Algebra-Based Design of a Feedback
Controller for the PMSG . . . . . . . . . . . . . . . ...... 391
7.5.4 Differential Flatness of the Distributed
PMSG Model . . . . . . . . . . . . . . . . . . . . . . . ...... 393
7.5.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . ...... 397

8 Differential Flatness Theory for Electric Motors

and Actuators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403
8.2 Flatness-Based Adaptive Control of DC Motors . . . . . . . . . . . 404
8.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
8.2.2 Dynamics and Linearization of the DC
Motor Model . . . . . . . . . . . . . . . . . . . . . . . . . . ... 405
8.3 Flatness-Based Control of Induction Motors
in Cascading Loops. . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 409
8.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 409
8.3.2 A Cascading Loops Scheme for Control
of Field-Oriented Induction Motors . . . . . . . . . . . ... 410
8.3.3 A Flatness-Based Control Approach for Induction
Motors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 414
8.3.4 Implementation of the EKF for the Nonlinear
Induction Motor Model . . . . . . . . . . . . . . . . . . . ... 415
8.3.5 Unscented Kalman Filtering for Induction
Motor Control . . . . . . . . . . . . . . . . . . . . . . . . . ... 416
8.4 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 418
8.5 Flatness-Based Adaptive Control of Electrostatic MEMS
Using Output Feedback . . . . . . . . . . . . . . . . . . . . . . . . . ... 422
8.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 422
8.5.2 Dynamic Model of the Electrostatic Actuator . . . . ... 423
8.5.3 Linearization of the MEMS Model Using
Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 425
xxiv Contents

8.5.4 Differential Flatness of the Electrostatic Actuator .... 427

8.5.5 Adaptive Fuzzy Control of the MEMS Model
Using Output Feedback . . . . . . . . . . . . . . . . . . .... 429
8.5.6 Lyapunov Stability Analysis. . . . . . . . . . . . . . . .... 434
8.5.7 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . .... 439

9 Differential Flatness Theory in Power Electronics. . . . . . . . . . . . . 443

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
9.2 Three-Phase Voltage Source Converters Control . . . . . . . . . . . 444
9.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 444
9.2.2 Linearization of the Converter’s Model Using
Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 446
9.2.3 Differential Flatness of the Voltage Source
Converter. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 449
9.2.4 Kalman Filter-Based Disturbance Observer
for the VSC Model . . . . . . . . . . . . . . . . . . . . . . . . . 453
9.2.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
9.3 Inverters Control. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
9.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 458
9.3.2 Dynamic Model of the Inverter. . . . . . . . . . . . . . . . . 459
9.3.3 Lie Algebra-Based Control of the Inverter’s Model . . . 463
9.3.4 Differential Flatness of the Inverter’s Model. . . . . . . . 466
9.3.5 Flatness-Based Control of the Inverter . . . . . . . . . . . . 468
9.3.6 State and Disturbances Estimation with Nonlinear
Kalman Filtering. . . . . . . . . . . . . . . . . . . . . . . . . . . 472
9.3.7 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 473
9.4 Distributed Inverters Synchronization . . . . . . . . . . . . . . . . . . 475
9.4.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 475
9.4.2 The Synchronization Problem for Parallel Inverters . . . 477
9.5 State and Disturbances Estimation of Parallel Inverters
with Nonlinear Kalman Filtering. . . . . . . . . . . . . . . . . . . . .. 482
9.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 483

10 Differential Flatness Theory for Internal Combustion Engines . . . 491

10.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
10.2 Flatness-Based Control of Valves in Marine Diesel Engines . . . 493
10.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
10.2.2 Dynamic Model of the Valve . . . . . . . . . . . . . . . . . . 494
10.2.3 Input–Output Linearization Using Lie Algebra . . . . . . 498
10.2.4 Input–Output Linearization Using Differential
Flatness Theory . . . . . . . . . . . . . . . . . . . . . . . . . .. 501
10.2.5 Disturbances Compensation with Derivative-Free
Nonlinear Kalman Filter . . . . . . . . . . . . . . . . . . . .. 504
10.2.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . .. 506
Contents xxv

10.3 Flatness-Based Control of Diesel Combustion Engines . ..... 511

10.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 511
10.3.2 Dynamic Model of the Turbocharged
Diesel Engine. . . . . . . . . . . . . . . . . . . . . . . . ..... 512
10.3.3 Nonlinear Control of the Diesel Engine Using
Lie Algebra . . . . . . . . . . . . . . . . . . . . . . . . . ..... 514
10.3.4 A Dynamic Extension-Based Feedback
Control Scheme . . . . . . . . . . . . . . . . . . . . . . ..... 517
10.3.5 Nonlinear Control of the Diesel Engine
Using Differential Flatness Theory . . . . . . . . . ..... 521
10.3.6 Disturbances Compensation Using
the Derivative-Free Nonlinear Kalman Filter . . . . . . . 525
10.3.7 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 527
10.4 Adaptive Control for Diesel Combustion Engines . . . . . . . . . . 528
10.4.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 528
10.4.2 Observer-Based Adaptive Fuzzy Control
for the Diesel Combustion Engine. . . . . . . . . . ..... 529
10.4.3 Application of Flatness-Based Adaptive Fuzzy
Control to the MIMO Diesel Engine Model . . . ..... 533
10.4.4 Lyapunov Stability Analysis. . . . . . . . . . . . . . ..... 538
10.4.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . ..... 543
10.5 Flatness-Based Control and Kalman Filtering
for the Spark-Ignited Engine . . . . . . . . . . . . . . . . . . . ..... 547
10.5.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 547
10.5.2 Dynamic Model of the SI Engine . . . . . . . . . . ..... 548
10.5.3 Feedback Linearizing Control of the SI Engine
Using Lie Algebra . . . . . . . . . . . . . . . . . . . . ..... 549
10.5.4 Feedback Linearizing Control of the SI Engine
Using Differential Flatness Theory . . . . . . . . . ..... 551
10.5.5 Compensation of Disturbances Using
the Derivative-Free Nonlinear Kalman Filter . . ..... 553
10.5.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . ..... 554
10.6 Flatness-Based Adaptive Fuzzy Control
of the Spark-Ignited Engine . . . . . . . . . . . . . . . . . . . . ..... 558
10.6.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . ..... 558
10.6.2 Flatness-Based Adaptive Fuzzy Control
for SI Motors . . . . . . . . . . . . . . . . . . . . . . . . ..... 559
10.6.3 Lyapunov Stability Analysis. . . . . . . . . . . . . . ..... 562
10.6.4 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . ..... 565
10.7 Flatness-Based Control and Kalman Filtering
of the Air–Fuel Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
10.7.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 566
10.8 Dynamic Model of the Air–Fuel Ratio System . . . . . . . . . . . . 567
10.8.1 The Air and Fuel Flow Models. . . . . . . . . . . . . . . . . 567
xxvi Contents

10.8.2 Dynamics of the Air–Fuel Ratio System . . . . . . . ... 569

10.8.3 Differential Flatness of the Air–Fuel Ratio System ... 570
10.8.4 Flatness-Based Control of the Air–Fuel
Ratio System . . . . . . . . . . . . . . . . . . . . . . . . . . ... 572
10.8.5 Compensation of Uncertainties with the
Derivative-Free Nonlinear Kalman Filter . . . . . . . ... 573
10.8.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 577

11 Differential Flatness Theory for Chaotic Dynamical Systems. . . . . 579

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 579
11.2 Flatness-Based Control of Chaotic Dynamical Systems . . . . . . 580
11.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
11.2.2 Differential Flatness of Chaotic Dynamical
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 581
11.2.3 Flatness-Based Adaptive Fuzzy Control
for Chaotic Systems . . . . . . . . . . . . . . . . . . . . . . . . 585
11.2.4 Design of the Feedback Controller . . . . . . . . . . . . . . 585
11.2.5 Approximators of Unknown System Dynamics . . . . . . 587
11.2.6 Lyapunov Stability Analysis. . . . . . . . . . . . . . . . . . . 588
11.2.7 Nonlinear Feedback Control of Chaotic Systems
Based on Fuzzy Local Linearization . . . . . . . . . . ... 591
11.2.8 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 593
11.3 Differential Flatness Theory for Chaos-Based
Communication Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
11.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 596
11.3.2 Structure of the Chaotic Communication System. . . . . 598
11.3.3 Differential Flatness Theory . . . . . . . . . . . . . . . . . . . 600
11.3.4 Estimation in Chaotic Modulators with Nonlinear
Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . . . . ... 601
11.3.5 Channel Equalization and Synchronization
Using Dual Kalman Filtering . . . . . . . . . . . . . . . ... 602
11.3.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 605

12 Differential Flatness Theory for Distributed

Parameter Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 613
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 613
12.2 Pointwise Flatness-Based Control of Distributed Parameter
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 615
12.2.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 615
12.2.2 Nonlinear 1D Wave-Type Partial Differential
Equations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 616
12.2.3 Sine-Gordon Nonlinear PDE in the Model
of the Josephson Junction . . . . . . . . . . . . . . . . . ... 617
Contents xxvii

12.2.4 Current Equation in a Josepshon

Transmission Line. . . . . . . . . . . . . . . . . . . . . . . . . . 618
12.2.5 State-Space Description of the Nonlinear
Wave Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 619
12.2.6 Solution of the Control and Estimation Problem
for Nonlinear Wave Dynamics . . . . . . . . . . . . . . . . . 622
12.2.7 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 625
12.3 Control of Heat Diffusion in Arc Welding Using Differential
Flatness Theory and Nonlinear Kalman Filtering. . . . . . . . . . . 627
12.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
12.4 Dynamic Model of the Arc Welding Process . . . . . . . . . . . . . 631
12.5 State-Space Description of the Nonlinear Heat Diffusion
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
12.6 Solution of the Control and Estimation Problem
for Nonlinear Heat Diffusion . . . . . . . . . . . . . . . . . . . . . . . . 635
12.6.1 Solution of the Control Problem . . . . . . . . . . . . . . . . 635
12.6.2 Solution of the Estimation Problem . . . . . . . . . . . . . . 637
12.7 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 639
12.8 Fault Detection and Isolation in Distributed Parameter
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
12.8.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
12.8.2 Estimation of Nonlinear Wave Dynamics. . . . . . . . . . 643
12.8.3 Equivalence Between Kalman Filters
and Regressor Models . . . . . . . . . . . . . . . . . . . . . . . 645
12.8.4 Change Detection with the Local Statistical
Approach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 646
12.8.5 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 651
12.9 Application to Condition Monitoring of Civil
and Mechanical Structures . . . . . . . . . . . . . . . . . . . . . . . . . . 656
12.9.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 656
12.9.2 Dynamical Model of the Building—Mechanical
Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 657
12.10 Differential Flatness of the Multi-DOF
Building’s Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 659
12.10.1 Damage Detection with the Use
of Statistical Criteria . . . . . . . . . . . . . . . . . . . . . . . . 662
12.10.2 Disturbances Estimation with the Derivative-Free
Nonlinear Kalman Filter . . . . . . . . . . . . . . . . . . . . . 664
12.10.3 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 666
xxviii Contents

13 Differential Flatness Theory in the Background of Other

Control Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 671
13.1 Differential Flatness Theory in the Background
of Backstepping Control . . . . . . . . . . . . . . . . . . . . . . . . ... 671
13.1.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 671
13.1.2 Flatness-Based Control Through Transformation
into the Canonical Form . . . . . . . . . . . . . . . . . . ... 673
13.1.3 A New Approach to Flatness-Based Control
for Nonlinear Dynamical Systems . . . . . . . . . . . . . . . 674
13.1.4 Closed-Loop Dynamics . . . . . . . . . . . . . . . . . . . . . . 677
13.1.5 Comparison to Backstepping Control. . . . . . . . . . . . . 679
13.1.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 680
13.2 Differential Flatness and Optimal Control . . . . . . . . . . . . . . . 686
13.3 Boundary Control of Nonlinear PDE Dynamics Using
Differential Flatness Theory . . . . . . . . . . . . . . . . . . . . . . ... 687
13.3.1 Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 687
13.3.2 Transformation of the PDE Model into a Set
of Nonlinear ODEs . . . . . . . . . . . . . . . . . . . . . . ... 688
13.3.3 Differential Flatness of the Nonlinear PDE Model. ... 691
13.3.4 Computation of a Boundary Conditions-Based
Feedback Control Law . . . . . . . . . . . . . . . . . . . ... 693
13.3.5 Closed-Loop Dynamics . . . . . . . . . . . . . . . . . . . ... 695
13.3.6 Simulation Tests . . . . . . . . . . . . . . . . . . . . . . . . ... 697

References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 701

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
Acronyms

AC Alternate Current
AGV Automatic Ground Vehicle
ARE Algebraic Riccati Equation
ARMA Auto Regressive Moving Average
ARMAX Auto Regressive Moving Average with noise
AUV Autonomous Underwater Vessel
BER Bit Error Rate
CRLB Cramer-Rao Lower Bound
DC Direct Current
DFIG Doubly-Fed Induction Generator
DFKF Derivative-Free nonlinear Kalman Filtering
DOF Degrees of Freedom
DPS Distributed Parameter Systems
EKF Extended Kalman Filtering
KF Kalman Filtering
MSE Mean Square Error
MMSE Minimum Mean Square Estimator
MIMO Multi-Input Multi-Output
ODE Ordinary Differential Equation
PDE Partial Differential Equation
PMSG Permanent Magnet Synchronous Generator
SI Spark Ignited
SISO Single-input Single Output
SNR Signal-to-Noise Ratio
UAV Unmanned Aerial Vehicle
UKF Unscented Kalman Filtering
USV Unmanned Surface Vessel
VSC Voltage Source Converter

xxix
Chapter 1
Nonlinear Dynamical Systems
and Global Linearizing Control Methods

1.1 Introduction

This chapter overviews the basics of systems theory which can be used in the
modeling of nonlinear dynamics. To understand the oscillatory behavior of nonlinear
dynamical systems, benchmark examples of oscillators are given. Moreover, the fol-
lowing properties are analyzed: phase diagram, isoclines, attractors, local stability,
bifurcations of fixed points, and chaos properties.
Next, the chapter outlines differential geometry and Lie algebra-based control
as a predecessor to differential flatness theory-based control. First, the differential
geometric approach and the Frobenius theorem are analyzed. Next, the concept of
input–output linearization is introduced and its association to transformation of nor-
mal forms is explained. Additionally, the concept of input-state linearization is pre-
sented and the stages of its implementation are explained. Necessary and sufficient
conditions for applying input-state linearization and input–output linearization are
provided.

1.2 Characteristics of the Dynamics of Nonlinear Systems

The main features characterizing the stability of nonlinear dynamical systems are
defined as follows [238, 555]:
1. Finite escape time: It is the finite time within which the state-vector of the nonlinear
system converges to infinity.
2. Multiple isolated equilibria: A linear system can have only one equilibrium to
which converges the state vector of the system in steady-state. A nonlinear system
can have more than one isolated equilibria (fixed points). Depending on the initial
state of the system, in steady-state the state vector of the system can converge to one
of these equilibria.

© Springer International Publishing Switzerland 2015 1

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_1
2 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

3. Limit cycles: For a linear system to exhibit oscillations it must have eigenvalues on
the imaginary axis. The amplitude of the oscillations depends on initial conditions.
In nonlinear systems one may have oscillations of constant amplitude and frequency,
which do not depend on initial conditions. These types of oscillations are known as
limit cycles.
4. Subharmonic, harmonic, and almost periodic oscillations: A stable linear system
under periodic input produces an output of the same frequency. A nonlinear system,
under periodic excitation can generate oscillations with frequencies that are several
times smaller (subharmonic) or multiples of the frequency of the input (harmonic).
It may also generate almost periodic oscillations with frequencies that are not nec-
essarily multiples of a basis frequency (almost periodic oscillations).
5. Chaos: A nonlinear system in steady-state can exhibit behavior that is not character-
ized as equilibrium, periodic oscillation, or almost periodic oscillation. This behavior
is characterized as chaos. As time advances the behavior of the system changes in a
random-like manner, and this depends on the initial conditions. Although the dynamic
system is deterministic it exhibits randomness in the way it evolves in time.
6. Multiple modes of behavior: It is possible for the same dynamical system to
exhibit simultaneously more than one of the aforementioned characteristics (1)–(5).
Thus, a system without external excitation may exhibit simultaneously more than
one limit cycles. A system receiving a periodic external input may exhibit harmonic
or subharmonic oscillations, or an even more complex behavior in steady state which
depends on the amplitude and frequency of the excitation.

Example 1: Oscillations of a pendulum (Fig. 1.1).

The equation of the rotational motion of the pendulum under friction is given as

ml 2 θ̈ = mgsin(θ )l − klv ⇒
ml 2 θ̈ = mgsin(θ )l − kl 2 θ̇ ⇒ (1.1)
ml θ̈ = mgsin(θ ) − kl θ̇

By defining the state variables x1 = θ and x2 = θ̇ one has the state-space description

ẋ1 = x2
(1.2)
ẋ2 = gl sin(x1 ) − k
m x2

To compute the equilibrium one has

ẋ1 = ẋ2 = 0⇒
(1.3)
x2 = 0 and x1 = nπ n = 0, ±1, ±2, · · ·

Indicative forms of oscillations for the pendulum are defined as follows:

1.2 Characteristics of the Dynamics of Nonlinear Systems 3

Fig. 1.1 Pendulum

performing oscillations

(a) The effect of friction is neglected, therefore one has

ẋ1 = x2
(1.4)
ẋ2 = gl sin(x1 )

(b) An external torque T is applied to the pendulum

ẋ1 = x2
(1.5)
ẋ2 = gl sin(x1 ) − mk x2 + 1
ml 2
T

Example 2: Tunnel diode circuit (Fig. 1.2)

The following equations hold:

1 d Vc (t)
Vc = Ic (s)⇒Ic (s) = sC Vc (s)⇒Ic (t) = C (1.6)
sC dt
Fig. 1.2 Tunnel diode
circuit
4 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

d I L (t)
VL (s) = sLI L (s)⇒VL (t) = L (1.7)
dt

I L (s) = IC + I R (1.8)

I R (s) = h(V R ) (1.9)

The following state variables are defined x1 = VC and x2 = i L . From Eq. (1.8) it
holds that
I L − IC − I R = 0⇒IC = I L − I R ⇒IC = x2 − h(x1 ) (1.10)

Moreover, it holds that

d IL
I L R − E + VL + VC = 0⇒E = VL + VC + i L R⇒E = L + VC + I L R⇒
dt
1 1
E = L ẋ2 + VC − x2 R⇒ẋ2 = [−VC − Rx2 + E]⇒ẋ2 = [−x1 − Rx2 + u]
L L
(1.11)
Additionally, from Eq. (1.6) it holds that

d Vc (t) 1
= Ic (t) (1.12)
dt C
By replacing Eq. (1.10) into Eq. (1.12) one obtains

d Vc (t) 1
= [x2 − h(x1 )] (1.13)
dt C
From Eqs. (1.11) and (1.13) one has the description of the system is state-space form
as
ẋ1 = C1 [x2 − h(x1 )]
(1.14)
ẋ2 = L1 [−x1 − Rx2 + u]

The associated equilibrium is computed from the condition ẋ1 = 0 and ẋ2 = 0 which
gives
−h(x1 ) + x2 = 0
(1.15)
−x1 − Rx2 + u = 0

and consequently
E−x1
−h(x1 ) + R =0
E−x1 (1.16)
x2 = R

Therefore, finally the equilibrium point is computed from the solution of the relation

E 1
h(x1 ) = − x1 (1.17)
R R
1.2 Characteristics of the Dynamics of Nonlinear Systems 5

Fig. 1.3 Spring-mass

system

Example 3: Spring-mass system (Fig. 1.3)

It holds that
m ẍ = F − F f − Fsp ⇒
(1.18)
m ẍ = F − b ẋ − kx

One can also consider a model with nonlinear spring dynamics given by

g(x) = k(1 − a 2 x 2 )x |ax| < 1 model of softening spring

(1.19)
g(x) = k(1 + a 2 x 2 )x x > xthr es model of hardening spring

The combination of a hardening spring, a linear viscous damping term, and a periodic
external force F = Acos(ωt) results in the Duffing oscillator

m ẍ + c ẋ + kx + ka 2 x 3 = Acos(ωt) (1.20)

The combination of a linear spring, a linear viscous damping term, a dry friction
term, and a zero external force generates the following oscillator model:

m ẍ + kx + c ẋ + η(x, ẋ) = 0 (1.21)

where
⎧
⎪
⎨ μk mg·sign(ẋ) if |ẋ > 0|
η(x, ẋ) = −kx if |ẋ| = 0 and |x|≤μs mg/k (1.22)
⎪
⎩
−μs mgsign(x) if ẋ = 0 and |x| > μs mg/k

By defining the state variables x1 = x and x2 = ẋ one has

ẋ1 = x2
(1.23)
ẋ2 = − mk x1 − mc x2 − m η(x, ẋ)
1
6 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

For x2 = ẋ > 0 one obtains the state-space description

ẋ1 = x2
(1.24)
ẋ2 = − mk x1 − mc x2 − μk g

For x2 = ẋ < 0 one obtains the state-space description

ẋ1 = x2
(1.25)
ẋ2 = − mk x1 − mc x2 + μk g

1.3 Computation of Isoclines

An autonomous second order system is described by two differential equations of

the form
ẋ1 = f 1 (x1 , x2 )
(1.26)
ẋ2 = f 2 (x1 , x2 )

The method of the isoclines consists of computing the slope (ratio) between f 2 and
f 1 for every point of the trajectory of the state vector (x1 , x2 ).

f 2 (x1 , x2 )
s(x) = (1.27)
f 1 (x1 , x2 )

The case s(x) = c describes a curve in the x1 − x2 plane along which the trajectories
ẋ1 = f 1 (x1 , x2 ) and ẋ2 = f 2 (x1 , x2 ) have a constant slope.
The curve s(x) = c is drawn in the x1 − x2 plane and along this curve one also
draws small linear segments of length c. The curve s(x) = c is known as an isocline.
The direction of these small linear segments is according to the sign of the ratio
f 2 (x1 , x2 )/ f 1 (x1 , x2 ).
Example 1:
The following simplified pendulum equation is considered:

ẋ1 = x2
(1.28)
ẋ2 = −sin(x1 )

The slope s(x) is given by the relation

f 2 (x1 , x2 ) sin(x2 )
s(x) = ⇒s(x) = − (1.29)
f 1 (x1 , x2 ) x2

Setting s(x) = c it holds that the isoclines are given by the relation
1.3 Computation of Isoclines 7

Fig. 1.4 Isoclines diagram 5

for s(x) = −sin(x
x2
1) c=−1/4
4 c=−1/3
c=−1/2
3 c=−1.0
c=1.0

2 c=1/2
c=1/3
c=1/4
1

x2
0

−1

−2

−3

−4

−5
−5 0 5
x1

1
x2 = − sin(x1 ) (1.30)
c
For different values of c one has the following isoclines diagram depicted in Fig. 1.4.
Example 2:
The following oscillator model is considered, being free of friction and with state-
space equations:
ẋ1 = x2
(1.31)
ẋ2 = −0.5x2 − sin(x1 )

To compute isoclines one has

s(x) = −0.5x2x−sin(x 1)
= c⇒ − 0.5x2 − sin(x1 ) = cx2 ⇒
2 (1.32)
(0.5 + c)x2 = sin(x1 )⇒x2 = − 0.5+c1
sin(x1 )

For different values of parameter c the isoclines are depicted in Fig. 1.5.
8 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Fig. 1.5 Isoclines diagram 5

for s(x) = −0.5x2x−sin(x
2
1) c1=−3/4
4 c2=−5/6
c3=−1
3 c4=−3/2
c5=1/2

2 c6=0
c7=−1/6
c8=−1/4
1

x2
0

−1

−2

−3

−4

−5
−5 0 5
x1

1.4 Basic Features in the Study of Nonlinear Dynamics

The basic features that are important in the study of nonlinear dynamics are (i)
equilibria (fixed points), (ii) limit cycles, (ii) phase diagrams, (iv) periodic orbits, and
(v) bifurcations of fixed points [238, 555, 584, 605]. The definition of these features
is given through examples in the case of nonlinear dynamical systems models.

1.4.1 The Phase Diagram

One can consider the nonlinear model with the two state variables, V and η. The
dynamics of the nonlinear model can be written as

dV
dt = f (V, η)
dη (1.33)
dt = g(V, η)

The phase diagram consists of the points on the trajectories of the solution of the
associated differential equation, i.e., (V (tk ), η(tk )).
At a fixed point or equilibrium it holds that f (V R , η R ) = 0 and g(V R , η R ) = 0.
The closed trajectories are associated with periodic solutions. If there are closed
trajectories the ∃T > 0 such that (V (tk ), η(tk )) = (V (tk + T ), η(tk + T )).
1.4 Basic Features in the Study of Nonlinear Dynamics 9

Another useful parameter is the nullclines. The V -nullcline is characterized by

the relation V̇ = f (V, η) = 0. The η-nullcline is characterized by the relation
η̇ = g(V, η) = 0. The fixed points (or equilibria) are found on the intersection of
nullclines.

1.4.2 Stability Analysis of Nonlinear Systems

1.4.2.1 Local Linearization

A manner to examine stability in nonlinear dynamical systems is to perform local

linearization around an equilibrium. Assume the nonlinear system ẋ = f (x, u) and
f (x0 , u) = 0, that is, x0 is the local equilibrium. The linearization of f (x, u) with
respect to u around x0 (Taylor series expansion) gives the equivalent description

ẋ = Ax + Bu (1.34)

where ⎛ ∂ f1 ∂ f1 ⎞
∂x ∂x· · · ∂∂ xfn1
⎜ ∂ f12 ∂ f22 ⎟
· · · ∂∂ xfn2 ⎟
⎜ ∂ x1 ∂ x2
A = ∇x f = ⎜ ⎟ |x=x0 (1.35)
⎝··· ··· ··· ···⎠
∂ fn ∂ fn ∂ fn
∂ x1 ∂ x2 · · · ∂ xn

and ⎛ ∂ f1 ∂ f1 ∂ f1 ⎞
∂u ∂u· · · ∂u
⎜ ∂ f21 ∂ f22∂ f2 ⎟
n
· · · ∂u
⎜ ∂u 1 ∂u 2 ⎟
B = ∇u f = ⎜ n ⎟ | x=x (1.36)
⎝··· ··· ··· ···⎠ 0

∂ fn ∂ fn ∂ fn
∂u 1 ∂u 2 · · · ∂u n

The eigenvalues of matrix A define the local stability features of the system:
Example 1:
Assume the nonlinear system

d2x dx
2
+ 2x + 2x 2 − 4x = 0 (1.37)
dt dt

By defining the state variables x1 = x, x2 = ẋ the system can be written in the

following form:
ẋ1 = x2
(1.38)
ẋ2 = −2x1 x2 − 2x12 + 4x1

It holds that ẋ = 0 if f (x) = 0 that is (x1∗,1 , x2∗,1 ) = (0, 0) and (x1∗,2 , x2∗,2 ) = (2, 0).
Round the first equilibrium (x1∗,1 , x2∗,1 ) = (0, 0) the system’s dynamics is written as
10 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods
   
01 ẋ1 01 x1
A = ∇x f = or = (1.39)
40 ẋ2 40 x2

The eigenvalues of the system are λ1 = 2 and λ2 = −2. This means that the fixed
point (x1∗,1 , x2∗,1 ) = (0, 0) is an unstable one.
Next, the fixed point (x1∗,2 , x2∗,2 ) = (2, 0) is analyzed. The associated Jacobian
matrix is computed again. It holds that

0 1
A = ∇x f = (1.40)
−4 −4

The eigenvalues of the system are λ1 = −2 and λ2 = −2. Consequently, the fixed
point (x1∗,2 , x2∗,2 ) = (2, 0) is a stable one.

1.4.2.2 Lyapunov Stability Approach

The Lyapunov method analyzes the stability of a dynamical system without the need
to compute explicitly the trajectories of the state vector x = [x1 , x2 , . . . , xn ]T .
Theorem: The system described by the relation ẋ = f (x) is asymptotically stable in
the vicinity of the equilibrium x0 = 0 if there is a function V (x) such that
(i) V (x) to be continuous and to have a continuous first order derivative at x0
(ii) V (x) > 0 if x=0 and V (0) = 0
(iii) V̇ (x) < 0, ∀x=0.
The Lyapunov function is usually chosen to be a quadratic (and thus positive) energy
function of the system, however, there in no systematic method to define it.
Assume now that ẋ = f (x) and x0 = 0 is the equilibrium. Then the system is
globally asymptotically stable if for every ε > 0, ∃δ(ε) > 0, such that if ||x(0)|| < δ
then ||x(t)|| < ε, ∀t≥0.
This means that if the state vector of the system starts in a disk of radius δ then
as time advances it will remain at the same disk, as shown in Fig. 1.6. Moreover, if
lim t→∞ ||x(t)|| = x0 = 0, then the system is globally asymptotically stable.
Example 1: Consider the system

ẋ1 = x2
(1.41)
ẋ2 = −x1 − x32

The following Lyapunov function is defined:

V (x) = x12 + x22 (1.42)

The equilibrium point is (x1 = 0, x2 = 0). It holds that V (x) > 0∀(x1 , x2 )=
(0, 0) and V (x) = 0 for (x1 , x2 ) = (0, 0). Moreover, it holds that
1.4 Basic Features in the Study of Nonlinear Dynamics 11

Fig. 1.6 Global stability and global asymptotic stability

V̇ (x) = 2x1 ẋ1 + 2x2 ẋ2 = 2x1 x2 + 2x2 (−x1 − x23 )⇒

(1.43)
V̇ (x) = −2x24 < 0 ∀ (x1 , x2 )=(0, 0)

Therefore, the system is asymptotically stable and lim t→∞ (x1 , x2 ) = (0, 0).
Example 2: Consider the system

ẋ1 = −x1 (1 + 2x1 x22 )

(1.44)
ẋ2 = x13 x2

The following Lyapunov function is considered:

1 2
V (x) = x + x22 (1.45)
2 1
The equilibrium point is x1 = 0, x2 = 0. It holds that

V̇ (x) = x1 ẋ1 + 2x2 ẋ2 = −x12 (1 + 2x1 x22 + 2x2 (x13 x2 ))⇒
(1.46)
V̇ (x) = −x12 < 0 ∀ (x1 , x2 )=(0, 0)

Therefore, the system is asymptotically stable and lim t→∞ (x1 , x2 ) = (0, 0).

1.4.3 Stability Analysis of Nonlinear Models

Local stability of the previously described nonlinear model can be studied around the
associated equilibria. Local linearization can be performed around equilibria. Using
the set of differential equations ẋ = h(x) that describe the nonlinear system and
performing Taylor series expansion, that is, ẋ = h(x)⇒ẋ = h(x0 )|x0 + ∇x h(x −
x0 ) + · · · .
12 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Consider for instance a nonlinear model (Duffing oscillator) under no external

force  
ẋ1 f (x1 , x2 )
= (1.47)
ẋ2 g(x1 , x2 )

kα 2 x 3
where f (x1 , x2 ) = x2 and g(x1 , x2 ) = − mc x2 − mk x1 − m 1 . The fixed points of
this model are computed from the condition ẋ1 = 0 and ẋ2 = 0, and are found to be
(x1∗ , x2∗ ) = (0, 0). The Jacobian matrix ∇x h = M is given as
 
∂f ∂f
J= ∂ x1 ∂ x2 (1.48)
∂g ∂g
∂ x1 ∂ x2

which results in the matrix

 
0 1
J= 3kα 2 x12 (1.49)
− mk − m − mc

At the fixed point (x1∗ , x2∗ ) = (0, 0) one obtains det (λI − J ) = λ2 + mc λ + mk which
for c2 < 4 km has imaginary roots with negative real part. Consequently, the system
is locally stable at the operating point and exhibits attenuating oscillations.
The eigenvalues of matrix M define stability around fixed points (stable or unstable
fixed point). To this end, one has to find the roots of the associated characteristic
polynomial which is given by det (λI − J ) = 0 where I is the identity matrix.

1.5 Phase Diagrams and Equilibria of Nonlinear Models

1.5.1 Phase Diagrams for Linear Dynamical Systems

The following autonomous linear system is considered:

ẋ = Ax (1.50)

The eigenvalues of matrix A define the system dynamics. Some terminology asso-
ciated with fixed points is as follows:
A fixed point for the system of Eq. (1.50) is called hyperbolic if none of the
eigenvalues of matrix A has zero real part. A hyperbolic fixed point is called a saddle
if some of the eigenvalues of matrix A have real parts greater than zero and the
rest of the eigenvalues have real parts less than zero. If all of the eigenvalues have
negative real parts then the hyperbolic fixed point is called a stable node or sink. If
all of the eigenvalues have positive real parts then the hyperbolic fixed point is called
1.5 Phase Diagrams and Equilibria of Nonlinear Models 13

Fig. 1.7 Phase diagram of 10

initial state variables x1 , x2
of a second order linear 8
autonomous system with
negative eigenvalues, where 6
λ1 < λ2 < 0
4

x2
0

−2

−4

−6

−8

−10
−15 −10 −5 0 5 10 15
x1

an unstable node or source. If the eigenvalues are purely imaginary then one has an
elliptic fixed point which is said to be a center.
Case 1: Both eigenvalues of matrix A are real and unequal, that is, λ1 =λ1 =0. For
λ1 < 0 and λ2 < 0 the phase diagram for z 1 and z 2 is shown in Fig. 1.7.
In case λ2 is smaller than λ1 the term eλ2 t decays faster than eλ1 t .
For λ1 > 0 > λ2 the phase diagram of Fig. 1.8 is obtained:
In the latter case there are stable trajectories along eigenvector v1 and unstable
trajectories along eigenvector v2 of matrix A. The stability point (0, 0) is said to be
a saddle point.
When λ1 > λ2 > 0 then one has the phase diagrams of Fig. 1.9:
Case 2: Complex eigenvalues:
Typical phase diagrams in the case of stable complex eigenvalues are given in
Fig. 1.10.
Typical phase diagrams in the case of unstable complex eigenvalues are given in
Fig. 1.11.
Typical phase diagrams in the case of imaginary eigenvalues are given in Fig. 1.12.
Case 3: Matrix A has nonzero eigenvalues which are equal to each other. The asso-
ciated phase diagram is given in Fig. 1.13.
14 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Fig. 1.8 Phase diagram of 2

initial state variables x1 , x2
of a second order linear
1.5
autonomous system with real
eigenvalues, where
λ1 > 0 > λ2 1

0.5

x2
0

−0.5

−1

−1.5

−2
−30 −20 −10 0 10 20 30
x1

Fig. 1.9 Phase diagram of 3

state variables x1 , x2 of a
second order linear
autonomous system with real 2
eigenvalues λ1 > λ2 > 0

1
x2

−1

−2

−3
−30 −20 −10 0 10 20 30
x1
1.5 Phase Diagrams and Equilibria of Nonlinear Models 15

Fig. 1.10 Phase diagram of 15

state variables x1 , x2 of a
second order linear
autonomous system with 10
complex stable eigenvalues

x2
0

−5

−10

−15
−10 −5 0 5 10
x1

Fig. 1.11 Phase diagram of 10

x 10
state variables x1 , x2 of a 1
second order linear
autonomous system with 0.8
complex unstable
eigenvalues 0.6

0.4

0.2
x2

−0.2

−0.4

−0.6

−0.8

−1
−8 −6 −4 −2 0 2 4 6 8
x1 x 10
9
16 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Fig. 1.12 Phase diagram of 25

state variables x1 , x2 of a 2nd
order linear autonomous 20
system with imaginary
eigenvalues 15

10

x2
0

−5

−10

−15

−20

−25
−15 −10 −5 0 5 10 15
x1

Fig. 1.13 Phase diagram of 10

state variables x1 , x2 of a
second order linear 8
autonomous system with
identical stable eigenvalues 6

2
x2

−2

−4

−6

−8

−10
−15 −10 −5 0 5 10 15
x1
1.5 Phase Diagrams and Equilibria of Nonlinear Models 17

1.5.2 Multiple Equilibria for Nonlinear Dynamical Systems

A nonlinear system can have multiple equilibria as shown in the following example.
Consider for instance the model of a pendulum under friction

ẋ1 = x2
(1.51)
ẋ2 = − gl sin(x1 ) − K
m x2

The associated phase diagram is designed for different initial conditions and is given
in Fig. 1.14.
For the previous model of the nonlinear oscillator, local linearization around
equilibria with the use of Taylor series expansion enables analysis of the local stability
properties of nonlinear dynamical systems

ẋ1 = f 1 (x1 , x2 )
(1.52)
ẋ2 = f 2 (x1 , x2 )

which gives

ẋ1 = f 1 ( p1 , p2 ) + α11 (x1 − p1 ) + α12 (x2 − p2 ) + h.o.t

(1.53)
ẋ2 = f 2 ( p1 , p2 ) + α21 (x1 − p1 ) + α22 (x2 − p2 ) + h.o.t

Fig. 1.14 Phase diagram of 6

state variables x1 , x2 of a
second order nonlinear
oscillator that exhibits 4
multiple equilibria

2
x2

−2

−4

−6
−30 −20 −10 0 10 20 30
x1
18 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

where
∂ f 1 (x1 ,x2 ) ∂ f 1 (x1 ,x2 )
α11 = ∂ x1 |x1 = p1 ,x2 = p2 α12 = ∂ x2 |x1 = p1 ,x2 = p2
∂ f 2 (x1 ,x2 ) ∂ f 2 (x1 ,x2 ) (1.54)
α21 = ∂ x1 |x1 = p1 ,x2 = p2 α22 = ∂ x2 |x1 = p1 ,x2 = p2

For the equilibrium it holds that

f 1 ( p1 , p2 ) = 0
(1.55)
f 2 ( p1 , p2 ) = 0

Next, by defining the new variables y1 = x1 − p1 and y2 = x2 − p2 one can rewrite

the state-space equation as
ẏ1 = ẋ1 = α11 y1 + α12 y2 + h.o.t.
(1.56)
ẏ2 = ẋ2 = α21 y1 + α22 y2 + h.o.t.

By omitting the higher order terms one can approximate the initial nonlinear system
with its linearized equivalent
ẏ1 = α11 y1 + α12 y2
(1.57)
ẏ2 = α21 y1 + α22 y2

which in matrix form is written as

ẏ = Ay (1.58)

where   
∂ f1 ∂ f1
α11 α12 ∂ x1 ∂ x2 ∂f
A= = ∂ f2 ∂ f2 |x= p = ∂ x |x= p (1.59)
α21 α22 ∂ x1 ∂ x2

Matrix A = ∂∂ xf is the Jacobian matrix of the system computed at point (x1 , x2 ) =

( p1 , p2 ). It is anticipated that the trajectories of the phase diagram of the linearized
system in the vicinity of the equilibrium point will be also close to the trajectories
of the phase diagram of the nonlinear system.
Therefore, if the origin (equilibrium) in the phase diagram of the linearized system
is (i) a stable node (matrix A has stable real eigenvalues), (ii) a stable focus (matrix A
has stable complex eignevalues), (iii) a saddle point (matrix A has some eigenvalues
with negative real part while the rest of the eigenvalues have positive real part),
then one concludes that the same properties hold for phase diagram of the nonlinear
system.
Example: A nonlinear oscillator of the following form is considered:

ẋ = f (x)⇒
  (1.60)
ẋ1 x2
=
ẋ2 −sin(x1 ) − 0.5x2
1.5 Phase Diagrams and Equilibria of Nonlinear Models 19

The associated Jacobian matrix is


∂f 0 1
∂x = (1.61)
−cos(x1 ) −0.5

There are two equilibrium points (0, 0) and (π, 0). The linearization around the
equilibria gives
 
0 1 0 1
A1 = A2 =
−1 −0.5 1 −0.5
(1.62)
with eigenvalues with eigenvalues

λ1,2 = −0.25± j0.97 λ1 = −1.28 λ2 = 0.78

Consequently, the equilibrium (0, 0) is a stable focus (matrix A1 has stable complex
eignevalues) and the equilibrium (π, 0) is a saddle point (matrix A2 has an unstable
eigenvalue).

1.5.3 Limit Cycles

A dynamical system is considered to exhibit limit cycles when it admits the nontrivial
periodic solution
x(t + T ) = x(t) ∀ t≥0 (1.63)

for some T > 0 (the trivial periodic solution is the one associated with x(t) =
constant). An example about the existence of limit cycles is examined in the case of
the Van der Pol oscillator. The state equations of the oscillator are

ẋ1 = x2
(1.64)
ẋ2 = −x1 + ε(1 − x!2 )x2

Next the phase diagram of the Van der Pol oscillator is designed for three different
values of parameter ε, namely ε = 0.2, ε = 1 and ε = 5.0. In Figs. 1.15 and 1.16,
it can be observed that in all cases there is a closed trajectory to which converge all
curves of the phase diagram that start from points far from it.
To study conditions under which dynamical systems exhibit limit cycles, a second
order autonomous nonlinear system is considered next, given by
 
ẋ1 f 1 (x1 , x2 )
= (1.65)
ẋ2 f 2 (x1 , x2 )
20 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Fig. 1.15 Phase diagram of 2.5

the Van der Pol oscillator for
ε = 0.2 2

1.5

0.5

x2
0

−0.5

−1

−1.5

−2

−2.5
−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5
x1

Fig. 1.16 Phase diagram of 3

the Van der Pol oscillator for
ε=1
2

1
x2

−1

−2

−3
−2.5 −2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5
x1
1.5 Phase Diagrams and Equilibria of Nonlinear Models 21

Next, the following theorem defines the appearance of limit cycles in the phase
diagram of dynamical systems [238, 555].
Theorem 1 (Poincaré-Bendixson): If a trajectory of the nonlinear system of Eq. (1.65)
remains in a finite region Ω, then one of the following is true: (i) the trajectory goes
to an equilibrium point, (ii) the trajectory tends to a limit cycle, (iii) the trajectory is
itself a limit cycle.
Moreover, the following theorem provides a sufficient condition for the nonexis-
tence of limit cycles:
Theorem 2 (Bendixson): For the nonlinear system of Eq. (1.65), no limit cycle can
exist in a region Ω of the phase plane in which ∂∂ xf11 + ∂∂ xf22 does not vanish and does
not change sign.

1.6 Bifurcations in Nonlinear Dynamics

1.6.1 Bifurcations of Fixed Points of Nonlinear Models

As the parameters of the nonlinear model of the system are changed, the stability of
the equilibrium point can also change and also the number of equilibria may vary.
Values of these parameters at which the locus of the equilibrium (as a function of the
parameter) changes and different branches appear are known as critical or bifurca-
tion values. The phenomenon of bifurcation has to do with quantitative changes in
parameters leading to qualitative changes in the system’s properties [94, 137, 194,
258, 287].
Another issue in bifurcation analysis has to do with the study of the segments of the
bifurcation branches in which the fixed points are no longer stable but either become
unstable or are associated with limit cycles. The latter case is called Hopf bifurcation
and the system’s Jacobian matrix has a pair of complex imaginary eigenvalues.

1.6.2 Saddle-Node Bifurcations of Fixed Points

in a One-Dimensional System

The considered dynamical system is given by ẋ = μ − x 2 . The fixed points of the

√
system result from the condition ẋ = 0 which for μ > 0 gives x ∗ = ± μ. The first
√ √
fixed point x = μ is a stable one, whereas the second fixed point x = − μ is an
unstable one. The phase diagram of the system is given in Fig. 1.17. Since there is
one stable and one unstable fixed point the associated bifurcation (locus of the fixed
points in the phase plane) will be a saddle-node one.
The bifurcations diagram is given next. The diagram shows how the fixed points
of the dynamical system vary with respect to the values of parameter μ. In the above
case it represents a parabola in the μ − x plane as shown in Fig. 1.18.
22 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Fig. 1.17 Phase diagram

and fixed points of the
system ẋ = μ − x 2 .
Converging arrows denote a
stable fixed point

Fig. 1.18 Saddle-node

bifurcation diagram

√
For μ > 0 the dynamical system has two fixed points located at ± μ. One fixed
point is stable and is associated with the upper branch of the parabola. The other
fixed point is unstable and is associated with the lower branch of the parabola. The
value μ = 0 is considered to be a bifurcation value and the point (x, μ) = (0, 0) is a
bifurcation point. This particular type of bifurcation where one branch is associated
with fixed points and the other branch is not associated to any fixed points is known
as saddle-node bifurcation.

1.6.3 Pitchfork Bifurcation of Fixed Points

In pitchfork bifurcations the number of fixed points varies with respect to the values
of the bifurcation parameter. The dynamical system ẋ = x(μ − x 2 ) is considered.
1.6 Bifurcations in Nonlinear Dynamics 23

Fig. 1.19 Phase diagram

and fixed points of a system
exhibiting pitchfork
bifurcations. Converging
arrows denote a stable fixed
point

Fig. 1.20 Pitchfork

bifurcation diagram

The associated fixed points are found by the condition ẋ = 0. For μ < 0 there is
one fixed point at zero which is stable. For μ = 0 there is still one fixed point at
zero which is still stable. For μ > 0 there are three fixed points, one at x = 0, one at
√ √
x = + μ which is stable and one at x = − μ which is also stable. The associated
phase diagrams and fixed points are presented in Fig. 1.19.
The bifurcations diagram is given next. The diagram shows how the fixed points
of the dynamical system vary with respect to the values of parameter μ. In the above
case, it represents a parabola in the μ − x plane as shown in Fig. 1.20.
24 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

1.6.4 The Hopf Bifurcation

Bifurcation of the equilibrium point means that the locus of the equilibrium on the
phase plane changes due to variation in the system’s parameters. Equilibrium x ∗ is
a hyperbolic equilibrium point if the real parts of all its eigenvalues are nonzero. A
Hopf bifurcation appears when the hyperbolicity of the equilibrium point is lost due
to variation of the system parameters and the eigenvalues become purely imaginary.
By changing the values of the parameters at a Hopf bifurcation, an oscillatory solution
appears [373, 576].
The stages for finding Hopf bifurcations in nonlinear dynamical systems are
described next. The following autonomous differential equation is considered:

dx
= f 1 (x, λ) (1.66)
dt

where x is the state vector x∈R n and λ∈R m is the vector of the system parameters. In
Eq. (1.66) a point x ∗ satisfying f 1 (x ∗ ) = 0 is an equilibrium point. Therefore, from
the condition f 1 (x ∗ ) = 0 one obtains a set of equations which provide the equilibrium
point as function of the bifurcating parameter. The stability of the equilibrium point
can be evaluated by linearizing the system’s dynamic model around the equilibrium
point and by computing eigenvalues of the Jacobian matrix. The Jacobian matrix at
the equilibrium point can be written as

∂ f 1 (x)
J f1 (x ∗ ) = |x=x ∗ (1.67)
∂x
and the determinant of the Jacobian matrix provides the characteristic equation which
is given as

det (μi In − J f1 (x ∗ ) ) = λn1 + α1 λin−1 + · · · + αn−1 λi + αn = 0 (1.68)

where In is the n×n identity matrix, μi with i = 1, 2, . . . , n denotes the eigenvalues

of the Jacobian matrix D f 1 (x ∗ ). From the requirement the eigenvalues of the sys-
tem to be purely imaginary one obtains a condition, i.e. values that the bifurcating
parameter should take, for the appearance of Hopf bifurcations.
As example, the following nonlinear system is considered:

ẋ1 = x2
(1.69)
ẋ2 = −x1 + (m − x12 )x1

Setting ẋ1 = 0 and ẋ2 = 0 one obtains the system’s fixed points. For m≤1 the
system has the fixed point (x1∗ , x2∗ ) = (0, 0). For m > 1 the system has the fixed
√ √
points (x1∗ , x2∗ ) = (0, 0), (x1∗ , x2∗ ) = ( m − 1, 0), (x1∗ , x2∗ ) = (− m − 1, 0). The
system’s Jacobian is
1.6 Bifurcations in Nonlinear Dynamics 25

Fig. 1.21 Phase diagram 2

and fixed points of a system
exhibiting Hopf bifurcations
1.5

0.5

x2
0

−0.5

−1

−1.5

−2
−2 −1.5 −1 −0.5 0 0.5 1 1.5 2 2.5
x1

Fig. 1.22 Hopf bifurcation 4

diagram
3

0
x

−1

−2

−3

−4
1 2 3 4 5 6 7 8 9 10 11
m
26 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

0 1
J= (1.70)
−1 + m − 3x12 0

If m≤1, the eigenvalues of the Jacobian at the fixed point (x1∗ , x2∗ ) = (0, 0) will
be imaginary. Thus the system exhibits Hopf bifurcations, as shown in Fig. 1.21. If
m > 1 then the fixed point (x1∗ , x2∗ ) = (0, 0) is an unstable one. On the other hand,
√
if m > 1 the eigenvalues of the Jacobian at the fixed points (x1∗ , x2∗ ) = ( m − 1, 0)
√
and (x1∗ , x2∗ ) = (− m − 1, 0) will be imaginary and the system exhibits again Hopf
bifurcations.
The bifurcations diagram is given next. The diagram shows how the fixed points
of the dynamical system vary with respect to the values of parameter μ. In the above
case it represents a parabola in the μ − x plane as shown in Fig. 1.22.

1.7 Predecessors of Differential Flatness Theory

1.7.1 The Differential Geometric Approach

Before proceeding to the analysis of the differential geometric approach for the
control of nonlinear systems the following definitions are given [216, 238]:
Vector Field: A mapping f : D→R n where D⊂R n is a domain, is said to be a vector
field on D. A vector field is a n-dimensional column.
Covector Field: The transpose of a vector field is said to be a covector field. A
covector field is an n-dimensional row.
Inner Product: If f and w are, respectively, a vector field and a covector field on D,
then the inner product < w, f > is defined as


n
< w, f >= w(x) f (x) = wi (x) f i (x) (1.71)
i=1

Differential: Let h : D→R. The differential of h is a covector field defined as

∂h ∂h ∂h ∂h
dh = =[ , ,..., ] (1.72)
∂x ∂ x1 ∂ x1 ∂ xn

Lie Derivatives: Let h : D→R and f : D→R n . The Lie derivative of h, with respect
to f or along f , is denoted as L f h and is defined as

∂h
Lfh = f (x) (1.73)
∂x
This is the meaning of computing the derivatives of h along the trajectories of the
system ẋ = f (x). The previous notation can be generalized towards the computation
1.7 Predecessors of Differential Flatness Theory 27

of Lie derivatives of higher order with respect to the same vector field, or with respect
to a new vector field. Thus, one has

∂L f h
L g L f h(x) = g(x) (1.74)
∂x
while it also holds

L 0f h(x) = h(x)
∂h n
L f h(x) = ∂h
∂ x1 f 1 +
1 ∂h 2
∂ x2 f 2 + ··· + ∂ xn f n
··· (1.75)
∂ L k−1
L kf h(x) = L f L k−1
f = ∂x
f
f (x)

Using the Lie derivative notation one can arrive at the definition of the relative degree
for nonlinear dynamical systems. Consider for example the following single-input
single-output nonlinear system

ẋ = f (x) + g(x)u
(1.76)
y = h(x)

has relative degree r in a domain D0 if

L g L if h(x) = 0 for 0≤i≤r − 2 and L g L rf−1 h(x)=0 ∀ x∈D0 (1.77)

Lie Bracket: Let f and g be two vector fields on D⊂R n . The Lie Bracket of f and
g is written as [ f, g] and is a third vector field which is defined as

∂g ∂f
[ f, g](x) = f − g (1.78)
∂x ∂x

where ∂∂gx and ∂∂ xf are Jacobian matrices. The definition of the Lie Brackets can be
repetitively used as follows:

ad 0f g(x) = g(x)
ad f g(x) = [ f, g](x)
(1.79)
···
ad kf g(x) = [ f, ad k−1
f g](x)

For Lie Brackets the following properties hold:

Bilinear: Let f 1 , f 2 and g1 , g2 be vector fields and r1 , r2 be real numbers. Then,

[r1 f 1 + r2 f 2 , g1 ] = r1 [ f 1 , g1 ] + r2 [ f 2 , g2 ]
(1.80)
[ f 1 , r1 g1 + r2 g2 ] = r1 [ f 1 , g1 ] + r2 [ f 1 , g2 ]
28 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Skew commutative: [ f, g] = −[g, f ]

Jacobi identity: If f and g are vector fields and h is a real-valued function, then

L [ f,g] h(x) = L f L g h(x) = L g L f h(x) (1.81)

Diffeomorphism: A mapping T : D→R n is a diffeomorphism on D if (i) it is

invertible on D, that is there exists a function T −1 (x) such that T −1 (T (x)) = x
for all x∈D and (ii) both T (x) and T −1 (x) are continuously differentiable. If the
Jacobian matrix [ ∂∂Tx ] is nonsingular at a point x0 ∈D, then from the inverse function
theorem it is concluded that there is a neighborhood N of x0 , such that T restricted
to N is a diffeomorphism on N , T is said to be a global diffeomorphism if it is a
diffeomorphism on R n and T (R n ) = R n . T is a global diffeomorphism if and only if
(i) [∂ T /∂ x] is nonsingular for all x∈R n
(ii) T is proper, that is lim ||x||→∞ ||T (x)|| = ∞
Distribution: Let f 1 , f 2 , . . ., f k be vector fields on D⊂R n . At any fixed point x∈D,
f 1 (x), f 2 (x), . . ., f k (x) are vectors in R n and

Δ(x) = span{ f 1 (x), f 2 (x), . . . , f n (x)} (1.82)

is a subspace on R n . To each point x∈R n a subspace Δ(x) is assigned. One can refer
to this assignment using
Δ = span( f 1 , f 2 , . . . , f n ) (1.83)

and this is called a distribution.This implies that, Δ is the collection of all vector
spaces Δ(x) for xin D. The dimension of Δ(x) which is denoted as

dim(Δ(x)) = rank[ f 1 (x), f 2 (x), . . . , f k (x)] (1.84)

may vary with x. However, when Δ = span{ f 1 , f 2 , . . . , fr }, where { f 1 (x),

. . . , fr (x)} are linearly independent for all x∈D, then dim(Δ(x)) = r for all x∈D. In
that case Δ is said to be a nonsingular distribution on D, generated by f 1 , f 2 , . . . , fr .
It follows from the smoothness of the vector fields that every y∈Δ can be expressed as


r
g(x) = ci (x) f i (x) (1.85)
i=1

where ci (x) are smooth functions defined on D.

Involutive distribution: A distribution Δ is involutive if

g1 ∈Δ and g2 ∈Δ ⇒[g1 , g2 ] ∈ Δ (1.86)

If Δ is a nonsingular distribution on Δ, generated by f 1 , f 2 , . . . , fr then it can be

verified that Δ is an involutive distribution if and only if
1.7 Predecessors of Differential Flatness Theory 29

[ f i , f j ]∈Δ ∀ 1≤i, j≤r (1.87)

or equivalently

r
[ f i , f j ](x) = αi jk f k (x) ∀ i, j (1.88)
k=1

Equivalently, involutivity means that if one forms the Lie bracket of any pairs of vector
fields from the set [ f 1 , f 2 , . . . , f m ], the produced vector field can be expressed as a
linear combination of the original set of vector fields. It is noted that checking if a
set of vector fields [ f 1 , . . . , f m ] is involutive is equivalent to checking if

rank( f 1 (x), . . . , f m (x)) = rank( f 1 (x), . . . , f m (x), [ f i , f j ](x)) (1.89)

for all x and all i, j.

Codistribution: Codistributions are dual objects to distributions. They are defined
using covector fields and possess properties in analogy to those of distributions. A
particular codistribution is the one that annihilates a distribution Δ, which is denoted
as Δ⊥ , and which is defined as

Δ⊥ (x) = {w∈(R n )∗ | < w, v >= 0, ∀ v∈Δ(x)} (1.90)

where (R n )∗ is the n-dimensional space of new vectors.

Complete integrability: Let Δ be a nonsingular distribution on D, generated by
f 1 , . . . , f n . Then Δ is completely integrable, if for each x0 ∈D, there exists a neigh-
borhood N of x0 and n − r real-valued smooth functions h 1 (x), . . . , h n−r (x), such
that h 1 (x), . . . , h n−r (x) satisfy the partial differential equations
∂h j
f i (x) = 0, ∀ 1≤i≤r and 1≤ j≤n − r (1.91)
∂x

and the covector fields dh j (x) are linearly independent for all x∈D. Equivalently,
one can write
Δ⊥ = span{dh 1 , . . . , dh n−r } (1.92)

Finally, a key result in differential geometry is Frobenius’ theorem which states that:
A nonsingular distribution is completely integrable if it is involutive.

1.7.2 Elaboration on the Frobenius Theorem

The Frobenius theorem is important for the formal treatment of feedback linearization
of n-order nonlinear systems. It provides a necessary and sufficient condition for the
solvability of a special class of partial differential equations. Consider for example
the set of first-order partial differential equations
30 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

∂h ∂h ∂h
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 =0
∂h ∂h ∂h (1.93)
∂ g1 g1 + ∂ x2 g2 + ∂ x3 g3 =0

where f i (x1 , x2 , x3 ) and gi (x1 , x2 , x3 ), i = 1, 2, 3 are scalar functions of x1 , x2 , x3

and h(x1 , x2 , x3 ) is an unknown function. If a solution h(x1 , x2 , x3 ) exists for the
above set of partial differential equations, then it is said that the set of vector fields
[ f, g] where f = ( f 1 , f 2 , f 3 )T and g = (g1 , g2 , g3 )T is completely integrable.
It has to be determined when this set of partial differential equations is solvable.
The Frobenius theorem provides a relatively simple condition about this. Equa-
tion (1.93) has a solution h(x1 , x2 , x3 ) if and only if there exist scalar functions
α1 (x1 , x2 , x3 ) and α2 (x1 , x2 , x3 ), such that

[ f, g] = α1 f + α2 g (1.94)

which means that the Lie bracket of f and g can be expressed as a linear combination
of f and g. This is the involutivity condition on the vector fields [ f, g]. Thus, the
Frobenius theorem states that the set of vector fields [ f, g] is completely integrable,
if and only if it is involutive.
Next, the Frobenius theorem is restated, according to the previous definitions of
complete integrability and involutivity.
Frobenius theorem: Let f 1 , f 2 , . . . , f m be a set of linearly independent vector fields.
The set is completely integrable, if and only if it is involutive.

1.7.3 Input–Output Linearization

Input–output linearization is analyzed for nonlinear dynamical systems of the form

ẋ = f (x) + g(x)u
(1.95)
y = h(x)

where f : D→R n , g : D→R n and h : D→R are smooth, D⊂R n is a domain.

Moreover, y is the system’s output. By input–output linearization one means the
generation of the linear differential relation between the output y and a new equivalent
input v [498]. The basic approach of input–output linearization is to differentiate the
output function y repeatedly until the input u appears.
The single-input single-output system of Eq. (1.95) is considered [238]

ẋ = f (x) + g(x)u
(1.96)
y = h(x)

The system has relative degree r on D, which means

L g L if h(x) = 0 for 0≤i≤r − 2, and L g L rf−1 h(x)=0, ∀x∈D (1.97)

1.7 Predecessors of Differential Flatness Theory 31

The following two lemmas are applied to the system [238]:

Lemma 1: For all x∈D and all integers k and j such that k≥0 and 0≤r − 1 one has

0 if 0≤ j + k < r − 1
L ad i g L kf h(k) = r −1 (1.98)
f (−1) j L g L f h(x) =0 if j + k = r − 1

Proof : The proof is based on induction. For j = 0 Eq. (1.98) it holds because it
becomes equivalent to the definition of the relative degree of the system of Eq. (1.96).
Next, it is assumed that Eq. (1.98) holds for some j and it is proven for j + 1. Using
the Jacobi identity one has

L [ f,β] λ(x) = L f L β λ(x) − L β L f λ(x) (1.99)

for any real-valued function λ and any vector fields f and β. Taking λ = L kf ,
j
β = ad f g one obtains

L ad j+1 g L kf h(k) = L [ f,ad i g] L kf h(x) = L f L ad i g L kf h(x) − L ad i g L k+1

f h(x)
f f f j
(1.100)

In the previous equation, the first term of the right-hand side is equal to 0, that is,
L f L ad i g L kf h(x) = 0 because j + k + 1≤r − 1 or j + k < r − 1.
f
Moreover, using that Eq. (1.98) holds for j one has also

0 if 0≤ j + k + 1 < r − 1
L ad i g L k+1
f h(k) = (1.101)
f (−1) j L g L rf−1 h(x)=0 if j + k + 1 = r − 1

Consequently, one obtains


0 if 0≤ j + k + 1 < r − 1
L ad i+1 g L kf h(k) = (1.102)
f (−1) j+1 L g L rf−1 h(x)=0 if j + k + 1 = r − 1

which completes the proof of the Lemma.

Lemma 2: For all x∈D
• the row vectors dh(x, d L f )h(x), . . . , d L rf−1 (x) are linearly independent
• the column vectors g(x), ad f g(x), . . . , ad rf −1 g(x) are linearly independent
32 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

Proof : One has

⎛ ⎞
dh(x)  
⎝ ··· ⎠ g(x) · · · ad r −1 g(x) =
f
d L rf−1 h(x)
⎛ ⎞
L g h(x) L ad f g h(x) · · · ··· L ad r −1 g h(x) (1.103)
⎜ L L h(x) ⎟
f
⎜ g f ··· · · · L ad r −2 g h(x) ∗ ⎟
⎜ f ⎟
⎝ ··· ··· ··· ··· ··· ⎠
r −1
L g L f h(x) ∗ ··· ··· ∗

From the previous Lemma, the right-hand side matrix is written as

⎛ ⎞
0 ··· ··· 0 
⎜0 ··· ···  ∗⎟
⎜ ⎟
⎜· · · ··· ··· ··· · · ·⎟ (1.104)
⎜ ⎟
⎝0  ··· ··· ∗⎠
 ∗ ··· ··· ∗

where  denotes a nonzero element. Thus, the matrix is nonsingular. This provides
a proof for the Lemma, because if any of the two matrices appearing on the left side
of the equation had rank less than n, their product should be a singular matrix.
The previous lemma proves that r ≤n. Next, the main result of input–output lin-
earization will be formulated: A single-input single-output system of relative degree
r < n can be locally transformed into the canonical (normal) form. This is given in
the following theorem:
Theorem: The system of Eq. (1.96) is considered and it is assumed that it has relative
degree r < n in D. Then, for every x0 ∈D, there exist a neighborhood N of x0 and
smooth functions φ1 (x), . . . , φn−r (x), such that

L g φi (x) = 0, ∀ 1≤i≤n − r, ∀ x∈N (1.105)

and the mapping ⎛ ⎞

φ1 (x)
⎜ ··· ⎟
⎜ ⎟
⎜ φn−r (x) ⎟
T (x) = ⎜
⎜ h(x) ⎟
⎟ (1.106)
⎜ ⎟
⎝ ··· ⎠
L rf−1 h(x)

restricted to N , is a diffeomorphism on N .
Proof : The distribution Δ = span{g} is nonsingular, involutive, and has dimension
1. By Frobenius’ theorem Δ is completely integrable. Consequently, for every x0 ∈D
1.7 Predecessors of Differential Flatness Theory 33

there exist a neighborhood N1 of x0 and n − 1 smooth functions φ1 (x), . . . , φn−1 (x)

with linearly independent differentials such that

L g φi (x) = 0 for 1≤i≤n − 1 ∀ x∈N1 (1.107)

Equivalently,
Δ⊥ = span{dφ1 , . . . , dφn−1 } (1.108)

on N1 . Since
L g L if h(x) = 0, for 0≤i≤r − 2 (1.109)

and dh 1 , . . . , d L rf−2 h(x) are linearly independent, one can use h, . . . , L rf−2 h as part
of these n − 1 functions. In particular they can be denoted as φn−r +1 , . . . , φn−1 .
Since L g L rf−1 h(x)=0, one has

d L rf−1 h(x0 )∈Δ

/ ⊥ (x0 )⇒rank[ ∂∂Tx (x0 )] = n⇒
∂T (1.110)
∂ x (x 0 ) is nonsingular

2 of x 0 such that T (x), restricted to N2 , is

This shows that there is a neighborhood N
a diffeomorphism on N2 . Taking N = N1 N2 completes the theorem’s proof.

1.7.4 Elaborating on Input–Output Linearization

1.7.4.1 The Case of a Well-Defined Relative Degree

First, a region Ωx of the state-space is considered. Using the notations of differential

geometry, the process of repeated differentiation gives

ẏ = ∇h( f + gu) = L f h(x) + L g h(x)u = v (1.111)

If L g h(x)=0 for some x = x0 in Ωx , then by continuity this relation is verified in

a finite neighborhood Ω of x0 . In Ω the control input that is finally exerted on the
system is given by
1
u= (−L f h + v) (1.112)
Lgh

results in a linear relation between u and v, which is ẏ = v.

If, however, L g h(x) = 0 for all x∈Ωx one can attempt to differentiate ẏ once
more with respect to time, and this gives

ÿ = L 2f h(x) + L g L f h(x)u (1.113)

If L g L f h(x) is again zero for all x∈Ωx , additional differentiations with respect to
time can be attempted
34 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

y (i) = L if h(x) + L g L i−1

f h(x)u (1.114)

until for some integer r one arrives at

L g L i−1
f h(x) =0 (1.115)

for some x = x0 in Ωx . If the above relation indeed holds then one can compute the
stabilizing control input that is finally applied to the system

1
u= (−L rf h + v) (1.116)
L g L i−1
f h(x)

By applying the control input of Eq. (1.116) to the equivalent description of the
system’s dynamics one gets

y (r ) = L rf h(x) + L g L rf−1 h(x)u (1.117)

one arrives at the input–output linearized description of the system

y (r ) = v (1.118)

The number r of the differentiations required for the input u to appear is the relative-
degree of the system. One has r ≤n, where n is the system’s order. If r = n then
input–output linearization becomes equivalent to the input-state linearization of the
system.

1.7.4.2 Input–Output Linearization and Normal Forms

When the relative degree r is defined and r < n, the nonlinear system of Eq. (1.95)
can be transformed to the normal form using the new state vector [y, ẏ, . . . , y (r −1) ].
Let
μ = [μ1 , μ2 , . . . , μr ]T = [y, ẏ, . . . , y (r −1) ]T (1.119)

In a neighborhood Ω of a point x0 , the normal form of the system can be written as

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
μ̇1 0 1 0 ··· ··· 0 0 μ1 0
⎜ μ̇2 ⎟ ⎜ 0 0 1 ··· ··· 0 0⎟ ⎜ μ2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ μ̇3 ⎟ ⎜ 0 ··· ··· 0⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ 0 0 0 ⎟ ⎜ μ3 ⎟ ⎜ 0 ⎟
⎜ · · · ⎟ = ⎜· · · ··· ··· ··· ··· ··· ⎟ ⎜
· · ·⎟ ⎜ · · · ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ + ⎜· · ·⎟ (α(μ, ψ) + (μ, ψ)u)
⎜ · · · ⎟ ⎜· · · ··· ··· ··· ··· ··· ⎟ ⎜
· · ·⎟ ⎜ · · · ⎟ ⎜
⎟ ⎟ ¯
⎜ ⎟ ⎜ ⎜· · ·⎟
⎝μ̇r −1 ⎠ ⎝ 0 0 0 ··· ··· 0 1 ⎠ ⎝μr −1 ⎠ ⎝ 0 ⎠
μ̇r 0 0 0 ··· ··· 0 0 μr 1
(1.120)
1.7 Predecessors of Differential Flatness Theory 35

with internal dynamics defined by

ψ̇ = w(μ, ψ) (1.121)

while the output function is defined by

y = μ1 (1.122)

Variables μi and ψ j are referred to as normal coordinates or normal states in Ω

(or at x0 ). It is noted that the internal dynamics of the system do not depend on the
control input u.
To demonstrate that the nonlinear system of Eq. (1.95) can indeed be transformed
into the normal form of Eq. (1.120), it has to be shown not only that a suitable
transformation exists. Equivalently, a diffeomorphism has to be defined as

φ(x) = [μ1 , . . . , μr , ψ1 , . . . , ψn−r ]T (1.123)

such that Eqs. (1.120) and (1.121) are verified. To show that a transformation φ is a
diffeomorphism it suffices to show that its Jacobian is invertible, i.e., the gradients
∇μi and ∇ψ j are linearly independent.
First, it will be shown that the gradients ∇μi are linearly independent which
means that the components of μ can serve as state variables. This result states that
the system’s output and its first r − 1 derivatives can serve as new state variables.
Next, it will be shown that other n − r vector fields ψ j can be found to complete
the new state vector, which means that the gradients ∇μi and ∇ψ j are also linearly
independent.
The following Lemma holds about the linear independence of the gradients ∇μi
[498]:
Lemma: If the relative degree of the system of Eq. (1.95) is r in the region Ω, then
the gradients ∇μ1 , ∇μ2 , . . ., ∇μr are also linearly independent in Ω.
Proof : It is first noted that in terms of μ Eq. (1.115) can be written as

∇μi g = 0 1≤i < r

(1.124)
∇μr g=0

Next, the proof is provided for the case that r = 3, however, it can be generalized to
higher values of r . It is assumed that there exist smooth functions αi (x), such that
everywhere in Ω
a1 ∇μ1 + a2 ∇μ2 + a3 ∇μ3 = 0 (1.125)

Multiplying Eq. (1.125) by g yields

[a1 ∇μ1 + a2 ∇μ2 + a3 ∇μ3 ]g = 0 (1.126)

36 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

which from Eq. (1.125) implies that α3 = 0 (everywhere in Ω). Replacing a3 = 0

in Eq. (1.126) gives
[a1 ∇μ1 + a2 ∇μ2 ] = 0 (1.127)

Multiplying Eq. (1.127) by the Lie bracket ad f g, and using Jacobi’s identity, as well
as the definition of the relative degree one has

0 = a1 L ad f g μ1 + a2 L ad f g μ2 =
= α1 [L f L g h − L g L f h] + α2 [L f L g − L g L f ]L f h = (1.128)
−a2 L g L 2f h

which from the definition of the relative degree implies that a2 = 0, everywhere in
Ω. Replacing a2 = 0 in Eq. (1.127) implies that

a1 ∇μ1 = 0 (1.129)

Multiplying Eq. (1.129) by ad 2f g and using the definition of the relative degree, as
well as the following result from Jacobi’s identity

L ad 2 g h = L ad f ad f g h = L f L ad f g h − L ad f g L f h =
f
= L f [L f L g h − L g L f h] − [L f L g − L g L f ]L f h = (1.130)
= L 2f L g h − 2L f L g L f h + L g L 2f h

one gets

0 = a1 L ad 2 g μ1 = a1 [L 2f L g h − 2L f L g L f h + L g L 2f h] = a1 L g L 2f h (1.131)
f

which again from the definition of the relative degree implies that a1 = 0, everywhere
in Ω. This finally confirms that the gradients ∇μi are linearly independent. This
completes the Lemma’s proof .
Next, it is shown that there exist n − r more functions ψ j to complete the coor-
dinates transformation. The proof about this is as follows:
It is needed to show that the gradient vectors ∇ψ j can be found such that the
∇μi and ∇ψ j are all linearly independent. At x0 Eq. (1.124) indicates that the first
r − 1 vectors ∇μi , (i = 1, 2, . . . , r − 1) which have been shown to be linearly
independent, are all within an hyperplane orthogonal to g. Since the dimension of
such hyperplane is n − 1, one can find

n − r = (n − 1) − (r − 1) (1.132)

vectors in that hyperplane that are linearly independent of the ∇μi and linearly
independent of each other. These vectors can be called ∇ψ j ( j = 1, 2, . . . , n − r )
and by definition they verify
1.7 Predecessors of Differential Flatness Theory 37

∇ψ j g = 0 1≤ j≤n − r (1.133)

Through Frobenius theorem it can also be assured that there exist n − 1 linearly
independent gradient functions ∇h k k = 1, . . . , n − 1 which satisfy ∇h k g = 0.
Furthermore, from Eq. (1.124) it is assured that μr is not in the hyperplane orthogonal
to g. Thus the gradients ∇μi (i = 1, . . . , r ) and ∇ψ j ( j = 1, . . . , n − r ) are
all linearly independent. Thus the Jacobian matrix of the transformation given in
Eq. (1.123) is invertible.
By continuity, the Jacobian remains invertible in a neighborhood Ω1 of x0 .
Redefining Ω as the intersection of Ω1 and Ω, the transformation φ defines a dif-
feomorphism in Ω. Hence, in Ω this transformation is a state transformation which
brings the nonlinear system in the form of Eq. (1.120), with

α(μ, ψ) = L rf h(x) = L rf h[ψ −1 (μ, ψ)]

(1.134)
β(μ, ψ) = L g L rf−1 h(x) = L g L rf−1 h[ψ −1 (μ, ψ)]

while in Eq. (1.121) the input u does not appear, since the ψ j verifies

L g ψ j (x) = 0 ∀x ∈ Ω (1.135)

From a practical point of view, to find the vector field ψ that completes the trans-
formation into a normal form requires the nontrivial step of solving the set of partial
differential equations given in Eq. (1.133).

1.7.5 Input-State Linearization

A different approach for obtaining a linearized equivalent description of a nonlinear

dynamical system is input-state linearization.
Definition: A nonlinear system

ẋ = f (x) + g(x)u (1.136)

where f : Dr →R n and G : Dr →R n× p are sufficiently smooth on a domain Dx ⊂R n ,

is said to be input-state linearizable if there exists a diffeomorphism T : Dx ⇒R n
such that Dx = T (Dx ) contains the origin, and the change of variables z = T (x)
transforms the system of Eq. (1.136) into the form

ż = Az + Bβ −1 (x)[u − α(x)] (1.137)

The following single-input system is considered:

ẋ = f (x) + g(x)u (1.138)

38 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

where f : D→R n and g : D→R n are smooth functions on a domain D⊂R n .

The following Theorem gives necessary and sufficient conditions for the system of
Eq. (1.138) to be input-to-state linearizable near a point x0 ∈D [238].
Theorem: The system of Eq. (1.138) is input-state linearizable if and only if there
exists a domain Dx ⊂D such that
1. the matrix G(x) = [g(x), ad f g(x), . . . , ad n−1
f g(x)] has rank n for all x∈Dx
2. the distribution D = span{g, ad f g, . . . , ad n−2
f g} is involutive in Dx
Proof : The system of Eq. (1.138) is input-to-state linearizable if and only if there is
a real-valued function h(x) such that the system

ẋ = f (x) + g(x)u
(1.139)
y = h(x)

has relative degree n in Dx . That is h(x) satisfies

L g L if h(x) = 0 for 0≤i≤n − 2 and L g L n−1

f h(x) =0 ∀x∈D x (1.140)

The “only if” part follows from the definition of input-to-state linearization while the
“if part” follows from applying the change of variables which is given by z = T (x) =
[h(x), L f h(x), . . . , L n−1 T
f h(x)] . The previously given Lemma 2 confirms that the
Jacobian matrix [ ∂∂Tx ](x) is nonsingular for all x∈Dx . Hence, for each x0 ∈Dx , there
is a neighborhood N of x0 such that T (x) is restricted to N , is a diffeomorphism. For
the theorem’s proof it is also necessary to show that the existence of h(x) satisfying
Eq. (1.140) is equivalent to the previous stated conditions 1 and 2.
Necessity: Suppose there is h(x) satisfying Eq. (1.140) shows that rank(G) = n.
Then D is nonsingular and has dimension n − 1. From Eq. (1.98), with k = 0 and
r = n one has

L g h(x) = L ad f g h(x) = · · · = L ad (n−2) g h(x) = 0 (1.141)

f

which can also be written as

dh(x)[g(x), ad f g(x), . . . , ad n−2

f g(x)] = 0 (1.142)

The above equation implies that D ⊥ = span(dh). Hence, D is completely integrable

and it follows from Frobenius theorem that D is involutive.
Sufficiency: It is assumed that conditions 1 and 2 hold. Then D is nonsingular and
has dimension n − 1. By Frobenius’ theorem, there exists h(x) satisfying

L g h(x) = L ad f g h(x) = · · · = L ad n−2 g h(x) = 0 (1.143)

f
1.7 Predecessors of Differential Flatness Theory 39

Using the Jacobi identity, it can be verified that

L g h(x) = L g L f h(x) = · · · = L g L n−2

f h(x) = 0 (1.144)

Furthermore,

dh(x)G(x) = dh(x)[g(x), ad f g(x), . . . , ad n−1

f g(x)] = [0, . . . , 0, L ad n−1 g h(x)]
f
(1.145)

Since rank(G) = n and dh(x)=0, it must be true that L ad n−1 g h(x)=0. Using the
f
Jacobi identity, it can be verified that L g L n−1
f h(x) =0, which completes the proof of
the theorem.

1.7.5.1 Relation Between Input-State and Input–Output Linearization

The formal definition of input-to-state-linearization is as follows:

Using the transformed state-space description of the system of Eq. (1.137) and by
defining the new control input v = β −1 (x)[u − α(x)], one obtains

ż = Az + Bv (1.146)

where ⎛ ⎞ ⎛ ⎞
0 1 0 ··· 0 0
⎜0 0 1 ··· 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟
⎜· · · ··· ··· ··· · · ·⎟ ⎜ ⎟
A=⎜ ⎟ B = ⎜ · · ·⎟ (1.147)
⎜· · · ··· ··· ··· ⎟
· · ·⎟ ⎜ · · ·⎟
⎜ ⎜ ⎟
⎝0 0 0 ··· 1 ⎠ ⎝0⎠
0 0 0 ··· 0 1

The new state vector z is called linearizing state and the control input of Eq. (1.146) is
the linearizing control law. The transformed system dynamics given in Eqs. (1.146)
and (1.147) is in the so-called linear canonical form, which is both controllable and
observable.
The canonical form of Eq. (1.146) stands for a special case of the input–output
linearization, where the output function leads to a relative degree n. This means that
if a system is input–output linearizable with relative degree n, it is also input-state
linearizable. On the other hand, is a system is input-state linearizable with a first new
state z 1 representing the output, the system is also input–output linearizable with
relative degree n.
Consequently, the relation between input-state linearization and input–output lin-
earization can be summarized as follows:
Lemma: An nth order nonlinear system is input-state linearizable, if and only if, there
exists a scalar function z 1 (x) such that the system’s input–output linearization with
z 1 (x) as output has relative degree n.
40 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

However, the above lemma provides no guidance about how to find the desirable
output function z 1 (x).

1.7.5.2 Conditions for Input-State Linearization

It should be specified which are the nonlinear systems of the form of Eq. (3.44) which
can be subjected to input-state linearization. The above question stands for one of the
most important issues in feedback linearization theory. The necessary and sufficient
conditions for input–output linearization of a dynamical system have been stated in
the theorem given in Sect. 1.7.5. Next, the proof of this theorem is revisited [498].
Theorem: The nonlinear system of Eq. (3.44) with f (x) and g(x) being smooth vector
fields, is input-state linearizable, if and only if, there exists a region Ω such that the
following conditions hold:
• the vector fields [g, ad f g, . . . , ad n−1
f g] are linearly independent in Ω
• the set [g, ad f g, . . . , ad n−2
f g] is involutive in Ω

Prior to proving the theorem some remarks about the aforementioned conditions are
given:
The first condition can be interpreted as a controllability condition for the nonlin-
ear system of Eq. (1.139). For linear systems, the vector fields [g, ad f g, . . . , ad n−1
f g]
become [b, Ab, . . . , An−1 b] and independence implies that the associated matrix is
a full-rank one which is also invertible. It can be also shown that if a system’s linear
approximations in a closed connected region Ω∈R are all controllable, then under
some mild smoothness assumptions, the system can be driven from any point in Ω
to any point in Ω (the inverse does not hold because, a nonlinear system can be
controllable whereas its linear approximation may not be).
The involutivity condition is less intuitive. It is trivially satisfied for linear systems
(which have constant vector fields) but does not hold in general in the nonlinear case.
Next, to proceed to the Theorem’s proof, the following Lemma is analyzed:
Lemma: Let z(x) be a smooth function in a region Ω. Then, in Ω, the set of equations

L g z = L g L f z = · · · = L g L kf z = 0 (1.148)

is equivalent to
L g z = L g L ad f gz = · · · = L g L ad k gz = 0 (1.149)
f

for any positive integer k.

Proof : The Lemma’s proof is based on induction. When k = 0 the result is obvious.
When k = 1, then from Jacobi’s identity one obtains

L ad f g z = L f L g z − L g L f z = 0 − 0 = 0 (1.150)
1.7 Predecessors of Differential Flatness Theory 41

When k = 2, the using again Jacobi’s identity then one obtains

L ad 2 g z = L 2f L g z − 2L f L g L f z + L g L 2f z = 0 − 0 + 0 = 0 (1.151)
f

Repeating the procedure, it can be shown by induction that Eq. (1.150) implies
Eq. (1.151) for any k. Finally, using Jacobi’s identity in an inverse manner it can
be shown that from Eq. (1.151) one can arrive at Eq. (1.150).
Proof of Theorem: First the necessity condition is proven. It is assumed that there
exist a state transformation z = z(x) and an input transformation u = α(x) + β(x)v,
such that z and v satisfy Eq. (1.146). The first row of Eq. (1.146) can be rewritten as

∂z 1
ż 1 = ( f + gu) = z 2 (1.152)
∂x
Continuing, in a similar manner with the rest of the elements of the state vector one
has the following differential equations:

∂z 1 ∂z 1
f + gu = z 2 (1.153)
∂x ∂x
∂z 2
∂x f + ∂z
∂ x gu = z 3
2
(1.154)
··· ··· ···

∂z n ∂z n
f + gu = v (1.155)
∂x ∂x
Since z 1 , . . . , z n are independent of u, while v is not, from the above equations it
can be concluded that

L g z 1 = L g z 2 = · · · = L g z n−1 = 0 and L g z n =0 (1.156)

L f z i = z i+1 i = 1, 2, . . . , n − 1 (1.157)

The above equations on z i can be compressed into one single equation concerning
only z 1 . Using the Lemma proven above and Eq. (1.156) one gets

∇z 1 ad kf g = 0 k = 0, 1, 2, . . . , n − 2 (1.158)

Moreover, by continuing as in the proof of the previous Lemma it can be shown that

∇z 1 ad n−1
f g = (−1)n−1 L g z n (1.159)

which implies that

∇z 1 ad n−1
f g=0 (1.160)
42 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

From Eqs. (1.158)–(1.160) one can conclude that the state vector fields [g, ad f g,
. . . , ad n−1
f g] are linearly independent. Indeed, if for some number i(i≤n − 1) there
existed scalar functions α1 (x), . . . , αi−1 (x), such that


i−1
ad if g = αk ad kf g (1.161)
k=0

one would also have


n−2
ad n−1
f g= αk ad kf g (1.162)
k=0

Equation (1.162) together with Eq. (1.159) would imply that


n−2
∇z 1 ad n−1
f g= αk ad kf g = 0 (1.163)
k=0

which is in contradiction to Eq. (1.160).

From Eqs. (1.158)–(1.160) it is also possible to conclude that the vector fields
are involutive. This follows from the existence of a scalar function z 1 satisfying the
n − 1 partial differential equations of Eq. (1.158) and from the necessity part of the
Frobenius theorem. This completes the necessity part of the present Theorem proof.
Next, it will be proven that the two conditions in the present Theorem are also
sufficient for the state linearizability of the system of Eq. (1.139), which means that
it is possible to find a state transformation and an input transformation such that
Eq. (1.146) is satisfied. The stages of the proof are as follows: Since the involutiv-
ity condition is satisfied, from the Frobenius theorem there exists a nonzero scalar
function z 1 (x) satisfying

L g z 1 = L ad f g z 1 = · · · = L ad n−2 g z 1 = 0 (1.164)
f

According to the previous Lemma, the above equation can be written as

L g z 1 = L g L f z 1 = · · · = L g L n−2
f z1 = 0 (1.165)

Consequently, if one uses as new state variables the elements of the vector z =
[z 1 , L f z 1 , . . . , L n−1
f z 1 ] , the first n − 1 state equations give
T

ż k = z k+1 k = 1, . . . , n − 1 (1.166)

while the last equation becomes

ż n = L nf z 1 + L g L n−1
f z1u (1.167)
1.7 Predecessors of Differential Flatness Theory 43

The next question that arises is if the gain function L g L n−1

f z 1 can be equal to zero.
Since the vector fields [g, ad f g, . . . , ad n−1
f g] are linearly independent in Ω, and
using Eq. (1.165), one obtains in accordance to the proof of the previous Lemma that

L g L n−1
f z 1 = (−1)
n−1
L ad n−1 g z 1 (1.168)
f

and consequently it should hold

L ad n−1 g z 1 (x)=0 ∀x∈Ω (1.169)

f

Otherwise, the nonzero vector ∇z 1 should satisfy

∇z 1 [g, ad f g, . . . , ad n−1
f g] = 0 (1.170)

which implies that ∇z 1 would be orthogonal to n linearly independent vectors, which

is a contradiction. For the system’s dynamics described in Eq. (1.167), and by denot-
ing as new control input v = L nf z 1 + L g L n−1
f z 1 u, the stabilizing feedback control
law that is finally applied to the system is
−1
u = (L g L n−1
f z 1 ) (−L f z 1 + v)
n
(1.171)

1.7.6 Stages in the Implementation of Input-State

Linearization

The stages for implementing input-state linearization of dynamical systems of the

form of Eq. (1.139) and for designing the associated feedback control law, are as
follows:
(i) the vector fields [g, ad f g, . . . , ad n−1
f g] for the given system is constructed.
(ii) it is checked if the controllability and involutivity conditions are satisfied.
If both conditions are satisfied, then one should find the first state z 1 (which is the
output function that leads to input–output linearization of relative degree n) from
Eqs. (1.164) and (1.165), that is

∇z 1 ad if = 0 i = 0, 1, . . . , n − 2
(1.172)
∇z 1 ad n−1
f =0

Next, the state transformation z(x) = [z 1 , L f z 1 , . . . , L n−1

f z 1 ] and the input trans-
T

formation u = α(x) + β(x)v are computed where

44 1 Nonlinear Dynamical Systems and Global Linearizing Control Methods

L nf z 1
α(x) = −
L g L n−1
f z1 (1.173)
β(x) = 1
L g L n−1
f z1

(n)
The transformed control input that makes the linearized system z 1 = v be a Hurwitz
stable one, i.e., to have all its poles in the left complex semi-plane is given by

(n)
v = z 1,d − K 1 (z 1(n−1) − z 1,d
(n−1)
) − K 2 (z 1(n−2) − z 1,d
(n−2)
) − ···−
(1.174)
−K n−1 (ż 1 − ż 1,d ) − K n (z 1 − z 1,d )

1.7.7 Input–Output and Input-State Linearization

for MIMO Systems

The previously analyzed approaches to input–output and input-state linearization

which were formulated for SISO systems can be also extended to MIMO systems.
Input–output linearization of MIMO systems is obtained similarly to the SISO case,
by differentiating the selected outputs yi until the inputs re-appear. It is assumed that
ri stands for the smallest order of output derivations, such that at least one of the
(r )
inputs appears in yi i , that is,

(ri )

m
yi = L rfi h i + L g j L rfi −1 h i u j (1.175)
j=1

with L g j L rfi −1 h i (x)=0, for at least one j, in a neighborhood Ω of point x0 . Following

this procedure of successive derivations for all chosen outputs of the system, one
obtains the input–output linearized description of the system’s dynamics in matrix
form
⎛ ri ⎞ ⎛ r 1 ⎞ ⎛ ⎞⎛ ⎞
y1 L f h 1 (x) L g1 L rf1 −1 h 1 (x) · · · L g1m L rfm −1 h m (x) u1
⎜···⎟ ⎜ ··· ⎟ ⎜ ⎟⎜ ⎟
1
⎜ ⎟=⎜ ⎟ ⎜ · · · · · · · · · ⎟ · · ·⎟
⎝···⎠ ⎝ ··· ⎠ + ⎜ ⎟⎜
⎝ ··· ··· ··· ⎠ ⎝· · ·⎠
rm
ym L rfm h m (x) L gm L r1 −1 h 1 (x) · · · L gm L rm −1 h m (x) um
1 f m f
(1.176)

In the above description the linearized dynamical system is a square one, that is, the
number of its outputs is equal to the number of control inputs. Therefore, through
the followed procedure one succeeds at the same time linearization and decoupling
of the system’s dynamics. Thus, finally the linearized system is written in the form

Ỹ = F̃ + G̃u⇒
(1.177)
Ỹ = v
1.7 Predecessors of Differential Flatness Theory 45

and the feedback control law that makes the elements of the output vector Ỹ converge
to the desirable setpoints is u = G̃ −1 (v− F̃). The above system is said to have relative
degree (r1 , . . . , rm ) and the scalar r = r1 + . . . + rm is the total relative degree.

1.7.8 Dynamic Extension

In certain cases, and despite successive derivations of the system’s outputs yi , i =

1, 2, . . . , m the control inputs may not reappear and thus matrix G̃ given in Eq. (1.177)
may be singular. To handle this case, the so-called dynamic extension or dynamic
feedback linearization is performed. This means that the system’s state vector is
extended by considering as additional state vector elements some of the control
inputs and their derivatives. For the extended state-space description it becomes
possible to perform input–output linearization and to obtain an equivalent linearized
and decoupled form of it. This control approach has as a result to generate a control
input that contains integral terms of the initial state vector of the system.
Without loss of generality, one can consider dynamic extension in the case of a
system with two inputs and two outputs and that matrix G̃ given in Eq. (1.177) is
of rank 1. Assuming that one column of matrix G̃ becomes equal to zero, we get a
description of the system (after successive differentiations) in form
   r
(r )
y1 1 L f1 h 1
(r2 ) = L r2 h + g̃1 u 1 (1.178)
2y f 2

By considering the control input as additional state variable, and by differentiating

Eq. (1.178) once more with respect to time while also using the initial system’s
dynamics, one arrives at the description
    
(r +1)
y1 1 b1 (x, u 1 ) ẽ1,1 (x, u 1 ) ẽ1,2 (x, u 1 ) u̇ 1
(r2 +1) = b (x, u ) + ẽ (x, u ) ẽ (x, u ) u
(1.179)
y2 2 1 2,1 1 2,2 2 2

which is also written as

  
y1(r1 +1) u˙1
(r2 +1) = B̃(x, u 1 ) + Ẽ(x, u 1 ) u = ṽ (1.180)
y2 2

The control input ṽ is chosen to place the poles of the input–output linearized system
to the left complex semiplane. The control input [u̇ 1 , u 2 ]T is given by

u̇ 1
= Ẽ −1 (ṽ − B̃(x, u 1 )) (1.181)
u2

After computing u˙1 from the aboverelation, the control input

 that is finally applied to
the dynamical system is given by [ u̇ 1 , u 2 ]T . The term u̇ 1 implies that the integral
of the initial state vector of the system is used for implementing the feedback control.
Chapter 2
Differential Flatness Theory
and Flatness-Based Control

2.1 Introduction

First, the chapter analyzes flatness-based control for lumped parameter systems,
that is systems which are described by ordinary differential equations. The chapter
overviews the definition and properties of differential flatness and presents basic
classes of differentially flat systems. It is explained that all dynamical systems which
satisfy differential flatness properties can be transformed through a change of vari-
ables into the linear canonical form. The first section of the chapter presents examples
of single-input dynamical systems which are written into the linear canonical form by
using the differential flatness theory diffeomorphism and the design of the associated
feedback control loop is explained. The case of MIMO differentially flat dynamical
system is also examined. It is shown that differentially flat systems which admit static
feedback linearization can be transformed into the linear canonical form. Moreover, it
is shown that for MIMO differentially flat systems, that admit only dynamic feedback
linearization, it is again possible to succeed transformation to the linear canonical
form and subsequently to design state feedback controllers.
Next, the chapter examines flatness-based control for distributed parameter sys-
tems, that is, systems which are described by partial differential equations. Unlike
control of lumped parameter systems, distributed parameter systems control has been
less investigated. Such systems are described by partial differential equations and
the associated boundary conditions and play a critical role in several engineering
problems, such as vibrating structures, flexible-link robots, waveguides and optical
fibers, heat conduction, etc. Differential flatness theory enables the solution of such
control problems. A flatness-based control method for distributed parameter systems
proposes the decomposition of the desirable trajectory into a series of a reference flat
output and its derivatives, and enables to compute control commands that succeed
trajectory tracking. One can also consider flatness-based control for PDE systems
which are based on the transformation of the PDE model into a finite differences
models and the associated state-space description in a canonical form. The chapter

© Springer International Publishing Switzerland 2015 47

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_2
48 2 Differential Flatness Theory and Flatness-Based Control

overviews main findings and methods on flatness-based control of systems exhibiting

a spatiotemporal (2D) dynamics.

2.2 Definition of Differentially Flat Systems

2.2.1 The Background of Differential Flatness Theory

Differential flatness theory and flatness-based control were introduced in the late
1980s by Michel Fliess and coresearchers and since then they keep on being devel-
oped and on providing efficient solutions to advanced control and state estimation
problems [153].
The definition of a differentially flat system is as follows: A system ẋ = f (x, u)
with state vector x ∈ R n , input vector u ∈ R m where f is a smooth vector field, is
differentially flat if there exists a vector y ∈ R m in the form

y = h(x, u, u̇, . . . , u (r ) ) (2.1)

such that

x = φ(y, ẏ, . . . , y (q) )

(2.2)
u = α(y, ẏ, . . . , y (q) )

where h, φ and α are smooth functions. This means that the new system’s description
is given by the m algebraic variables yi , i = 1, 2, . . . , m. The definition of the flat
output given above was y = h(x, u, u̇, . . . , u (r ) ). If the flat output is exclusively a
function of the state vector x then the system is a 0-flat one. However, there may be
a need to express the flat output as a function of not only the state vector x but also
as a function of the control u and of its derivatives. For instance, the latter holds in
the case of dynamic feedback linearization and in the application of the so-called
dynamic extension. This means that the state vector of the system is extended by
considering as additional state variables the control inputs and its derivatives.
Equation (2.2) shows that the state vector of the differentially flat system and its
control inputs can be expressed as function of the flat output and of the flat output’s
derivatives. The basic question that arises in the study of differential flatness is
whether, given the differential equations that describe the nonlinear system dynamics
ẋ = f (x, u), there exists a function y = h() given by y = h(x, u, u̇, . . . , u (r ) ), such
that the state vector of the system x and the control input u can be expressed as
functions of y and of its derivatives, as in Eq. (2.2).
This problem was initially set by D. Hilbert in 1912, for the second-order Monge’s
equation
d2 y dy dz
= F(x, y, z, , ) (2.3)
dx2 dx dx
2.2 Definition of Differentially Flat Systems 49

which is an underdetermined differential equation system (there is one differential

equation with respect to x, connecting two functions of x, namely y(x) and z(x)).
Hilbert speaks about a solution that is not based on the computation of integrals.
He has shown that this problem is related to the classification of underdetermined
differential equation systems through a group of transformations called “invertible
without integral.”
Soon afterwards, Èlie Cartan, reworks on the question set by Hilbert and shows
how the calculations on the Pfaff systems, permit to classify second-order Monge
equations which admit a solution without integral. He deduced an explicit description
for all curves in the Euclidean space R 3 for which the curvature ratio and the torsion
are constant. The results of Èlie Cartan are complete for Pfaff systems of codimension
2, which are underdetermined differential equation systems of codimension 1, i.e.,
systems in which only one arbitrary function intervenes (otherwise stated, systems
having only one control input). Èlie Cartan also suggested the notion of absolute
equivalence; however, he did not define it with precision. He also noted that for
Pfaff systems of codimension equal or greater than 2 (which means systems having
at least two control inputs) the solution of the above problem becomes extremely
complicated.
Next a formal definition will be given about the analogy in terms of the con-
trol theory, of what was described above as underdetermined differential equation
systems which can be solved without integration.

2.2.2 Differential Flatness for Finite Dimensional Systems

As noted in Eqs. (2.1) and (2.2) differential flatness is a structural property of a

class of nonlinear dynamical systems, denoting that all system variables (such as
state vector elements and control inputs) can be written in terms of a set of specific
variables (the so-called flat outputs) and their derivatives. The following nonlinear
system is considered:
ẋ(t) = f (x(t), u(t)) (2.4)

The time variable is t ∈ R, the state vector is x(t) ∈ R n with initial conditions x(0) =
x0 , and the input variable is u(t) ∈ R m . Next, the main principles of differentially
flat systems are given [465, 535]:
The finite dimensional system of Eq. (2.4) can be written in the general form of an
ordinary differential equation (ODE), i.e., Si (w, ẇ, ẅ, . . . , w(i) ), i = 1, 2, . . . , q.
The entity w is a generic notation for the system variables (these variables are,
for instance, the elements of the system’s state vector x(t) and the elements of
the control input u(t)) while w(i) , i = 1, 2, . . . , q are the associated derivatives.
Such a system is said to be differentially flat if there is a collection of m functions
y = (y1 , . . . , ym ) of the system variables and of their time derivatives, i.e., yi =
φ(w, ẇ, ẅ, . . . , w(αi ) ), i = 1, . . . , m satisfying the following two conditions [152,
340, 362, 364, 422]:
50 2 Differential Flatness Theory and Flatness-Based Control

1. There does not exist any differential relation of the form R(y, ẏ, . . . , y (β) ) = 0
which implies that the derivatives of the flat output are not coupled in the sense of an
ODE, or equivalently it can be said that the flat output is differentially independent.
2. All system variables (i.e., the elements of the system’s state vector w and the
control input) can be expressed using only the flat output y and its time derivatives
wi = ψi (y, ẏ, . . . , y (γi ) ), i = 1, . . . , s. An equivalent definition of differentially
flat systems is as follows:
Definition: The system ẋ = f (x, u), x ∈ R n , u ∈ R m is differentially flat if there
exist relations

h : R n × (R m )r +1 → R m ,
φ : (R m )r → R n and (2.5)
ψ : (R m )r +1 → R m

such that

y = h(x, u, u̇, . . . , u (r ) ),
x = φ(y, ẏ, . . . , y (r −1) ), and (2.6)
u = ψ(y, ẏ, . . . , y (r −1) , y (r ) ).

This means that all system dynamics can be expressed as a function of the flat output
and its derivatives; therefore, the state vector and the control input can be written as

x(t) = φ(y(t), ẏ(t), . . . , y (r −1) (t)), and

(2.7)
u(t) = ψ(y(t), ẏ(t), . . . , y (r ) (t))

Next, an example is given to explain the design of a differentially flat controller for
finite dimensional systems of known parameters.
Example 1: Flatness-based control for a nonlinear system of known parameters [263].
Consider the following model:

ẋ1 = x3 − x2 u
ẋ2 = −x2 + u (2.8)
ẋ3 = x2 − x1 + 2x2 (u − x2 )

x22
The flat output is chosen to be y1 = x1 + 2 . Thus one gets:

x2
y1 = x1 + 22
y2 = ẏ1 = (x3 − x2 u) + x2 (u − x2 ) = x3 − x22 (2.9)
y3 = ẏ2 = ÿ1 = x2 − x1 + 2x2 (u − x2 ) − 2x2 (u − x2 ) = −x1 + x2
(3)
v = ẏ3 = y1 = −x3 + x2 u − x2 + u = −x2 − x3 + u(1 + x2 )

It can be verified that property (1) holds, i.e., there does not exist any differential
relation of the form R(y, ẏ, . . . , y (β) ) = 0, and this implies that the derivatives of
2.2 Definition of Differentially Flat Systems 51

the flat output are not coupled. Moreover, it can be shown that property (2) also
holds i.e., the components w of the system (elements of the system’s state vector and
control input) can be expressed using only the flat output y and its time derivatives
wi = ψi (y, ẏ, . . . , y (γi ) ), i = i, . . . , s.
(3)
For instance to calculate x1 with respect to y1 , ẏ1 , ÿ1 and y1 the relation of ÿ1 is
used, i.e.,
x12 + 2x1 (1 + ÿ1 ) + ÿ12 − 2y1 = 0 (2.10)
√
√ solutions are derived, i.e., x1 = −(1+ ÿ1 − 1 + 2(y1 + ÿ1 ))
from which two possible
and x1 = −(1 + ÿ1 + 1 + 2(y1 + ÿ1 )). Keeping the biggest out of these two solu-
tions one obtains: √
x1 = −(1 + ÿ1 ) + 1 + 2(y1 + ÿ1 )
x2 = ÿ1 + x1
x3 = ẏ1 + ÿ12 + 2x1 ÿ1 + x12 (2.11)
(3)
y1 + ÿ12 + ÿ1 + ẏ1 +x1 +2x1 ÿ1 +x12
u= 1+x1 + ÿ1

The computation of the equivalent model of the system in the linear canonical form
is summarized as follows: By finding the derivatives of the flat output one gets a
set of equations which can be solved with respect to the state variables and the
control input of the initial state-space description of the system. First, the binomial
of variable x1 given in Eq. (2.10) is solved providing x1 as a function of the flat output
and its derivatives. Next, using the expression for x1 and Eq. (2.11), state variable
x2 is also written as a function of the flat output and its derivatives. Finally, using
the expressions for both x1 and x2 and Eq. (2.11), state variable x3 is written as a
function of the flat output and its derivatives. Thus one can finally express the state
vector elements and the control input as function of the flat output and its derivatives,
which completes the proof about differential flatness of the system.
From Eq. (2.11) it can be concluded that the initial system of Eq. (2.8) is indeed
differentially flat. Using the flat output and its derivatives, the system of Eq. (2.8) can
be written in Brunovsky (canonical) form:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
y 010 y1 0
d ⎝ 1⎠ ⎝
y2 = 0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (2.12)
dt y 000 y3 1
3

where the new control input is v = f (x) + g(x)u. Therefore, a transformation of the
system into a linear equivalent description is obtained and then it is straightforward to
design a controller based on linear control theory. Thus, given the reference trajectory
[x1∗ , x2∗ , x3∗ ]T , one can find the transformed reference trajectory [y1∗ , ẏ1∗ , ÿ1∗ ]T and
select the appropriate control input v that succeeds tracking of the reference setpoints.
Knowing v, the control u of the initial system can be found. Knowing v the control
input that is actually applied to the system is u = g −1 (x)[v − f (x)].
52 2 Differential Flatness Theory and Flatness-Based Control

It is noted that for linear systems, the property of differential flatness is equivalent
to that of controllability. Next, two examples of differentially flat MIMO dynamical
systems are given. It is shown that the definition of the differential flat outputs also
enables to transform the system into the Brunovksy (canonical) form:
Example 2: Differential flatness of a nonlinear spring–damper–mass system which
consists of two masses.
The spring–damper–mass model is described in Fig. 2.1. The dynamic equations
of the model are given by [89]

M1 ẍ1 = − f K 1 (x) − f B1 (x)+

+ f K 2 (x) + f B2 (x) + u 1 + 0.2u 2
(2.13)
M2 ẍ2 = − f K 2 (x) − f B2 (x) + 0.25u 1 + u 2

where M1 and M2 are the masses of the system, x(t) = [x1 , ẋ1 , x2 , ẋ2 ]T is the state
vector which has as elements the positions and the velocities of the two masses, and
f K 1 (x) and f K 2 (x) are spring forces defined by the following equations

f K 1 (x) = K 10 + ΔK 1 x13
(2.14)
f K 2 (x) = K 20 + ΔK 2 (x2 − x1 )3

f B1 (x) and f B2 (x) are friction forces which are defined as

f B1 (x) = b10 ẋ1 + Δb1 ẋ12

(2.15)
f B2 (x) = b20 (ẋ2 − ẋ1 ) + Δb2 (ẋ2 − ẋ1 )2

Fig. 2.1 A spring–damper–

mass system consisting of
two masses
2.2 Definition of Differentially Flat Systems 53

The following flat outputs are defined

y1 = x1 y2 = x2 (2.16)

Obviously, it holds
x1 = y1 x2 = y2
(2.17)
ẋ1 = ẏ1 ẋ2 = ẏ2

Thus, it is observed that the state variables of the spring-damper-mass model can be
expressed as functions of the flat outputs and of the associated derivatives. Moreover,
the following relations can be obtained about the control inputs of the model

M1 ẍ1 =
−(K 10 x1 + ΔK 1 x13 ) − (b10 ẋ1 + Δb1 ẋ12 )+
(2.18)
+(K 20 (x2 − x1 ) + ΔK 2 (x2 − x1 )3 )+
+(b20 (ẋ2 − ẋ1 ) + Δb2 (ẋ2 − ẋ1 )2 ) + u 1 + 0.2u 2

M2 ẍ2 =
−(K 20 (x2 − x1 ) + ΔK 2 (x2 − x1 )3 ) − (b20 (ẋ2 − ẋ1 )+ (2.19)
Δb2 (ẋ2 − ẋ1 )2 + 0.25u 1 + u 2

Using the definition of the flat outputs in Eqs. (2.18) and (2.19), one obtains

M1 ÿ1 =
−(K 10 y1 + ΔK 1 y13 ) − (b10 ẏ1 + Δb1 ẏ12 )+
(2.20)
+(K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 )3 )+
+(b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 ) + u 1 + 0.2u 2

M2 ÿ2 =
−(K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 )3 ) − (b20 ( ẏ2 − ẏ1 )+ (2.21)
+Δb2 ( ẏ2 − ẏ1 )2 + 0.25u 1 + u 2

or equivalently

M1 ÿ1 + (K 10 y1 + ΔK 1 y13 ) + (b10 ẏ1 +

+Δb1 ẏ12 ) − (K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 )3 )− (2.22)
−(b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 ) = u 1 + 0.2u 2

M2 ÿ2 + (K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 )3 )+

+(b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 = (2.23)
= 0.25u 1 + u 2
54 2 Differential Flatness Theory and Flatness-Based Control

Defining the matrix of the coefficients of the control inputs and its inverse as
   
1 0.2 1.0526 −0.2105
A= A−1 = (2.24)
0.25 1 −0.2632 1.0526

one obtains the following relation about the control inputs and the flat outputs defined
in Eq. (2.16)     
u1 1.0526 −0.2105 q1
= (2.25)
u2 −0.2632 1.0526 q2

where q1 = M1 ÿ1 + (k10 y1 + ΔK 1 y13 ) + (b10 ẏ1 + Δb1 ẏ12 ) + (K 20 (y2 − y1 ) +

ΔK 2 (y2 − y1 )3 ) + (b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 ) and q2 = M2 ÿ2 + (K 20 (y2 −
y1 ) + ΔK 2 (y2 − y1 )3 ) + (b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 ).
The previous relation between the state variables of the system and the flat outputs
enables to write the system in the Brunovsky (canonical) form

ÿ1 =
M1 {−(K 10 y1 + ΔK 1 y1 ) − (b10 ẏ1 + Δb1 ẏ1 )}+
1 3 2
(2.26)
M1 {(K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 ) )+
1 3

+b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )} + M11 u 1 + 0.2 M11 u 2

2

ÿ2 =
M2 {−(K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 ) )−
1 3
(2.27)
−(b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 ) )} + 0.25 M12 u 1 +
2 1
M2 u 2

Consequently, one has

ÿ1 = f 1 (y1 , ẏ1 , y2 , ẏ2 ) + g11 (x)u 1 + g12 (x)u 2 (2.28)

ÿ2 = f 2 (y1 , ẏ1 , y2 , ẏ2 ) + g21 (x)u 1 + g22 (x)u 2 (2.29)

with
f 1 (y1 , ẏ1 , y2 , ẏ2 ) = f 1 (x) =
M1 {−(K 10 y1 + ΔK 1 y1 ) − (b10 ẏ1 + Δb1 ẏ1 )}+
1 3 2
(2.30)
M1 {(K 20 (y2 − y1 ) + ΔK 2 (y2 − y1 ) )+
1 3

+b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 )2 }

g11 (y1 , ẏ1 , y2 , ẏ2 ) = g12 (x) = 1

M1
(2.31)
g12 (y1 , ẏ1 , y2 , ẏ2 ) = g12 (x) = 0.2
M1
2.2 Definition of Differentially Flat Systems 55

f 2 (y1 , ẏ1 , y2 , ẏ2 ) = f 2 (x) =

M2 {−(K 20 (y2 − y1 )) + ΔK 2 (y2 − y1 ) −
1 3
(2.32)
−(b20 ( ẏ2 − ẏ1 ) + Δb2 ( ẏ2 − ẏ1 ) )}
2

g21 (y1 , ẏ1 , y2 , ẏ2 ) = g21 (x) = 0.25

M2 (2.33)
g22 (y1 , ẏ1 , y2 , ẏ2 ) = g22 (x) = M12

Thus, one obtains the following canonical form description for the nonlinear system
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẋ1 0 1 0 0 x1 0 0  
⎜ẋ2 ⎟ ⎜0 0 0 0⎟ ⎜ x 2 ⎟ ⎜1 0⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟ v1 (2.34)
⎝ẋ3 ⎠ ⎝0 0 0 1⎠ ⎝x3 ⎠ ⎝0 0⎠ v2
ẋ4 0 0 0 0 x4 0 1

where
v1 = f 1 (x) + g11 (x)u 1 + g12 (x)u 2
(2.35)
v2 = f 2 (x) + g21 (x)u 1 + g22 (x)u 2

or equivalently
v = F(x) + G(x)u (2.36)

where      
v1 u1 f 1 (x)
v= , u= ,F = ,
v2 u2 f 2 (x)
  (2.37)
g (x) g12 (x)
G = 11
g21 (x) g22 (x)

Example 3: Differential flatness of the VTOL aircraft (vertically take-off and landing
aircraft).
The dynamic model of the vertically take-off and landing aircraft (Fig. 2.2) is
described by the following set of equations [561]

ẍ = u 1 sin(θ ) − εu 2 cos(θ )
z̈ = u 1 cos(θ ) + εu 2 sin(θ ) − 1 (2.38)
θ̈ = u 2
56 2 Differential Flatness Theory and Flatness-Based Control

Fig. 2.2 The vertically

take-off and landing aircraft
model

where ε is a small parameter. Defining y1 and y2 two smooth functions satisfying

the condition ÿ12 + ( ÿ2 + 1)2 = 0, e.g., y1 = x − εsin(θ ), y2 = z + εcos(θ )

ε ÿ1
x = y1 −
,
ÿ1 +( ÿ2 +1)2
2

z = y2 −
ε ÿ2 +1 , (2.39)
ÿ12 +( ÿ2 +1)2

θ = tan −1 ( ÿ2ÿ+1
1
)

Since all state variables are expressed as functions of the flat outputs and their deriv-
atives, one can also write the control inputs u 1 and u 2 as functions of the flat outputs
and their derivatives. This confirms that the model of the vertically take-off and
landing aircraft is differentially flat. The system can be written in the Brunovsky
form
(4) (4)
y1 = v1 , y2 = v2 (2.40)

where the control inputs v1 and v2 are defined as

v1 = η̇2 sin(θ ) + 2η2 θ̇cos(θ )+

+η1 u 2 cos(θ ) − η1 θ̇ 2 sin(θ )
(2.41)
v2 = η̇2 cos(θ ) − 2η2 θ̇ sin(θ )−
−η1 u 2 sin(θ ) − η1 θ̇ 2 cos(θ )

with variables η1 and η2 being defined as

η1 = u 1 − εθ̇ 2 , η2 = η̇1 (2.42)

2.2 Definition of Differentially Flat Systems 57

The previous Brunovsky-form model of the vertically take-off and landing aircraft
can be also written using state-space equations
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1,1 0 1 0 0 0 0 0 0 y1,1 0 0
⎜ ẏ1,2 ⎟ ⎜0 0 1 0 0 0 0 0⎟ ⎜ y1,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ẏ1,3 ⎟ ⎜0 0 0 1 0 0 0 0⎟ ⎜ ⎟ ⎜ 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ y1,3 ⎟ ⎜0 ⎟ 
⎜ ẏ1,4 ⎟ ⎜0 0 0 0 0 0 0 ⎟ ⎜
0⎟ ⎜ y1,4 ⎟ ⎜ 0⎟
⎜ ⎟ ⎜ ⎟ + ⎜1 ⎟ v1
⎜ ẏ2,1 ⎟ = ⎜0 0 0 0 0 1 0 ⎟ ⎜
0⎟ ⎜ y2,1 ⎟ ⎜
⎟ 0⎟
(2.43)
⎜ ⎟ ⎜ ⎜0 ⎟ v2
⎜ ẏ2,2 ⎟ ⎜0 0 0 0 0 0 1 0⎟ ⎜ y2,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ẏ2,3 ⎠ ⎝0 0 0 0 0 0 0 1⎠ ⎝ y2,3 ⎠ ⎝0 0⎠
ẏ2,4 0 0 0 0 0 0 0 0 y2,4 0 1

2.3 Properties of Differentially Flat Systems

2.3.1 Equivalence and Differential Flatness

2.3.1.1 Representation of System Dynamics as Vector Fields of Infinite

Dimension

A basic property of differentially flat systems is that through a change of variables

(diffeomorphism) they can be transformed to an equivalent description, which is
the linear canonical (Brunovsky) form. This is analyzed next and stands for the
Lie-Backlünd isomorphism to equivalence and differential flatness [153].
First a dynamical system of the form

ẋ = f (x) x ∈ X ⊂ R n (2.44)

is considered. This is described by the couple (X, f ), where X is defined in R n and

f is a vector field on X . A trajectory is considered to be the function t→ x(t) such
that ẋ(t) = f (x(t)) ∀t ≥ 0. One can consider also the output mapping x → h(x)
for which holds
d dh dh
h(x(t)) = (x(t)) · ẋ(t) = (x(t)) · f (x(t)) ∀ t ≥ 0 (2.45)
dt dt dt

dt (x(t)) · f (x(t)) is also called time

Equation (2.45) gives the total derivative, i.e., dh
derivative of the function h, and is noted by ḣ.
Next, the notions of the total derivative and of the time derivative of function h
can be generalized to the case of a nonlinear system with control input

ẋ = f (x, u) (2.46)
58 2 Differential Flatness Theory and Flatness-Based Control

where X × U ∈ R n × R m . Here, f is no longer a vector field but is an infinite

collection of vector fields parameterized by u. Actually, for every u the function
x → f u (x) = f (x, u) is a vector field on X . One can also consider the case of
dynamic feedback in which the system’s state vector is extended by defining as state
vector element the control input (and its derivatives), while the equations describing
the system’s dynamics are extended as follows:

ẋ = f (x, u)
(2.47)
u̇ = v

In the latter case, in place of the state-space X one has the state-space X × U .
Assume now that the solution of ẋ = f (x, u) is the function t → (x(t, u(t))) which
has values in X × U , such that ẋ(t) = f (x(t), u(t)) ∀ t ≥ 0. Moreover, one can
consider the infinite function

t → ξ(t) = (x(t), u(t), u̇(t), . . .) (2.48)

which takes values in X × U × Rm ∞ , where R ∞ = R m × R m × · · · represents an

m
infinite product formed by vectors defined in R m . A point of Rm ∞ is of the form
(1) (2)
u , u , . . . with u ∈ R , using that ẋ = f (x, u). This function also satisfies the
i m

relation
ξ̇ (t) = ( f (x(t), u(t)), u̇(t), ü(t), . . .) ∀ t ≥ 0 (2.49)

and this can be interpreted as a trajectory of an infinite vectors field

(x, u, u (1) , . . .) → F(x, u, u (1) , . . .) = ( f (x, u), u (1) , u (2) , . . . , ) (2.50)

According to the above, the dynamical system ẋ = f (x, u) is defined by the space
X × U × Rm ∞ and an infinite number of vector fields F on that space. As in the case

of an autonomous dynamical system ẋ = f (x), for the nonautonomous dynamical

system ẋ = f (x, u) one can also define the time derivative of the smooth output
function (x, u, u (1) , . . .) → h(x, u, u (1) , . . . , u (k) ), which is written as

ḣ(x, u, u (1) , . . . , u (k+1) ) := Dh · F =

(2.51)
= ∂∂hx f (x, u) + ∂h ∂u u
(1) + ∂h u (2) + · · ·
∂u 1

If h() depends on a finite number of variables then the above sum becomes finite
too. For the above type of functions, one can define the so-called Fréchet topology
which implies computations on smooth functions which are defined in k copies of
R m , for k being sufficiently large.
After a change of variables, a differentially flat system is written in the canonical
Brunovsky form. Thus, it becomes equivalent to the trivial system Rm ∞ , F with
m
(1) (2)
coordinates y, y , y , . . . , and the vector fields

(y (1) , y (2) , y (3) , . . .) = Fm (y, y (1) , y (2) , . . .) (2.52)

2.3 Properties of Differentially Flat Systems 59

which describes systems composed by m chains of integrators (this is the canonical

Brunovsky form for linear controllable systems).

2.3.1.2 Equivalence of Systems

Two dynamical systems are considered to be equivalents if there exists an invertible

relationship which exchanges their trajectories. The definition of equivalent systems
is as follows:
Definition: Two systems (M, F) and (N , G) (notation referring to state vector and
vector field, respectively) are equivalent in ( p, q) ∈ (M, N ), if and only if, there
exists an endogenous transformation from a neighborhood of p to a neighborhood
of q. The two systems (M, F) and (N , G) are equivalent if the equivalence holds
for all pairs of points ( p, q) of the space M × N .
Using coordinates, the previous notions are expressed as follows: Considering the
two systems (X × U × Rm ∞ , F) and (Y × V × R ∞ , G) describing the initial system
s
dynamics
ẋ = f (x, u) with (x, u) ∈ X × U ⊂ R n × R m (2.53)

and the equivalent system dynamics

ẏ = g(y, v) with (y, v) ∈ Y × V ⊂ R r × R s (2.54)

The vector fields F, G are defined as

F(x, u, u (1) , . . .) = ( f (x, u), u (1) , u (2) , . . .)

(2.55)
G(y, v, v(1) , . . .) = (g(y, v), v(1) , v(2) , . . .)

If the two systems are equivalent, the endogenous transformation takes the form

Ψ (x, u, u (1) , . . .) = (ψ(x, ū), β(x, ū), β̇(x, ū), . . .) (2.56)

where ū is the abbreviated notation ū = (u, u (1) , . . . , u (k) ). Similarly, one can define
the inverse endogenous transformation

Φ(y, v, v(1) , . . .) = (φ(y, v̄), α(y, v̄), α̇(y, v̄), . . .) (2.57)

Since Φ and Ψ are inverse to each other, one has

ψ(φ(y, v̄), ᾱ(y, v̄)) = y and φ(ψ(x, ū), β̄(x, ū)) = x

(2.58)
β(φ(y, v̄), ᾱ(y, v̄)) = v and α(ψ(x, ū), β̄(x, ū)) = u
60 2 Differential Flatness Theory and Flatness-Based Control

About vector fields F and G which are related to the derivatives of x and y, respec-
tively, one has
f (φ(y, v̄), α(y, v̄)) = Dφ(y, v̄) · ḡ(y, v̄) (2.59)

where ḡ corresponds to (g, v(1) , . . . , v(k) )

g(ψ(x, ū), β(x, ū)) = Dψ(x, ū) · f¯(x, ū) (2.60)

where f¯ corresponds to ( f, u (1) , . . . , u (k) ). Additionally, it can be stated that if the

trajectory of the first system is denoted as

t → (x(t), u(t)) (2.61)

then the trajectory of the second system becomes

t → (y(t), v(t)) = (ψ(x(t), ū(t)), β(x(t), ū(t))) (2.62)

Example 1: The VTOL aircraft (Vertical Take-off and Landing Aircraft) studied in
Sect. 2.2.2 is revisited. State-space transformation for the VTOL model provides an
example of equivalence between dynamical systems descriptions.
The initial dynamic model of the VTOL aircraft is given by

ẍ = −u 1 sin(θ ) + εu 2 cos(θ )
z̈ = u 1 cos(θ ) + εu 2 sin(θ ) − 1 (2.63)
θ̇ = u 2

This system is first shown to be globally equivalent to the transformed model

ÿ1 = −ξ sin(θ )
ÿ2 = ξ cos(θ ) − 1 (2.64)
θ̇ = u 2

where ξ and θ stand for the new control inputs. Indeed, choosing

X := (x, z, ż, ż, θ, θ̇ ) and Y := (y1 , y2 , ẏ1 , ẏ2 )

(2.65)
U := (u 1 , u 2 ) V = (ξ, θ )

and with the abbreviated notations given above one can define the functions of the
transformed state vector and of the transformed control input Y = ψ(X, Ū ) and
V = β(X, Ū ), as
⎛ ⎞
x − εsin(θ )  
⎜ z + εcos(θ ) ⎟
⎟ and β(X, Ū ) = u 1 − εθ̇
2
ψ(X, U ) = ⎜
⎝ẋ − εθ̇ cos(θ )⎠ (2.66)
θ
ż − εθ̇ sin(θ )
2.3 Properties of Differentially Flat Systems 61

which finally enable to obtain the endogenous transformation Ψ . The inverse trans-
formation Φ is given by X = φ(Y, V̄ ) and U = α(Y, V̄ ) according to the following
⎛ ⎞
y1 + εsin(θ )
⎜ y1 − εcos(θ ) ⎟
⎜ ⎟  
⎜ ẏ1 + εθ̇ cos(θ )⎟ ξ + εθ̇ 2
⎜
φ(Y, V̄ ) = ⎜ ⎟
⎟ and α(Y, V̄ ) = (2.67)
⎜ ẏ2 − εθ̇ sin(θ )⎟ θ̈
⎝ θ ⎠
θ̇

Continuing with the vertical take-off and landing aircraft (VTOL) and using the
dynamics of the aircraft which has been defined in Eq. (2.63), it is found that this
system admits as flat output

y = (x − εsin(θ ), z + εcos(θ )) (2.68)

To express the state variables and the control input as functions of the flat output and
its derivatives (this enables to find also the elements of the inverse transformation of
Φ, that is X = φ( ȳ) and U = α(y)), the following implicit relations are used:

(y1 − x)2 + (y2 − z)2 = ε2

(y1 − x)( ÿ2 + 1)(y2 − z) ÿ1 = 0 (2.69)
( ÿ2 + 1)sin(θ ) + ÿ1 cos(θ ) = 0

Solving the above set of equations with respect to x, z, and θ , one gets

ÿ1
x = y1 ±ε

ÿ12 +( ÿ2 +1)2

z = y2 ±ε
( ÿ2 +1) (2.70)
ÿ12 +( ÿ2 +1)2
θ= −1
tan ( ÿ2 + ÿ11 )

One has simply to differentiate so as to obtain ẋ, ż, θ̇ , and u as functions of the deriv-
atives of the flat output y. Singularity of the system is avoided if ÿ12 + ( ÿ2 + 1)2 = 0.
Next, the system is transformed to the linear canonical form described in Eq. (2.43)
through dynamic feedback linearization. The feedback control consists of the fol-
lowing elements
ξ̇ = −v1 sin(θ ) + v2 cos(θ ) + ξ θ̇ 2
u 1 = ξ + εθ̇ 2 (2.71)
u 2 = − ξ (v1 cos(θ ) + v2 sin(θ ) + 2ξ̇ (θ̇))
1
62 2 Differential Flatness Theory and Flatness-Based Control

which transforms the system into the linear canonical form

(4)
y1 = v1
(2.72)
y2(4) = v2

Thus, now one has the change of coordinates

(x, z, θ, ẋ, ż, θ̇ , ξ, ξ̇ ) → (y, ẏ, ÿ, y (3) ) (2.73)

The only singularity which may appear in this feedback control loop is when ξ = 0
that is ÿ12 + ( ÿ2 + 1)2 = 0 (which practically means that the aircraft is in free fall).
Example 2: The underactuated vessel dynamics. The model of an autonomous hov-
ercraft provides another example about equivalence between an initial complicated
nonlinear description of its dynamics and the linear canonical (Brunovsky form)
[461].
The hovercraft model is obtained from the generic ship’s model, after setting
specific values for the elements of the inertia and Coriolis matrix and after reducing
the number of the available control inputs. The state-space equation of the nonlinear
underactuated hovercraft model (Fig. 2.3) is given by

ẋ = ucos(ψ) − vsin(ψ)
ẏ = usin(ψ) + vcos(ψ)
ψ̇ = r
(2.74)
u̇ = v · r + τu
v̇ = −u · r − βv
ṙ = τr

Fig. 2.3 Diagram of the

underactuated hovercraft’s
kinematic model
2.3 Properties of Differentially Flat Systems 63

where x and y are the cartesian coordinates of the vessel, ψ is the orientation angle,
u is the surge velocity, v is the sway velocity, and r is the yaw rate. Coefficient β
is a function of elements of the inertia matrix and hydrodynamic damping matrix
of the vessel. The control inputs are the surge force τu and the yaw torque τr . The
hovercraft’s model is also written in the matrix form:
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ ucos(ψ) − vsin(ψ) 00
⎜ ẏ ⎟ ⎜ usin(ψ) + vcos(ψ) ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ 
⎜ψ̇ ⎟ ⎜ r ⎟ ⎜0 0⎟ τu
⎜ ⎟=⎜ ⎟+⎜ ⎟ (2.75)
⎜ u̇ ⎟ ⎜ vr ⎟ ⎜1 0⎟ τr
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ v̇ ⎠ ⎝ −ur − βv ⎠ ⎝0 0⎠
ṙ 0 01

or equivalently, one has the description

x̃˙ = f˜(x̃) + g̃(x̃)ṽ (2.76)

The system’s state vector is denoted as x̃ = [x, y, ψ, u, v, r ]T , f (x̃) ∈ R 6×1 , and

g̃(x̃) = [g̃a , g̃b ] ∈ R 6×2 , while the control input is the vector ṽ = [τu , τr ]T .
The system’s state vector can be extended by including as additional state variables
the control input τu and its first derivative τ̇u . These are denoted as z 1 = τu and
z 2 = τ̇u . The extended state-space description of the system becomes
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ ucos(ψ) − vsin(ψ) 0 0
⎜ ẏ ⎟ ⎜ usin(ψ) + vcos(ψ) ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ψ̇ ⎟ ⎜ r ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ 
⎜ u̇ ⎟ ⎜ vr + z 1 ⎟ ⎜0 0⎟
⎜ ⎟=⎜ ⎟+⎜ ⎟ τ̈u (2.77)
⎜ v̇ ⎟ ⎜ −ur − βv ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ τr
⎜ ṙ ⎟ ⎜ 0 ⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ż 1 ⎠ ⎝ z2 ⎠ ⎝0 0⎠
ż 2 0 1 0

or equivalently, one has the description

ż = f (z) + g(z)ṽ (2.78)

The extended system’s state vector is denoted as z = [x, y, ψ, u, v, r, z 1 , z 2 ]T . More-

over, one has f (z) ∈ R 8×1 , and g(z) = [ga , gb ] ∈ R 8×2 , while the control input is
the vector ṽ = [τ̈u , τr ]T .
It can be proven that the model of the underactuated vessel given in Eq. (2.74) is a
differentially flat one. This means that all its state variables and the control inputs can
be written as functions of a single variable, which is the flat output. In the hovercraft’s
case, the flat output is the vector of the vessel’s cartesian coordinates, that is

ỹ = [y1 , y2 ] = [x, y] (2.79)

64 2 Differential Flatness Theory and Flatness-Based Control

It holds that

ẍ = u̇cos(ψ) − u · sin(ψ) · ψ̇ − v̇sin(ψ) − v · cos(ψ)ψ̇

(2.80)
ÿ = u̇sin(ψ) + u · cos(ψ) · ψ̇ + v̇cos(ψ) − v · cos(ψ)ψ̇

Moreover, it holds that

ẍ + β ẋ = cos(ψ)(u̇ − vψ̇ + βu) + sin(ψ)(−u ψ̇ − v̇ − βv)

(2.81)
ÿ + β ẏ = cos(ψ)(v̇ + u ψ̇ + βv) + sin(ψ)(−vψ̇ + u̇ + βu)

Using Eqs. (2.81) and (6.166), and after computing that

u ψ̇ + v̇ + βv = u · r − ur − βv + βv = 0
(2.82)
u̇ − vψ̇ + βu = vr + τu − vr + βu = τu + βu

one obtains
ÿ+β ẏ cos(ψ)0+sin(ψ)(τu +βu)
ẍ+β ẋ = cos(ψ)(τu +βu)−sin(ψ)0 →
(2.83)
ÿ+β ẏ
ẍ+β ẋ = tan(ψ) → ψ = atan −1 ( ẍ+β
ÿ+β ẏ
ẋ )

Thus, through Eq. (2.83) it is proven that the state variable ψ (heading angle of the
vessel) is a function of the flat output and of its derivatives.
From Eq. (2.81) one also has that

(ẍ + β ẋ)2 + ( ÿ + β ẏ)2 = (τu + βu)2 (2.84)

Moreover, it holds that

ẋ(ẍ + β ẋ) = (ucos(ψ) − vsin(ψ))cos(ψ)(τu + βu)

(2.85)
ẏ( ÿ + β ẏ) = (usin(ψ) + vcos(ψ))sin(ψ)(τu + βu)

while using Eq. (2.84) and after intermediate computations one finally obtains

ẋ(ẍ + β ẋ) + ẏ( ÿ + β ẏ) = u(τu + βu) (2.86)

Dividing Eq. (2.86) with the square root of Eq. (2.84) one obtains

u(τu +βu)
√ẋ(ẍ+β ẋ)+ ẏ( ÿ+β ẏ) = (τu +βu) (2.87)
(ẍ+β ẋ) +( ÿ+β ẏ)
2 2

which finally give

u = √ẋ(ẍ+β ẋ)+ ẏ( ÿ+β ẏ)
(2.88)
(ẍ+β ẋ) +( ÿ+β ẏ)
2 2
2.3 Properties of Differentially Flat Systems 65

It also holds that

ẏ ẍ − ẋ ÿ = (usin(ψ) + vcos(ψ))(u̇cos(ψ) − usin(ψ)ψ̇−

−v̇sin(ψ) − vcos(ψ)ψ̇) − (ucos(ψ) − vsin(ψ))(u̇sin(ψ) + ucos(ψ)ψ̇+
+v̇cos(ψ) − vsin(ψ)ψ̇)
(2.89)

which after intermediate computations and substitution of the derivative variables

from Eq. (2.75) give
ẏ ẍ − ẋ ÿ = v(βu + τu ) (2.90)

From Eqs. (2.90) and (2.84) one obtains

ẏ ẍ − ẋ ÿ
v=  (2.91)
(ẍ + β ẋ)2 + ÿ + β ẏ)2

From the state-space equations it holds that

r = ψ̇ (2.92)

where from Eq. (2.83) one has that

ÿ + β ẏ
ψ = atan −1 ( ) (2.93)
ẍ + β ẋ

which means that r is also a function of the flat output and of its derivatives. This
can be also confirmed analytically. Indeed from Eq. (2.93) it holds that

cos 2 (ψ)ψ̇ + sin 2 (ψ)ψ̇ (y (3) + β ψ̈)(ẍ + β ẋ) − ( ÿ + β ẏ)(x (3) + β ẍ)
=
cos 2 (ψ) (ẍ + β ẋ)2
(2.94)
which also gives

ψ̇ (y (3) + β ψ̈)(ẍ + β ẋ) − ( ÿ + β ẏ)(x (3) + β ẍ)

= (2.95)
cos 2 (ψ) (ẍ + β ẋ)2

while using that

1
= tan 2 (ψ) + 1 (2.96)
cos2 ψ

one obtains that

(ẍ + β ẋ)2
cos2 ψ = (2.97)
(ẍ + β ẋ)2 + ( ÿ + β ẏ)2
66 2 Differential Flatness Theory and Flatness-Based Control

Thus, from Eqs. (2.95) and (2.92) one has that

(y (3) + β ψ̈)(ẍ + β ẋ) − ( ÿ + β ẏ)(x (3) + β ẍ)

r = ψ̇ ⇒ r = cos 2 (ψ) (2.98)
(ẍ + β ẋ)2

which after intermediate operations gives

y (3) (ẍ + β ẋ) − x (3) ( ÿ + β ẏ) − β 2 (ẍ ẏ − ÿ ẋ)

r= (2.99)
(ẍ + β ẋ)2 + ( ÿ + β ẏ)2

Equivalently, from the state-space equations one has that


τu = u̇ − v · r ⇒ τu = d
dt
√ẋ(ẍ+β ẋ)+ ẏ( ÿ+β ẏ) −
(ẍ+β ẋ)2 +( ÿ+β ẏ)2
(2.100)
y (3) (ẍ+β ẋ)−x (3) ( ÿ+β ẏ)−β 2 (ẍ ẏ− ÿ ẋ)
−√ ẏ ẍ−ẋ ÿ
· (ẍ+β ẋ)2 +( ÿ+β ẏ)2
(ẍ+β ẋ)2 +( ÿ+β ẏ)2

which after intermediate operations gives

τu = √ẍ(ẍ+β ẋ)+ ÿ( ÿ+β ẏ)

(2.101)
(ẍ+β ẋ) +( ÿ+β ẏ)
2 2

Finally, for the control input τr it holds that τr = ṙ and using Eq. (2.99) this implies
that τr is also a function of the flat output and of its derivatives. This can be also
shown analytically according to the following:

τr = ṙ ⇒ τr =

y (4) (ẍ+βx)−x (4) ( ÿ+β ẏ)+β(y (3) ẍ−x (3) ÿ)−β 2 (x (3) ẏ−y (3) ẋ)
[(ẍ+β ẋ)2 +( ÿ+β ẏ)2 ]· (2.102)
(3)
−2 [y (ẍ+β ẋ)−x 3 ( ÿ+β ẏ)−β 2 (ẍ ẏ− ÿ ẋ)]
[(ẍ+β ẋ)2 +( ÿ+β ẏ)2 ]2
·
·{(ẍ + β ẋ)(x + β ẍ) + ( ÿ + β ẏ)(y (3) + β ÿ)}
(3)

Through Eq. (2.102) it is confirmed that that all state variables and the control input
of the hovercraft’s model can be written as functions of the flat output and of its
derivatives. Consequently, the vessel’s model is a differential flat one.
Next, it will be shown that a flatness-based controller can be developed for the
hovercraft’s model. It has been shown that it holds

ẍ = u̇cos(ψ) − usin(ψ)ψ̇ − v̇sin(ψ) − vcos(ψ)ψ̇ → ẍ = (vr + τu )cos(ψ) −

usin(ψ)r − (−ur − βv)sin(ψ) − vcos(ψ)r → ẍ = τu cos(ψ) + βvsin(ψ)
2.3 Properties of Differentially Flat Systems 67

By differentiating once more with respect to time and after intermediate operations
one finally obtains
x (3) = τ̇u cos(ψ) − τu sin(ψ)r +
(2.103)
+β(−ur − βv)sin(ψ) + βvcos(ψ)r

Similarly one has

ÿ = u̇sin(ψ) + ucos(ψ)ψ̇ + v̇cos(ψ) − vsin(ψ)ψ̇ → ÿ = (vr + τu )sin(ψ) +

ucos(ψ)r + (−ur − βv)cos(ψ) − vsin(ψ)r → ÿ = τu sin(ψ) − βvcos(ψ)

As explained in Eq. (2.104), the state vector of the system is extended so as to include
as new state variables the control input τu and its first derivative τ̇u . The new state
variables are denoted as z 1 = τu and z˙1 = τ̇u . Using that the extended state-space
description of the system is
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ ucos(ψ) − vsin(ψ) 0 0
⎜ ẏ ⎟ ⎜ usin(ψ) + vcos(ψ) ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ψ̇ ⎟ ⎜ r ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ 
⎜ u̇ ⎟ ⎜ vr + z 1 ⎟ ⎜0 0⎟
⎜ ⎟=⎜ ⎟+⎜ ⎟ τ̈u (2.104)
⎜ v̇ ⎟ ⎜ −ur − βv ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ τr
⎜ ṙ ⎟ ⎜ 0 ⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ż 1 ⎠ ⎝ z2 ⎠ ⎝0 0⎠
ż 2 0 1 0

or equivalently, one has the description

ż = f (z) + g(z)ṽ (2.105)

The system’s state vector is again denoted as z = [x, y, ψ, u, v, r, z 1 , z 2 ]T , f (z) ∈

R 8×1 , and g(z) = [ga , gb ] ∈ R 8×2 , while the control input is the vector ṽ =
[τ̈u , τr ]T .
The extended state-space description of the system given in Eq. (2.104) or in its
compact form described by Eq. (2.105), is used. By differentiating once more with
respect to time and after intermediate operations one finally obtains

y (3) = z 2 sin(ψ) + z 1 cos(ψ)r + βur cos(ψ)+

(2.106)
+β 2 vcos(ψ) + βvsin(ψ)r

It can be noticed that in the equations of the third-order derivatives for both x and
y only the control input τu and its derivative τ̇u appear, while the control input τr
is missing. Therefore, differentiation of x (3) once more with respect to time is per-
formed. This gives
68 2 Differential Flatness Theory and Flatness-Based Control

x (4) = τ̈u cos(ψ) − 2z 2 sin(ψ)r − z 1 cos(ψ)r 2 − z 1 sin(ψ)τr − βvr 2 sin(ψ) −

βz 1rsin(ψ) − βuτr sin(ψ) − βur 2 cos(ψ) + β 2 ur sin(ψ) − β 3 vsin(ψ) − β 2 vr cos
(ψ) − βur 2 cos(ψ) + β 2 vr cos(ψ) − βvr 2 sin(ψ) + βvcos(ψ)τr

while after substituting the time derivative according to Eq. (2.74) and after regroup-
ing terms one obtains a description of the form

x (4) = [−2z 2 sin(ψ)r − z 1 cos(ψ)r 2 − βvr 2 sin(ψ) − βz 1rsin(ψ) − βur 2 cos(ψ)+

+β 2 ur sin(ψ) − β 3 vsin(ψ) − β 2 vr cos(ψ) − βur 2 cos(ψ) + β 2 vr cos(ψ)−
−βvr 2 sinψ] + [cos(ψ)]τ̈u + [−z 1 sin(ψ) − βusin(ψ) + βvcos(ψ)]τr
(2.107)

Consequently, the fourth derivative of x is finally written in the form

x (4) = L 4f y1 + L ga L 3f y1 τ̈u + L gb L 3f y1 τr (2.108)

where

L 4f y1 = −2z 2 sin(ψ)r − z 1 cos(ψ)r 2 − βvr 2 sin(ψ) − βz 1 rsin(ψ)−

−βur 2 cos(ψ) + β 2 ur sin(ψ) − β 3 vsin(ψ) − β 2 vr cos(ψ) − βur 2 cos(ψ)+
+β 2 vr cos(ψ) − βvr 2 sinψ
(2.109)

L ga L 3f y1 = cos(ψ) (2.110)

L gb L 3f y1 = −z 1 sin(ψ) − βusin(ψ) + βvcos(ψ) (2.111)

In a similar manner, differentiating once more with respect to time the expression
about y (3) one gets

y (4) = ż 1 cos(ψ)r − z 1 sin(ψ)ψ̇r + z 1 cos(ψ)ṙ −

−β v̇r sin(ψ) − βvṙ sin(ψ) − βvr cos(ψ)ψ̇−
−β u̇r cos(ψ) − βu ṙ cos(ψ) + βur sin(ψ)ψ̇+ (2.112)
+β 2 v̇cos(ψ) − β 2 vsinψ ψ̇+
+ż 2 sin(ψ) + z 2 cos(ψ)ψ̇

while after substituting the time derivative according to Eq. (6.166) and after regroup-
ing terms one obtains a description of the form

y (4) = [z 2 r cos(ψ) − z 1 r 2 sin(ψ) + βur 2 sin(ψ) + β 2 vr sin(ψ) − βvr 2 cos(ψ)]−

−βvr 2 cos(ψ) − βz 1 r cos(ψ) + βur 2 sin(ψ) − βur cos(ψ) + β 2 vcos(ψ)−
−β vr sin(ψ) + z 2 r cos(ψ)] + [sin(ψ)]τ̈u + [z 1 cos(ψ) − βvsin(ψ) − βucos(ψ)]τr
2

(2.113)
2.3 Properties of Differentially Flat Systems 69

Thus y (4) can be also written in the form

y (4) = L 4f y2 + L ga L 3f y2 τ̈u + L gb L 3f y2 τr (2.114)

L 4f y2 = [z 2 r cos(ψ) − z 1r 2 sin(ψ) + βur 2 sin(ψ) − β 2 vr sin(ψ) − βvr 2 cos(ψ)]−

−βvr 2 cos(ψ) − βz 1r cos(ψ) + βur 2 sin(ψ) − βur cos(ψ) + β 2 vcos(ψ)−
−β 2 vr sin(ψ) + z 2 r cos(ψ)]
(2.115)
and
L ga L 3f y2 = sin(ψ) (2.116)

L gb L 3f y2 = z 1 cos(ψ) − βvsin(ψ) − βucos(ψ) (2.117)

Consequently, the aggregate input–output linearized description of the system

becomes
x (4) = L 4f y1 + L ga L 3f y1 τ̈u + L gb L 3f y1 τr
(2.118)
y (4) = L 4f y2 + L ga L 3f y2 τ̈u + L gb L 3f y2 τr

while by defining the new control inputs

v1 = L 4f y1 + L ga L 3f y1 τ̈u + L gb L 3f y1 τr
(2.119)
v2 = L 4f y2 + L ga L 3f y2 τ̈u + L gb L 3f y2 τr

the following description for the input–output linearized hovercraft model is obtained

x (4) = v1
(2.120)
y (4) = v2

For the dynamics of the linearized equivalent model of the system, the following
new state variables can be defined

z 1,1 = x z 1,2 = ẋ z 1,3 = ẍ z 1,4 = x (3)

(2.121)
z 2,1 = y z 2,2 = ẏ z 2,3 = ÿ z 2,4 = y (3)

and the state-space description of the system becomes

ż = Az + Bv
(2.122)
zm = C z
70 2 Differential Flatness Theory and Flatness-Based Control

or equivalently
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1,1 0 1 0 0 0 0 0 0 z 1,1 0 0
⎜ż 1,2 ⎟ ⎜0 0 1 0 0 0 0 0⎟ ⎜z 1,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ż 1,3 ⎟ ⎜0 0 0 1 0 0 0 0⎟ ⎜ ⎟ ⎜ 0⎟
⎜ ⎟ ⎜ ⎟ ⎜z 1,3 ⎟ ⎜0 ⎟ 
⎜ż 1,4 ⎟ ⎜0 0 0 0 0 0 0 ⎟ ⎜
0⎟ ⎜z 1,4 ⎟ ⎜ 0⎟
⎜ ⎟ ⎜ ⎟ + ⎜1 ⎟ v1
⎜ż 2,1 ⎟ = ⎜0 0 0 0 0 1 0 ⎟ ⎜
0⎟ ⎜z 2,1 ⎟ ⎜
⎟ 0⎟
(2.123)
⎜ ⎟ ⎜ ⎜0 ⎟ v2
⎜ż 2,2 ⎟ ⎜0 0 0 0 0 0 1 0⎟ ⎜z 2,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ż 2,3 ⎠ ⎝0 0 0 0 0 0 0 1⎠ ⎝z 2,3 ⎠ ⎝0 0⎠
ż 2,4 0 0 0 0 0 0 0 0 z 2,4 0 1

while the associated measurement equation is

⎞ ⎛
z 1,1
⎜z 1,2 ⎟
⎜ ⎟
⎜ ⎟
 m   ⎜z 1,3 ⎟
z1 1 0 0 0 0 0 0 0 ⎜z 1,4 ⎟
⎜
⎟
= (2.124)
m
z2 00001000 ⎜ ⎜z 2,1 ⎟
⎟
⎜z 2,2 ⎟
⎜ ⎟
⎝z 2,3 ⎠
z 2,4

This completes Example 2.

An important property of the endogenous transformations is that the number of
control inputs in the initial and the final description of the system remains the same.
Theorem: If the two systems (X × U × Rm ∞ , F) and Y × V × R ∞ , G are equivalent,
s
then they will have the same number of control inputs, that is m = s.
Proof : Consider the truncation Φm of Φ on X × U × (R m )μ

Φμ : X × U × (R m+k )μ → Y × V × (R s )μ
(2.125)
(x, u, u 1 , . . . , u k+μ ) → (φ, α, α̇, . . . , α (μ) )

which means the first μ + 2 blocks of the components of Ψ , while μ is a sufficiently

large entity. Since Ψ is invertible, Ψmu is a submersion for all μ. Thus the dimension
of the space from which Ψ has emanated is superior or equal to the dimension of the
space of arrival. This is written as

n + m(k + μ + 1) ≥ s(μ + 1) ∀μ > 0 (2.126)

This implies m ≥ s. In a similar manner, it is shown that with the mapping Ψ it

holds s ≥ m. Consequently, one finally has m = s which means that the number of
control inputs in the two descriptions remains the same.
This definition of the equivalence is adapted to the equivalence between two so-
called diffieties. As it will be explained in detail in Sect. 2.6, the concept of a diffiety
2.3 Properties of Differentially Flat Systems 71

is as follows: Variable (X, τx ) comprising the state vector of the nonlinear system
and its successive derivatives with respect to time, as well as the Lie derivatives of
the mapping transformation along a vector field defined by the state vector elements
and their derivatives is called a manifold of jets or a diffiety.
Given two diffieties (M, CTM) and (N , CTN) it is said that a function Ψ from
M to N is Lie-Backlünd or a C-morphism, if the tangent (projection) function T Ψ
satisfies T Φ(CTM) ⊂ CTN. Moreover, if Ψ has an inverse function Φ, such that
T Ψ (C T N ) ⊂ CTM, then Ψ is a Lie-Backlünd isomorphism or a C isomorphism.
When such an isomorphism exists, the diffieties are called equivalent. Therefore,
an endogenous transformation is a special type of a Lie-Bakclünd transformation
which preserves the parametrization in time of the integral curves. It is also possible
to introduce a more general concept which is the orbital equivalence, about the Lie-
Bakclünd isomorphisms which preserve only the geometric locus of the integral
curves.

2.3.1.3 Differentially Flat Systems and Equivalence to the Trivial System

Definition: The dynamical system (M, F) where M: is the state vector and F: is a
vector field, is differentially flat in p (or respectively flat), if and only if it is equivalent
in p (respectively equivalent) with the trivial system (therefore it can be described
in the Brunovsky canonical form, as shown in the examples of Sect. 2.2.2).
The definition of equivalent systems can be used, and one can consider a differ-
entially flat system (X × U × Rm ∞ , F) associated with

ẋ = f (x, u) where (x, u) ∈ X × U ⊂ R n × R m (2.127)

By definition, the above system is equivalent to the trivial system Rn∞ , Fs , where the
endogenous transformation Ψ takes the form

Ψ (x, u, u (1) , . . .) = (h(x, u), ḣ(x, u), ḧ(x, u), . . .) (2.128)

In other terms, Ψ is the infinite extension of function h(). The inverse transformation
of Ψ is denoted as Φ and holds

Ψ ( ȳ) = (ψ(ψ̄), β(ψ̄), β̇(ψ̄) . . .) (2.129)

Since Φ and Ψ are inverse applications it holds that

φ(h̄(x, ū)) = x and α(h̄(x, ū)) = u (2.130)

Moreover, t → y(t) is a trajectory of y = v and

t → (x(t), u(t)) = (ψ(ψ̄(t)), β(ψ̄(t))) (2.131)

is a trajectory of ẋ = f (x, u).

72 2 Differential Flatness Theory and Flatness-Based Control

Definition: Assume that {M, F} is a system and that Ψ is the transformation that
brings it to the canonical form. The first component of Ψ is considered to be the flat
output.
Another important property of differentially flat systems is that the dimension of
the flat output is equal to the number of the control inputs (and thus linearization
and decoupling is succeeded). However, it is noted that the flat output is not unique.
If y is a flat output, then γ (y) is also a flat output, provided that γ is a smooth
diffeomorphism. For single-input systems, the flat output is a scalar s = m = 1.
It is easy to show that one can pass from one flat output to another by using only
a static change of variables. The multivariable case is clearly more complicated.
For example, for s = m = 2 it is easy to see that if y1 , y2 is a flat output then
(y1 , y2 + γ (y1(r ) )) is also a flat output, provided that γ () is smooth function and
r ≥ 0. Since the flat output of a system is not unique, it is preferable to select the flat
output which leads to simple computation for the linearization of the system and the
design of the feedback controller.

2.3.2 Differential Flatness and Trajectory Planning

2.3.2.1 Applications of Differential Flatness in Trajectory Planning

Next, it will be shown how differential flatness can be used to solve the problem
of trajectory planning (setpoints definition) for dynamical systems control. Assume
the following nonlinear system in which the control input u does not implement
feedback control
ẋ = f (x, u), x ∈ R n u ∈ R m (2.132)

with flat output y = h(x, u, u̇, . . . , u (r ) ). In such a manner, the system’s trajec-
tories (x(t), u(t)) are written as a function of the flat output y and its derivatives
x = φ(y, ẏ, . . . , y (q) ) and u = α(y, ẏ, . . . , y (q) ). The trajectory planning problem
contains a start and an arrival state. It is assumed that the flat output is a vector with
elements yi which are written as

yi (t) = Ai j λ j (t) (2.133)
j

where λ j , j = 1, . . . , N are basis functions. The problem of definition of the flat

output components yi becomes equivalent to finding its projections in the space
spanned by the basis functions λi .
2.3 Properties of Differentially Flat Systems 73

Next, it is assumed that the initial state of the system is x0 at t0 , while the final
state of the system is x f at t f . The coefficients Ai j should satisfy the following
 
yi (t0 ) = j Ai j λ j (t0 ) yi (t f ) = j Ai j λ j (t f )
··· ··· ··· (2.134)
(q)  (q) (q)  (q)
yi (t0 ) = j Ai j λ j (t0 ) yi (t f ) = j Ai j λ j (t f )

Without loss of generality it assumed that the vector of the flat output is of dimension
i = 1, that is y = [y1 ] (these results can be also generalized to the multidimensional
case). The following vectors are defined

(q)
ỹ0 = (y1 (τ0 ), . . . , y1 (τ0 ))
(q)
ỹ f = (y1 (τ f ), . . . , y1 (τ f )) (2.135)
ỹ = ( ỹ0 , ỹ f )

Using Eqs. (2.135) and (2.134) one has that

 
Λ(τ0 )
ỹ = A = ΛA (2.136)
Λ(τ j )

The elements of A should satisfy Eq. (2.136). The only condition for Eq. (2.136) to
have a solution, is that matrix Λ is a full rank one. This means that the space of
functions λ j has to be sufficiently rich. It can be concluded that the path planning
problem for the case of differentially flat systems ends up to linear algebra theory.

2.3.2.2 Planning Under Constraints

In this case, the objective is to find a trajectory going from point a to point b, which
will satisfy the constraints K (x, u, . . . , u (r ) ) ≤ 0 at each time instant. In the flat
coordinates, this consists in finding T > 0 and a function [0, T ] t → y(t) with
(y, . . . , y (q) ) to be given at t = 0 and T and to verify that ∀t ∈ [0, T ] the constraint
K (y, . . . , y (v) )(t) ≤ 0 will hold, for a certain v and a specific function K , obtained
from k, φ, and α, where k is the constraint condition and φ, α are the previously
defined mappings ẋ = φ(y, v̄) and u = α(y, v̄). The difficulty of this problem
increases when q = v = 0.
Next, it is simply assumed that the initial state Y0 and the final state Y f are equi-
libria. It is also assumed that the motion from the initial to the final state satisfies the
following constraint: there exists a path [0, 1] σ → Y (σ ) such that Y (0) and Y (1)
correspond to start and arrival equilibria , for all σ ∈ [0, 1], K (Y (σ ), 0, . . . , 0) < 0.
Therefore, there exists T > 0 and [0, T ] t → y(t) which is a solution to the
problem. It suffices to take Y (η(t/T )) where T is sufficiently large, and with η to be
a smooth increasing function [0, 1] s → η(s) ∈ [0, 1] with η(0) = 0, η(1) = 1,
and ddsηi = 0 for i = 1, . . . , max(q, v).
i
74 2 Differential Flatness Theory and Flatness-Based Control

The problem becomes more complicated in the case of trajectory planning for
systems subjected to nonholonomic constraints.

2.3.2.3 Planning of Trajectories with Singularities

With the previous analysis it has been shown that the endogenous transformation

Ψ (x, u, u (1) , . . .) = (h(x, ū), ḣ(x, ū), ḧ(x, ū), . . .) (2.137)

associated with the flat output y = h(x, ū) is defined everywhere, is smooth and
invertible in a manner that always enables state variable x and control input u to be
expressed as functions of the flat output and its derivatives

(y, ẏ, . . . , y (q) ) → (x, u) = φ(y, ẏ, . . . , y (q) ) (2.138)

However, a point of singularity may exist in the area of the state-space into which
the control tries to bring the system’s state vector (y is no longer invertible and
Eq. (2.138)) cannot be solved. As function φ cannot be defined in such a point, the
previous computations become meaningless. A manner to circumvent this problem
and to avoid the singularity is to choose a trajectory for the system

t → φ(y(t), ẏ(t), . . . , y (q) (t)) (2.139)

such that the point of potential singularity is bypassed. This procedure is depicted in
the following example:
Example: The following system with differentially flat dynamics is considered:

ẋ1 = u 1 ẋ2 = u 2 u 1 ẋ3 = x2 u 1 (2.140)

where the flat output is defined as y = (x1 , x3 ). In case that u 1 = 0 it holds that
ẋ1 = ẏ1 = 0 and a singularity arises because it holds
 
φ ẏ2 ÿ2 ẏ1 − ÿ1 ẏ2
(y, ẏ, ÿ) (x , x , x , u , u ) = y1 , ẏ1 , y2 , ẏ1 , (2.141)
→ 1 2 3 1 2 ẏ 3
1

and the inverse transformation φ is not defined at the points of singularity. However,
if one considers trajectories of the following form t → y(t) := (σ (t), p(σ (t))),
with σ and π being smooth functions he obtains that

y2 (t) p(σ (t))

= (2.142)
y1 (t) σ (t)

dp
ẏ2 (t) dσ (σ (t)) · σ̇ (t)
= (2.143)
ẏ1 (t) σ̇ (t)
2.3 Properties of Differentially Flat Systems 75

and
d2 p
ÿ2 ẏ1 − ÿ1 ẏ2 dσ 2
(σ (t)) · σ̇ 3 (t)
= (2.144)
ẏ13 σ̇ 3 (t)

Therefore, one can extend t → φ(y(t), ẏ(t), ÿ(t)) everywhere in time using
 
dp d2 p
t → σ (t), (σ (t)), p(σ (t)), σ̇ (t), (σ (t)) (2.145)
dσ dσ 2

Trajectory planning is performed in a similar manner as before: actually, functions

σ and p and their derivatives are constrained at the start and at the arrival point of
the trajectory, while elsewhere they are free. In this example, the avoidance of the
singularity is internally related with a symmetry of the system. Equations

ẋ1 = u 1 , ẋ2 = u 2 u 1 , ẋ3 = x2 u 1 (2.146)

are linear in u 1 . The transformation, t → t˜ = σ (t) and u 1 → ũ 1 = σ̇u(t)

1
, where only
the time t and the control input u change, leaves the equations invariants.

2.3.3 Differential Flatness, Feedback Control

and Equivalence

2.3.3.1 Closed-Loop Systems Under State Feedback Control

and Equivalence

The analysis of Sect. 2.3.2 referred to a system’s dynamics ẋ = f (x, u) in which

u was not specifically implementing a feedback control action. Next, the notion of
equivalence will be extended to the case of closed-loop control. To this end, the
initial system dynamics ẋ = f (x, u) and its equivalent description ẏ = g(y, v) are
considered
ẋ = f (x, u), (x, u) ∈ X × U ⊂ R n × R m
(2.147)
ẏ = g(y, v), (y, v) ∈ Y × V ⊂ R r × R p

∞ , F) and (Y × V × R ∞ , G), where

These correspond to the systems (X × U × Rm s
F and G are defined by

F(x, u, u (1) , . . .) := ( f (x, u), u (1) , u (2) , . . .)

(2.148)
G(y, v, v(1) , . . .) := (g(y, v), v(1) , v(2) , . . .)

Considering that these systems are equivalent, they have the same trajectories. A
question that arises is if a transition is possible from the system ẋ = f (x, u) to the
76 2 Differential Flatness Theory and Flatness-Based Control

system ẏ = g(y, v) by a feedback loop (primarily dynamic or alternatively static),

or inversely. For the transformed state vectors z and control inputs v one has

ż = α(x, z, v) z ∈ Z ⊂ R q
(2.149)
u = κ(x, z, v)

Next, the following theorem is introduced [338]:

Theorem: Assume that the systems ẋ = f (x, u) and ẏ = g(y, v) are equivalent.
Thus ẋ = f (x, u) can be transformed by a dynamic loop and change of coordinates
(dynamic feedback linearization) into

ẏ = g(y, v), v̇ = v(1) , v̇(1) = v(2) , . . . , v̇(μ) = w (2.150)

for an entity μ being sufficiently large. Inversely, ẏ = g(y, v) can be transformed by

a feedback loop (dynamic) and change of coordinates into

ẋ = g(x, u), u̇ = u (1) , u̇ 1 = u (2) , . . . , u̇ ( ) = w (2.151)

for an entity v being sufficiently large.

Proof : The notations F and G are used for the infinite dimensional vector fields
which represent the two systems. The equivalence shows that there exists an invertible
function
Φ(y, v̄) = (φ(y, v̄), α(y, v̄), α̇(y, v̄), . . .) (2.152)

such that
F(Φ(y, v̄)) = DΦ(y, v̄) · G(y, v̄) (2.153)

Next, one sets ȳ = (y, v, v(1) , . . . , v(μ) ) and w = v(μ+1) . For μ sufficiently large,
it holds that φ (respectively α) depends exclusively on ȳ (respectively on ȳ and w).
With these notations Φ is written as

φ( ỹ, w) = (φ( ỹ), α( ỹ, w̄), α̇( ỹ, w̄), . . .) (2.154)

and Eq. (2.153) implies in particular that

f (φ( ỹ, α( ỹ, w))) = Dφ( ỹ)ḡ( ỹ, w) (2.155)

where ḡ := (g, v(1) , . . . , v(k) ). Since Φ is invertible, and φ is of full rank it can thus
be completed by a function π to obtain a change of coordinates

ỹ → φ(ψ̃) = (φ( ỹ), π(ψ̃)) (2.156)

2.3 Properties of Differentially Flat Systems 77

Considering now the dynamic feedback, one has

u = α(φ −1 (x, z), w)

(2.157)
ż = Dπ(φ −1 (x, z)) · g̃(φ −1 (x, z), w)

which transforms ẋ = f (x, u) into

   
ẋ f (x, α(φ −1 (x, z), w))
= f˜(x, z, w) := (2.158)
z Dπ(φ −1 (x, z)) · g̃(φ −1 (x, z), w)

Using the previous example one has

   
f (φ( ỹ, α( ỹ, w))) Dφ( ỹ
f˜(φ( ỹ, w)) = = g̃( ỹ, w) = Dφ( ỹ) · g̃( ỹ, w)
Dπ( ỹ) · g̃( ỹ, w) Dπ( ỹ))
(2.159)

Thus, f˜ and g̃ are φ-conjugates which complete the proof.

A generic remark is that a differentially flat system is equivalent to a system in

the linear canonical (Brunovsky) form, and thus one arrives immediately at the fol-
lowing conclusion: A flat dynamics is linearizable by dynamic feedback and change
of coordinates.

2.3.3.2 Closed-Loop Systems and Equivalence Under Endogenous

Feedback

The dynamical system ẋ = f (x, u) is considered. The feedback loop

u = κ(x, z, w)
(2.160)
ż = α(x, z, w)

is called endogenous since the system’s dynamics in open loop ẋ = f (x, u) is

equivalent to the dynamics of the system in closed loop, the latter given by

ẋ = f (x, κ(x, z, w))

(2.161)
ż = α(x, z, w)

The term endogenous has the meaning that variables z and w are not included in the
state vector of the system (case also known as “zero dynamics”). Using the above one
arrives at a more refined definition of equivalence between systems and differential
flatness.
78 2 Differential Flatness Theory and Flatness-Based Control

Theorem: Two dynamics ẋ = f (x, u) and ẏ = g(y, v) are equivalent, if and only if,
ẋ = f (x, u) can be transformed by endogenous feedback and change of coordinates.

Corollary 1: A dynamics is flat, if and only if it is linearizable by dynamic endoge-

nous feedback and change of coordinates.

Corollary 2: Consider the system description

ẋ = f (x, κ(x, z, w))

(2.162)
ż = α(x, z, w)

where the constituents of the dynamic endogenous feedback are

u̇ = κ(x, z, w)
(2.163)
ż = α(x, z, w)

This system can be transformed by endogenous feedback into

ẋ = f (x, u), where u̇ = u (1) , . . . , u̇ (μ) = w (2.164)

for μ sufficiently large (thus finally the system’s control input contains multiple inte-
gral terms). This shows clearly the properties which are preserved by equivalence.
The property which is maintained by introducing integral terms and change of coor-
dinates is controllability. An endogenous feedback is actually reversible to the effect
of multiple integrators.

2.3.3.3 Trajectory Tracking in Feedback Control of Differentially

Flat Systems

The problem of trajectory tracking for the system ẋ = f (x, u) consists in finding a
controller u which will permit to the system state variables to track the reference set-
point t → (xr (t), u r (t)). It is convenient to add to the open loop control a correction
term Δu, which is a function of the state vector error Δx = x − xr . For differen-
tially flat systems, there is a systematic method to calculate Δu from the tracking
error Δx.
If the dynamics admits as flat output y = h(x, ū), one can use the corollary
that a differentially flat system is linearizable by (dynamic) feedback and change of
coordinates, into the linear canonical form y (μ+1) = w. For the linearized equivalent
(μ+1)
model, the control input is defined as yr − K Δ ỹ, where K is a gains vector and
Δy is a vector having as elements the flat output tracking error and its derivatives
up to order μ. By applying such a feedback control input, the tracking error dynamics
becomes
Δy (μ+1) = −K Δ ỹ (2.165)
2.3 Properties of Differentially Flat Systems 79

where yr (t) := (xr (t), ū r (t)) and ỹ := (y, ẏ, . . . , y μ ), and Δy represents the vector
of the flat output tracking error. Such a feedback control law assures asymptotic
tracking. There exists an invertible transformation

Φ( ȳ) = (φ( ȳ), α( ȳ), α̇( ȳ), . . .) (2.166)

which connects the infinite vector fields F(x, ū) := ( f (x, u), u, u (1) , . . .) and
G( ȳ) := (y, y (1) , . . .). From the proof of the theorem on the equivalence of two
systems (theorem given in Sect. 2.3.3.1), the above means that for the state vector
one has
x = φ( ỹr (t) + Δ ỹ) =
= φ( ỹr (t)) + Rφ (yr (t), Δ ỹ)Δ ỹ = (2.167)
= xr (t) + Rφ (yr (t), Δ ỹ)Δ ỹ

and for the control input one has

u = α( ỹr (t) + Δ ỹ, −K Δỹ) = 

(μ+1) Δ ỹ
= α( ỹr (t)) + Rα (yr (t), Δ ỹ) =
−KΔ ỹ  (2.168)
(μ+1) Δ ỹ
= u r (t) + Rα ( ỹr (t), yr (t), Δ ỹ, Δw)
−K Δ ỹ

where the following classical result of factorization has been used

1
Rφ (Y, ΔY ) := 0 Dφ(Y + tΔY )dt
1 (2.169)
Rα (Y, w, ΔY, Δw) := 0 Dα(Y + tΔY, w + tΔw)dt

Since Δy → 0 for t → ∞, this means that x → xr (t) and u → u r (t).

2.4 Flatness-Based Control and State Feedback for Systems

with Model Uncertainties

Up to now, for the systems used for analyzing differential flatness properties and
flatness-based control it has been assumed that an exact dynamic model was avail-
able. Flatness-based control can be also applied to systems characterized by model
uncertainties and exogenous disturbances. Denoting the coefficients vector of the
system as θ ∈ R p , the generalized system dynamics of Eq. (2.170) is written as

ẋ(t) = f (θ, x(t), u(t)) (2.170)

80 2 Differential Flatness Theory and Flatness-Based Control

In the generic case, the system’s coefficients θ are subject to uncertainty, i.e.,

θ = θ0 + θ̃ , θ̃i ∈ [θ i , θ̄i ], i = 1, . . . , p (2.171)

where θ0 is the nominal value of the parameter vector. In Eq. (2.170) the vector
field f : R p × R n × R m → R n is smooth. Denoting the flat output y(t) ∈ R
and parameters vector θ , the definition of differential flatness given in Sect. 2.2.2 is
rewritten as [189, 190]
y = h(θ, x, u, u̇, . . . , u (r ) )
x = φ(θ, y, ẏ, . . . , y (r −1) ) (2.172)
u = ψ(θ, y, ẏ, . . . , y (r ) )

where h, φ, and ψ are smooth functions defining mappings h : R p ×R n ×(R m )r +1 →

R m , φ : R p × (R m )r → R n and ψ : R p × (R m )r +1 → R m , respectively. It is also
assumed that the flat output y(t) is independent of the coefficients (parameters) vector
θ . Equation (2.172) shows that for every given trajectory of the flat output t → y(t)
the evolution of all other variables of the system t → x(t) and t → u(t) are given
without integration of any differential equation. Moreover, for a sufficiently smooth
desired trajectory of the flat output t → y ∗ (t) Eq. (2.172) can be used to design the
corresponding feed-forward control u ∗ (t) directly for the nominal system parameters
θ0 . The trajectory y ∗ is called the nominal trajectory, while the trajectory u ∗ is called
the nominal control. Thus, knowing the desirable system’s output, one can also find
the associated flat output y ∗ and subsequently the control input that makes the system
track the desirable trajectory is given by

u ∗ (t) = ψ(θ0 , y ∗ (t), ẏ ∗ (t), . . . , y ∗ (r ) (t)) (2.173)

which means that for each admissible nominal trajectory y ∗ (t) there is a nominal
control input u ∗ . The initial condition of the desired trajectory of the flat output
t → y ∗ (t) is defined as the vector y0∗ = [y ∗ (0), ẏ ∗ (0), . . . , y ∗ (r −1) (0)]T . The sys-
tem is consistent with respect to the initial conditions x0 if x0 = φ(θ, y0∗ ). The
following theorem has been proven [189]:

Theorem: If the desired trajectory of the flat output is consistent with the initial con-
dition x0 and θ = θ0 , then when applying the nominal control input of Eq. (2.173)
the system becomes equivalent, by a change of coordinates, to a linear system in
Brunovksy (canonical) form.

Without loss of generality, a single-input single-output flat system is considered.

It is easy to show that such a flat system can be represented as follows: Setting
y = [y, ẏ, . . . , y (n−1) ]T = [y1 , y2 , . . . , yn ]T the system can be transformed via the
diffeomorphism
y = h(θ, x) (2.174)
2.4 Flatness-Based Control and State Feedback for Systems … 81

where h = φ −1 with φ defined in Eq. (2.172), into the control normal (canonical)
form
ẏi (t) = yi+1 (t) i ∈ {1, . . . , n − 1}
(2.175)
ẏn (t) = α(θ, y(t), u(t))

where y (i) is the i − 1th order derivative of the flat output. From the second row of
Eq. (2.175), one can see that the nth order derivative of the flat output is finally written
as a nonlinear function of the parameters vector θ , of the flat output’s vector y(t) and
of the control input u(t). The system in canonical form described by Eq. (2.175) and
the system of Eqs. (2.170), (2.172), and (2.173) have the same solution t → y(t).
Defining the new control input v(t) = α(θ, y(t), u(t)), the Brunovsky form of
the transformed initial system is written as

ẏi (t) = yi+1 (t) i ∈ {1, . . . , n − 1}

(2.176)
ẏn (t) = v(t)

Once the system has been written in the Brunovksy (canonical) form of Eq. (2.176),
a control input that can assure tracking of any desirable trajectory ẏn∗ is given by

v = ẏn∗ + K T ē (2.177)

where the term K T ē corresponds to an error feedbackcontroller with an integral term

t
and thus processes the augmented error vector ē = [ 0 e1 (τ )dτ, e1 , e2 , . . . , en ]. The
feedback control term can be also of the form of a proportional-derivative controller
and in the latter case processes the error vector e = [e1 , e2 , . . . , en ], thus taking the
form
n
KTe = − kn+1−i ei (t), kn+1−i > 0 (2.178)
i=1

Thus, in case that the exact values of the model’s parameters θ are known and these
are equal to the nominal values θ0 , and no integration of the flat output error is
nin the feedback loop, i.e. the feedback control signal is of the form K e =
used T

− i=1 kn+1−i ei (t), the closed-loop system error vector dynamics become

ẏn = v ⇒ ẏn = ẏn∗ + K T e ⇒

ẏn = ẏn∗ − kn e1 − kn−1 e2 + · · · − k1 en ⇒
(2.179)
ẏn − ẏn∗ = −kn e1 − kn−1 e2 + · · · − k1 en ⇒
e + k1 e(n−1) + k2 e(n−2) + · · · + kn−1 ė + kn e = 0
(n)
82 2 Differential Flatness Theory and Flatness-Based Control

∗
where ei = yi − yi∗ = y (i−1) − y (i−1) . The dynamics of the flat output tracking
error can be also written in a matrix form
⎛ ⎞ ⎛ ⎞⎛ ⎞
ė1 0 1 0 ··· 0 e1
⎜ ė2 ⎟ ⎜ 0 0 1 ··· 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ e2 ⎟
⎜· · ·⎟ = ⎜ · · · · · · ··· ··· ··· ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜· · ·⎟ (2.180)
⎝· · ·⎠ ⎝ 0 0 0 · · · 1 ⎠ ⎝· · ·⎠
ėn −kn −kn−1 −kn−2 · · · −k1 en

In case of the error dynamics of Eq. (2.180) to succeed elimination of the tracking
error it suffices to select the controller gains k1 , k2 , . . . , kn so as the polynomial
e(n) + k1 e(n−1) + k2 e(n−2) + · · · + kn−1 ė + kn e = 0 to be a Hurwitz one.
When that the model’s parameters θ are not precisely known, or have deviated
from their nominal values θ0 , one can apply again a control input of the form described
in Eqs. (2.177) and (2.178), and for specific magnitude of the exogenous disturbances
or parametric uncertainties can expect again satisfactory tracking of the reference
trajectory. In such a case, the robustness features of the closed loop can be analyzed
with the use of results from interval polynomial stability theory (e.g. Kharitonov’s
theory) or other results on interval polynomial analysis (e.g. see [431]). The problem
of robustness of a flatness-based controller in case of parametric uncertainties or
exogenous perturbations has been studied in [189, 190, 433, 443].

2.5 Classification of Types of Differentially Flat Systems

2.5.1 Criteria About the Differential Flatness of a System

2.5.1.1 Definition of 0-Flat Systems

The existence of a computable criterion so as to decide if the dynamical system

ẋ = f (x, u), x ∈ R n , u ∈ R m is differentially flat remains open. This means that
there is no systematic method to construct a flat output. This situation is somehow
equivalent to the definition of Lyapunov functions for nonlinear dynamical systems.
For even though Lyapunov functions are both theoretically and practically important
for the study of the stability features of nonlinear dynamical systems, there is no
systematic manner to define them.
The main difficulty in the computation of the flat output y = h(x, u, . . . , y (r ) )
remains in its dependence to the derivatives of the control input u, up to order r
which can be arbitrarily large. To know if the order of r admits an upper bound
(which can be dependent on n and m) remains a question. It is not known if for a
certain dimension of the state vector and for a certain dimension of the control inputs
vector such a finite boundary exists. In the following it will be noted that a dynamical
system is r -flat if it admits a flat output depending up to the r th order derivative of
2.5 Classification of Types of Differentially Flat Systems 83

the control input. Equivalently, in a 0-flat system its flat output depends exclusively
on the elements of the system’s state vector.
The next example shows that such an upper bound should depend, in the best
case, linearly to the dimension of the state vector. Consider the dynamical system
(a1 ) (a2 )
x1 = u 1 x1 = u 2 ẋ3 = u 1 u 2 (2.181)

The flat output is taken to be

α1 i (α1 −i) u (i−1) ,
y1 = x3 + i=1 (−1) x 1 2 y2 = x2 (2.182)

This system is r -flat, with r = min(α1 , α2 ) − 1. There is no flat output depending

on the derivatives of u which has an order equal or less than r − 1.

2.5.1.2 Systems Which Can Be Linearized Static and Dynamic State

Feedback

It has been shown that dynamical systems which can be linearized with static state
feedback can be also transformed to the linear canonical (Brunovsky) form. Thus,
such systems are differentially flat, with the flat outputs to be exclusively functions
of the state vector x (0-flat system). It is noted that a differentially flat system is
not always a system which can be linearized with static state feedback. There exist
systems which can be linearized through dynamic feedback linearization, that is their
state vector is extended by including as additional state variables the control inputs
and their derivatives (see previous examples on the linearization of the VTOL and
the hovercraft model, given in Sect. 2.2.2). Such systems can be also written in the
linear canonical form, and thus are also differentially flat.

2.5.1.3 Systems with Only One Control Input

Single-input dynamical systems, that is

ẋ = f (x, u) x ∈ R n , u ∈ R (2.183)

can be linearized with static state feedback linearization and can be finally written
in the linear canonical form. Such systems are differentially flat.

2.5.1.4 Systems Affine-in-the-Input of Codimension 1

A system of the form

n−1
ẋ = f 0 (x) + g j (x)u j x ∈ R n (2.184)
j=1
84 2 Differential Flatness Theory and Flatness-Based Control

that is systems in which the number of control inputs is by one less than the dimen-
sion of the system’s state vector, are 0-flat, as long as they are controllable (strongly
accessible in x). The problem becomes more complicated when the system is not an
affine-in-the-input one. However, the latter systems can be finally transformed into
affine-in-the-input ones too.

2.5.1.5 Systems Without Drift Term

m
These are systems of the form ẋ = i=1 f i (x)u i . Within this class of systems, one
can distinguish several cases. The following theorem holds [340]

Theorem 1: Systems without drift and with two control inputs

ẋ = f 1 (x)u 1 + f 2 (x)u 2 (2.185)

are flat if and only if, the generic rank of E k is equal to k + 2, for k = 0, . . . , n − 2,
where E 0 = { f 1 , f 2 } and E k+1 = {E k , [E k , E k ]}, k > 0. A dynamical system
without drift and with two control inputs is always 0-flat. Such a system can be
written in the chained form (which is a precursor of the linear canonical form),
through static state feedback and a change of variables. For example, for a system
of the form
ẋ1 = u 1 ẋ2 = u 2 ẋ3 = x2 u 1 · · · ẋn = xn−1 u 1 (2.186)

is a differentially flat one and the flat output is y = (x1 , xn ).

2.5.1.6 Systems in Triangular Form

The state-space description of systems in triangular form is as follows:

ẋ1 = f 1 (x1 ) + g1 (x1 )x2

ẋ2 = f 2 (x1 , x2 ) + g2 (x1 , x2 )x3
ẋ3 = f 3 (x1 , x2 , x3 ) + g3 (x1 , x2 , x3 )x4
···
(2.187)
ẋi = f i (x1 , x2 , . . . , xi ) + gi (x1 , x2 , . . . , xi )xi+1
···
ẋn−1 = f n−1 (x1 , x2 , . . . , xn−1 ) + gn−1 (x1 , x2 , . . . , xn−1 )xn
ẋn = f n (x1 , x2 , . . . , xn ) + gn (x1 , x2 , . . . , xn )u

The flat output for such systems is y = x1 .

2.5 Classification of Types of Differentially Flat Systems 85

Theorem 2: Driftless systems with a state vector of dimension n and n − 2 control

inputs. The system of the form
n−2
ẋ = f i (x)u i (2.188)
i=1

is a differentially flat one if it is controllable (strongly accessible for almost all x).
More precisely, such a system is 0-flat is n is an odd number and is a 1-flat one if n
is an even number.

2.5.2 A Sufficient Condition for Showing

that a System Is Not Differentially Flat

The following theorem provides a sufficient condition for showing that dynamical
systems are not differentially flat [340]:

Theorem: The criterion of the regulated variety. Assume that the dynamical system
ẋ = f (x, u) is a differentially flat one. Then, the projection to the p-dimensional
space of its state-space equation p = f (x, u) is a subvariety regulated for all x.

This criterion signifies that the elimination of the control input u from the n
equations ẋ = f (x, u) leads to n − m equations of the form F(x, ẋ) = 0 with the
following property: For all (x, p) such that F(x, p) = 0 there exists α ∈ R n , α = 0
such that
∀λ ∈ R F(x, p + λa) = 0 (2.189)

Proof : Consider (x̄, ū, p̄) such that p̄ = f (x̄, ū). Generally, f is of the order m =
dim(u) with respect to u. If x is removed from the state-space description then one
arrives at n−m equations which are based exclusively on x and p, that is F(x, p) = 0.
Thus one has the constraint about the trajectory around (x̄, ū) : F(x, ẋ) = 0. If the
system is differentially flat it holds that

x = φ(y, ẏ, . . . , y (q) ) (2.190)

˙ . . . , ȳ (q) ) such that x̄ =

where variable y is the flat output. Thus, there exists ( ȳ, ȳ,
˙ (q)
φ( ȳ, ȳ, . . . , ȳ ). Then, the following identity holds

˙ . . . , ȳ (q) ), φ y ẏ + φ ẏ ÿ + · · · + φ y (q) y (q+1) ) = 0

F(φ( ȳ, ȳ, (2.191)
86 2 Differential Flatness Theory and Flatness-Based Control

Taking now that the derivatives of function y at t = 0 up to order q are given by

y (r ) (0) = ȳ (r ) r = 0, . . . , q and y α (0) = ȳ q + ξ where ξ is an arbitrary vector in
R n , one has the following identity for all ξ ∈ R n

F(x̄, x̄ + φ y α ξ ) = 0 (2.192)

(q) ˙ . . . , ȳ (q) . From the point (x̄, p̄) of F(x, p) = 0

with φ y to be evaluated at ȳ, ȳ,
passes the affine space which is parallel to the image of φ y (q) , that is a space of
dimension at most equal to m and at least equal to 1.
Therefore by showing that there does not exist any flat output with derivatives
which are not coupled in the sense of an ODE (or equivalently it can be said that
there does not exist a flat output being differentially independent), one can prove
that a dynamical system is not differentially flat. This result is particularly simple.
It stands, however, for an efficient method to show that certain multi-input systems
are not differentially flat.
Example 1: The system

ẋ1 = u 1 , ẋ2 = u 2 , ẋ3 = (u 1 )2 + (u 2 )3 (2.193)

is not a differentially flat one. This is because the subvariety p3 = p12 + p23 is not a
regulated one. There does not exist an α ∈ R 3 , α = 0 such that

∀λ ∈ R p3 + λα3 = ( p1 + λα1 )2 + ( p2 + λα2 )3 (2.194)

Actually, the cubic term implies that α2 = 0 and the quadratic term implies that
α1 = 0 and thus one has α3 = 0.

2.5.3 Liouvillian and Nondifferentially Flat Systems

There are some nonflat systems which do not satisfy differential flatness properties.
Such systems cannot be linearized by state feedback (neither by static feedback
linearization nor by dynamic feedback linearization). For such cases, the notion
of Liouvillian systems has been introduced, which can be seen as an extension of
differentially flat systems.
Nonflat systems are algebraically characterized by an integer called the defect.
This integer is well defined by introducing the notion of maximal linearizing output.
Using an informal definition, a set of variables y = (y1 , . . . , ym ) is a maximal
linearizing output if the number of system variables which can be expressed as a
differential function of y is maximal. Therefore, the maximal linearizing output
characterizes the largest subsystem which satisfies the differential flatness property.
This largest subsystem is called a flat subsystem.
2.5 Classification of Types of Differentially Flat Systems 87

The defect represents the number of the system’s variables which do not belong
to the flat subsystem. A nonlinear system is said to be Liouvillian or integrable by
quadratures if the elements which do not belong to the flat subsystem can be obtained
by elementary integrations of the elements of the flat subsystem. In the following
examples, the notion of integrability by quadratures and that of Liouvillian systems
is explained [340]

Example 1:

ẋ1 = x2 + x32 , ẋ2 = x3 , ẋ3 = u (2.195)

This system is not linearizable by static or dynamic state feedback. The system is of
defect equal to 1 and is found to be Liouvillian (there is a state variable which does
not belong to the flat subsystem). A maximum linearizing output is given by y = x2
and x1 is the only variable that does not belong
 to the flat subsystem. However, x1
can be obtained from the integration x1 = y + ẏ 2 , that is x1 is given by a pure
integral of a differential function of y.

Example 2: The following dynamic model is considered

ẋ1 = u 1 , ẋ2 = u 2 , ẋ3 = x1 u 2 − x2 u 1 , ẋ4 = x3 u 1 ẋ5 = x3 u 2 (2.196)

This system is nilpotent and completely controllable since it is driftless and sat-
isfies the Lie algebra rank condition. However, the system is not flat and can-
not be transformed into the canonical form. Nevertheless, there is a subsystem
within it defined by the first four equations, which is differentially flat and con-
sequently the system’s defect is equal to 1. The maximal linearizing output of
the system is given by y = (y1 , y2 ), where y1 = x1 and y2 = (zx1 − x4 )/3,
with z to be defined by z = (x1 x2 − x3 )/2. Indeed z = ẏ2 / ẏ1 , thus x4 =
y1 z − 3y2 = α(y1 , ẏ1 , y2 , ẏ2 ), x3 = ẋ4 / ẏ1 = b(y1 , ẏ1 , ÿ1 , y2 , ẏ2 , ÿ2 ), x2 =
(2z + x3 )/y1 = c(y1 , ẏ1 , ÿ1 , y2 , ẏ2 , ÿ2 ), u 2 = ẋ2 = q(y1 , . . . , y1(3) , y2 , . . . , y2(3) ).
  (3)
Since x5 = x3 u 2 = b(y1 , . . . , ÿ2 )q(y1 , . . . , y2 ) one gets that the above system
of Eq. (2.196) is a Liouvillian one.

2.6 Elaborated Criteria for Checking Differential Flatness

2.6.1 Implicit Control Systems on Manifolds of Jets

Up to now the Lie-Backlünd approach to equivalence and flatness of nonlinear

dynamical systems has been used. The Lie-Backlünd approach states that a sys-
tem is differentially flat if and only if it is equivalent to (it can be transformed to)
the trivial system, that is to a system in the linear canonical Brunovsky form. There
88 2 Differential Flatness Theory and Flatness-Based Control

are elaborated criteria that enable to determine if a system is differential flat [286].
These state that flatness is equivalent to the property of strong closedness of the ideal
1-form representing the differentials of all possible trivializations.
For a nonlinear dynamical system ẋ = f (x, u) (that is a system in the explicit
form) which is described by a set of ordinary differential equations, its implicit
description is obtained after eliminating the control input from the state-space model.
This takes the form:
F(x, ẋ) = 0 (2.197)

Next, a manifold description is introduced. Considering the solutions of Eq. (2.197)

the infinite dimensional manifold is considered: X = X × R∞ n or equivalently X =

X × R × R · · · R . It is assumed that the infinite set of global coordinates X is

n n n

given by

X = (x1 , . . . , xn , ẋ1 , . . . , ẋn , ẍ1 , . . . , ẍn , . . . , x1(k) , . . . , xn(k) ) (2.198)

Moreover, the so-called Cartan vector field is defined

n  ( j+1) ∂
τX = x ( j)
(2.199)
i=1 j≥0 i ∂ xi

while also the Lie derivative of function φ ∈ C ∞ (X, R) along vector field τX is
introduced
dφ n  ( j+1) ∂φ
= L τX φ = x ( j)
(2.200)
dt i=1 j≥0 i ∂x i

( j) ( j) ( j) ( j+1)
Since it holds that dtd
xi = L τX xi = ẋi = xi , the Cartan vector field acts
on coordinates as a shift to the right. Variable (X, τX ), comprising the state vector
of the nonlinear system and its successive derivatives with respect to time, as well
as the Lie derivatives of the mapping transformation function φ along a vector field
defined by the state vector elements and their derivatives is called manifold of jets of
infinite order or diffiety.
Thus, in regular implicit control systems, one has no longer the description of the
system in the state-space form but uses a description based on the state vector and
its derivatives, as well as the derivatives of the mapping φ. Next, the definition of a
regular implicit control system is given:
Definition: A regular implicit control system is defined as a triple (X, τX , F) with
X = X × R∞ n , τ is the trivial Cartan field on X and F ∈ C ∞ (TX; R n−m ), satisfying
X
∂F
rank( ∂ x = n − m) in a suitable open dense subset of TX.
2.6 Elaborated Criteria for Checking Differential Flatness 89

2.6.2 The Lie-Backlünd Equivalence for Implicit Systems

The notion of the Lie-Backlünd equivalence for explicit control systems is now
extended to implicit control systems (systems described by manifolds of jets or
diffieties). Two regular implicit control systems are defined [286].
Definition 1: Let X, τX , F and Y, τY , G be two regular implicit systems. These are
Lie-Backlünd equivalent at (x̄0 , ȳ0 ) ∈ X0 × Y0 if and only if
(i) There exist neighborhoods X0 of x̄0 ∈ X0 and ȳ0 ∈ Y0 and a one-to-one map-
ping Φ = (φ0 , φ1 , . . .) ∈ C ∞ (Y0 ; X 0 ) with C ∞ (Y0 ; X 0 ) inverse Ψ , satisfying
Φ( ȳ0 ) = x̄0 and such that the restrictions of the trivial Cartan fields τY|y0 are
Φ-related, namely Φ ∗ τY|y0 = τX|x0
(ii) The C ∞ (Y0 |X0 ) inverse mapping Ψ = (ψ0 , ψ1 , . . .) is such that Ψ (x̄0 ) = ȳ0
and Ψ ∗ τX|x0 = τY|y0 .
The mappings Φ and Ψ are called mutually inverse Lie-Backlünd isomorphisms at
(x0 , y0 ).
It holds that two explicit control systems are Lie-Backlünd equivalent if and
only if the associated implicit control systems are Lie-Backlünd equivalent. This is
analytically stated as follows:
Two implicit systems (X, τX , F) and (Y, τY , G) are given and two vector fields f
and g compatible with (X, τX , F) and Y, τY , G, respectively are defined. The corre-
sponding explicit systems ẋ = f (x, u) and ẏ = g(y, v) are differentially equivalent
or Lie-Backlünd equivalent at the pair (x0 , u 0 , u̇ 0 ), . . . , and (y0 , v0 , v̇0 ), . . . (with u 0
such that ẋ0 = f (x0 , u 0 ) and v0 such that ẏ0 = g(x0 , v0 )), if and only if (X, τX , F)
and (Y, τY , G) are Lie-Backlünd equivalent in accordance to the previous Definition
1, at (x0 , ẋ0 , . . .) and (y0 , ẏ0 , . . .).
Next, it is pointed out once more that an explicit system is flat, if and only if, it
is equivalent to the trivial system (linear canonical Brunovsky form). This enables
to define also necessary and sufficient condition for implicit systems to be differen-
tially flat.
Definition 2: The implicit system (X, τX , F) is flat at x̄0 if and only if there exists
y0 ∈ R∞ m , such that (X, τ , F) is Lie-Backlünd equivalent at ( x̄ , ȳ ) ∈ X × R n
X 0 0 0 ∞
to the m-dimensional trivial implicit system (R∞m , τ , 0). In this case, the mutually
m
inverse Lie-Backlünd isomorphisms Φ and Ψ are called inverse trivializations.
Having denoted the implicit system as F(x, ẋ), its differential can be defined next

n  
∂ Fi ∂ Fi
dF = dx j + d ẋ j i = 1, . . . , n − m (2.201)
j=1 ∂x j ∂ ẋ j

Moreover, one defines the smooth mapping Φ of the 1-form (vector) ω to X as

follows:
  ∂φi
j
(l)
Φ ∗ ω( ȳ) = ωij (Φ(ψ̄)) (l) (ψ̄)dyk (2.202)
k,i i, j ∂ yk
90 2 Differential Flatness Theory and Flatness-Based Control

j ( j)
where it holds that φi (ψ̄) = xi . Then the following theorem holds about the
differential flatness of the implicit system
Theorem: The implicit system (X, τX , F) is flat at x̄0 , with x̄0 ∈ X0 if and only if
there exists ȳ0 ∈ R∞ m and a local Lie-Backlünd isomorphism Φ from R m to X
∞ 0
satisfying Φ( ȳ0 ) = x̄0 and such that Φ ∗ · dF = 0.

2.6.3 Conditions for Differential Flatness of Implicit Systems

To formulate necessary and sufficient conditions for the differential flatness of

implicit systems some preliminary definitions on polynomial matrices are used.
The following polynomial matrices are introduced where the indeterminate is the
d
differential operator dt

∂F ∂F d or d(φ 0 ) ∂φ 0 d j
P(F) = ∂x + ∂ ẋ dt P(φ 0 ) = j=0 ∂ y ( j) dt j
(2.203)

with P(F) (respectively P(φ 0 )) to be of size (n − m) × n (respectively n × m).

Next, the following theorem about the Smith decomposition (or diagonal reduction)
of matrices is given [286]:
Theorem: Given a matrix M ∈ M p,q [ dt d
] (the module of p × q matrices over [ dt
d
]),
there exist matrices V ∈ U p [ dt ], such that
d

 p , 0 p,q−p ) if p ≤ q
VMU = (Δ
Δq (2.204)
VMU = if p > q
0 p−q,q

where 0 p,q− p (respectively 0 p−q,q ) is the q × (q − p) (respectively ( p − q) × q)

matrix with entries that are all zeros, and with Δ p a p × p (respectively Δq a q × q)
diagonal matrix with elements δ1 , . . . , δσ , 0, . . . , 0 are such that δi is a nonzero dt
d

polynomial for i = 1, . . . , σ and is divisor of δ j for all for all σ ≥ j ≥ i. Moreover,

Δ p (respectively Δq ) is unique up to multiplication by a regular diagonal matrix in
 p× p (respectively q×q ).
Since matrix P(F) of the differential description P(F) ∈ Mn−m,n [ dt d
] of the
implicit system admits the Smith decomposition, it can be written as

VP(F)U = (Δ, 0n−m,m ) (2.205)

Moreover, the definition of hyper-regular matrices is given:

Definition: Given a matrix M ∈ M p,q [ dt
d
], it is said that M is hyper-regular if and
only if its Smith decomposition gives (Δ p , 0 p,q− p ) = (I p , 0 p,q− p ) if p ≤ q and
2.6 Elaborated Criteria for Checking Differential Flatness 91
   
Δq Iq
= if p ≥ q (2.206)
0 p−q,q 0 p−q,q

Other definitions associated with the Smith decomposition of the differential matrix
P(F) of the implicit system F(x, ẋ) are as follows:
It holds that VP(F)U = (Δ, 0n−m,m ) and matrix U is denoted as R-Smith(P(F)).
Next one denotes  
0
Û = U n−m,m (2.207)
Im

Moreover, there exists matrix Q 0 defined as

 
0m,n−m Im
Q0 = U −1 ∈ L − Smith(Û ) (2.208)
In−m 0n−m,m

and the associated matrices which satisfy

 
Q̃ 0 = Im 0m,n−m (2.209)

Q̂ 0 = (0n−m,m , In−m )Q 0 (2.210)

Next, it is considered that U is the right matrix in the Smith decomposition of the
implicit system’s differential matrix P(F), that is U ∈ R − Smith(P(F)). Then the
following m-dimensional vector 1-form ω is defined
⎛ ⎞ ⎛  ⎞
n j α
ω1 (x̄) j=1 α≥0 Q 1,a ( x̄)d x j |X0
⎜ ⎟
ω(x̄) = ⎝ · · · ⎠ = Q̃ 0 d x|X0 = ⎝ ··· ⎠ (2.211)
n 
ωm (x̄) j=1 α≥0 Q
j
m,a ( x̄)d x α|
X0 j

Using the above one arrives at a criterion for defining flat outputs.
Definition: Variational flat outputs. The vector 1-form ω = Q̃ 0 d x|X0 , defined in
Eq. (2.211), using matrices Q 0 and Q̃ 0 defined in Eqs. (2.229) and (2.210), respec-
tively, is a flat output of the variational system.
The following relations are defined next:
Definition: It is said that the [ dt
d
]-ideal T generated by τ1 , . . . , τΓ is strongly closed,
if there exists a matrix M ∈ Um [ dτ d
] such that d(Mτ ) = 0.
Theorem: A necessary and sufficient condition for systems X, τX , F to be flat at
(x̄0 , ȳ0 ) (over the class of meromorphic functions) is that there exist U ∈ R −
Smith(P(F)) and Q ∈ L = Smith(Û ), with Û to be given by
 
0n−m.m
Û = U (2.212)
Im
92 2 Differential Flatness Theory and Flatness-Based Control

such that the [ dt

d
]-ideal Ω which is generated by the 1-forms ω1 , . . . , ωm defined
in Eq. (2.211) to be strongly closed in X0 .
The following operator is defined

D(H )κ = d(H κ) − H dκ (2.213)

for all m-dimensional vector p-form κ in Λ p (( X ))m , all p ≥ 1 and all H ∈ Um [ dt

d
].
Then, the following theorem holds [286]:
Theorem: The [ dt d
]-ideal Ω generated by the components of the vector 1-form ω
defined in Eq. (2.211) is strongly closed in X0 , or equivalently the system (X, τX , F)
is flat, if and only if, there exists μ ∈ L 1 (Λ(( X ))m ) and a matrix M ∈ Um [ dt
d
] such
that
dω = μω D(μ) = μ2 D(M) = −Mμ (2.214)

The condition given in Eq. (2.214) may be seen as a generalization in the framework
of manifolds of jets of infinite order of the well-known moving frame structure
equations. Furthermore, the general matrix solution μ = (μik )i,k=1,...,m of dω = μω,
with ω defined in Eq. (2.211) is given by

m or d(μ) j,k j,k (α) dβ

μik = (Γi,α,β + vi,α,β )ω j ∧ (2.215)
j=1 α,β=0 dτ β
with
 j,k k, j
vi,α,β = vi,α,β ∀i, j, k = 1, . . . , m ∀α, β = 0, . . . , or d(μ), α = β or j = k
k,k
vi,α,α arbitrary ∀i, k = 1, . . . , m ∀α = 0 · · · or d(μ)
(2.216)

the integer ord(μ) being arbitrary but otherwise finite and satisfying or d(μ) ≥
j,k
or d(Γ ), the Γi,α,β being given by the following relation

or d(Γ ) m j,k (α) (β)

dωi = Γ ω ∧ ωk (2.217)
α,β=0 j,k=1 i,α,β j

j,k
and where vi,α,β ’s are meromorphic functions depending on successive derivatives
of x.
Moreover, the following hold:
The [ dt
d
]-ideal Ω generated by the components of the vector 1-form ω defined
in Eq. (2.211) is strongly closed in X0 , or equivalently the system (X, τX , F) is flat,
if and only if, there exists μ ∈ L 1 (Λ(X)m ), and two matrices M ∈ Mm,m [ dt d
] and
N ∈ Mm,m [ dt ], such that
d

dω = μω, D(μ) = μ2 , D(M) = −Mμ, D(N ) = μN , MN = NM = I (2.218)

2.6 Elaborated Criteria for Checking Differential Flatness 93

The procedure for testing if a system is differentially flat and for computing the
associated flat outputs is:
1. Write the system in the implicit form F(x, ẋ) by eliminating the control input
from its state-space equations.
2. Compute the differential matrix of the system P(F).
3. Perform Smith decomposition of P(F) and retain the R-Smith matrix U .
4. Compute matrix Û from Eq. (2.228).
5. Compute matrices Q 0 and Q̃ 0 from Eqs. (2.210) and (2.231), respectively.
6. Compute the vector 1-form that is described in Eq. (2.211). The elements of the
ω vector are the flat outputs of the system.
If for some or d(μ), the algorithm produces an invertible M, a flat output is obtained
by integration of dy = Mω, which is possible since d(Mω) = 0. In the opposite
case, the system is not differentially flat.

2.6.4 Example of Elaborated Differential Flatness Criteria

to Nonlinear Systems

An example is given next, about the application of elaborated differential flatness

criteria for implicit systems.
The unicycle robot: The kinematic model of the unicycle robot is considered [286]:

ẋ = vcos(θ )
ẏ = vsin(θ ) (2.219)
θ̇ = vl tan(φ)

The system is written in implicit form F(x, ẋ) = 0. It has n = 3 states and m = 2
inputs, with n − m = 1 and thus it is equivalent to a single implicit equation which
is obtained
F(x, y, θ, ẋ, ẏ, θ̇ ) = ẋsin(θ ) − ẏcos(θ ) = 0 (2.220)

Next, one computes

 
∂F ∂F d ∂F ∂F d ∂F ∂F d
P(F) = ∂x + ∂ ẋ dt ∂ y + ∂ ẏ dt ∂θ + ∂ θ̇ dt
⇒
(2.221)
 
P(F) = sin(θ ) dt
d
, −cos(θ ) dt
d
, ẋcos(θ ) + ψ̇sin(θ )

Setting E = ẋcos(θ ) + ẏsin(θ ) one applied the Smith decomposition algorithm to

vector P. The Smith decomposition (or diagonal reduction) of a polynomial matrix
is as follows:
94 2 Differential Flatness Theory and Flatness-Based Control

Given a μ× polynomial matrix A over the noncumulative ring [ dt d

], there
exist matrices V ∈ Uμ [ dt ] and U ∈ U [ dt ], such that V AU = (Δ, 0) if μ ≥
d d

where Δ is a μ × μ (respectively × ) diagonal matrix, with diagonal elements

δ1 , . . . , δn ,0, . . . , 0 such that δi is a nonzero dt
d
-polynomial for i = 1, . . . , σ and is
divisor of δ j for all σ ≥ j ≥ 1.
 
δ
V AU = (2.222)
0

if μ ≥ , where Δ is a μ × μ (respectively × ) diagonal matrix, whose diagonal

elements δ1 , . . . , δn ,0, . . . , 0 are such that δi is a nonzero dt
d
polynomial
Moving the last column (of degree zero) to the first place by permutation with the
two others one gets
 
P(F)U0 = E, −cos(θ ) dt
d
, sin(θ ) dt
d
(2.223)

with ⎛ ⎞
001
U0 = ⎝0 1 0⎠ (2.224)
100

and next by right-multiplying P(F)U0 with U1 where

⎛1 cos(θ) d
⎞
E E dt − sin(θ)
E dt
d

U1 = ⎝ 0 1 0 ⎠ (2.225)
0 0 1

finally gives  
P(F)U = 1 0 0 (2.226)

with matrix U to be computed by

⎛ ⎞
0 0 1
U = U0 U1 = ⎝ 0 1 0 ⎠ (2.227)
1 cos(θ) d
E E dt − sin(θ)
E dt
d

It also holds that vector P(F) is hyper-regular and that

⎛ ⎞
  0 1
01,2
Û = U =⎝ 1 0 ⎠ (2.228)
I2 cos(θ)
E
d
dt − sin(θ)
E dt
d
2.6 Elaborated Criteria for Checking Differential Flatness 95

with I2 being the identity matrix in R 2 . Next by defining matrix Q 0 as

 
0m,n−m Im
Q0 = U −1 ∈ L − Smith(Û ) (2.229)
In=m 0n−m,m

and matrices Q̃ 0 , Q̂ 0 as  
Q̃ 0 = Im 0m,n−m Q 0 (2.230)
 
Q̂ 0 = 0n−m In−m Q 0 (2.231)

one has ⎛ ⎞
0 1 0
Q0 = ⎝ 1 0 0⎠ (2.232)
sin(θ) d
L dt − cos(θ)
L dt
d
1
 
010
Q̃ 0 = (2.233)
100

Multiplying Q 0 by the vector d X = [d x, dy, dθ ]T the last line of this product

becomes

Q0d X = E (sin(θ )d ẋ − cos(θ )d ψ̇ + ( ẋcos(θ ) + ψ̇sin(θ )dθ ))

1
⇒
Q 0 d X = E1 d(ẋsin(θ ) − ẏcos(θ )) ⇒ (2.234)
Q0d X = 0

which comes from using Eq. (2.220) about the implicit system dynamics. Next matrix
Q̃ 0 is defined as
Q̃ 0 = (Im , 0m,n−m )Q 0 (2.235)

and by multiplying Q̃ 0 with vector d X = [d x, dy, dθ ]T one gets

⎛ ⎞
dx  
ω1
Q̃ 0 ⎝dy ⎠ = (2.236)
ω2
dθ

which finally gives ω1 = dy and ω2 = d x. Thus, the ideal Ω that is generated by

Ω1 , Ω2 is trivially strongly closed with M = I2 . Consequently, one finally concludes
that the flat output of the system is y1 = y and y2 = x.
96 2 Differential Flatness Theory and Flatness-Based Control

2.7 Distributed Parameter Systems and Their

Transformation into the Canonical Form

Up to now the analysis on differential flatness properties was focused on lumped

parameter systems, that is systems described by ordinary differential equations. By
applying differential flatness theory it is also possible to transform into the canonical
form distributed parameter systems which are described by nonlinear partial differ-
ential equations of the parabolic, elliptic and hyperbolic type. This is particularly
significant for developing efficient nonlinear control and filtering methods for such
a type of systems. To demonstrate this, the model of nonlinear heat diffusion PDE
will be used as an example.

2.7.1 State-Space Description of a Heat Diffusion Dynamics

The following nonlinear heat diffusion equation is considered, describing the spa-
tiotemporal variations of temperature in a surface

∂φ ∂ 2φ
= K 2 + f (φ) + u(x, t) (2.237)
∂t ∂x
Using the approximation for the partial derivative in the partial differential equation
given in Eq. (12.38) one has

∂ 2φ φi+1 − 2φi + φi−1

= (2.238)
∂x2 Δx 2
and considering spatial measurements of variable φ along axis x at points x0 +
iΔx, i = 1, 2, . . . , N one has

∂φi K 2K K
= φi+1 − φi + φi−1 + f (φi ) + u(xi , t) (2.239)
∂t Δx 2 Δx 2 Δx 2
By considering the associated samples of φ given by φ0 , φ1 , . . . , φ N , φ N +1 one has

∂φ1
∂t = K
φ
Δx 2 2
− 2K
φ
+ Δx
Δx 2 1
K
2 φ0 + f (φ1 ) + u(x 1 , t)
∂φ2
∂t = K
φ
Δx 2 3
− 2K
φ
+ Δx 2 φ1 + f (φ2 ) + u(x2 , t)
Δx 2 2
K
∂φ3
∂t = K
φ
Δx 2 4
− 2K
φ
+ Δx
Δx 2 3
K
2 φ2 + f (φ3 ) + u(x 3 , t)
(2.240)
···
∂φ N −1
∂t = Δx
K
2 φ N − Δx 2 φ N −1 + Δx 2 φ N −2 + f (φ N −1 ) + u(x N −1 , t)
2K K
∂φ N
∂t = Δx 2 φ N +1 − Δx 2 φ N + Δx 2 φ N −1 + f (φ N ) + u(x N , t)
K 2K K
2.7 Distributed Parameter Systems and Their Transformation … 97

By defining the following state vector

 
x T = φ1 , φ2 , . . . , φ N (2.241)

one obtains the state-space description

ẋ1 = K
x
Δx 2 2
− 2K
x
Δx 2 1
+ ΔxK
2 φ0 + f (x 1 ) + u(x 1 )
ẋ2 = K
x
Δx 2 3
− 2K
x
Δx 2 2
+ Δx 2 x1 + f (x2 ) + u(x2 )
K

ẋ3 = K
x
Δx 2 4
− 2K
x
Δx 2 3
+ ΔxK
2 x 2 + f (x 3 ) + u(x 3 )
(2.242)
···
ẋ N −1 = Δx K
2 x N − Δx 2 x N −1 + Δx 2 x N −2 + f (x N −1 ) + u(x N −1 )
2K K

ẋ N = Δx 2 φ N +1 − Δx 2 x N + Δx
K 2K K
2 x N −1 + f (x N ) + u(x N )

Next, the following state variables are defined

y1,i = xi
(2.243)
y2,i = ẋi

and the state-space description of the system becomes as follows:

ẏ1,1 = Δx K
2 y1,2 − Δx 2 y1,1 + Δx 2 φ0 + f (y1,1 ) + u(y1,1 )
2K K

ẏ1,2 = Δx K
2 y1,3 − Δx 2 y1,2 + Δx 2 y1,1 + f (y1,2 ) + u(y1,2 )
2K K

ẏ1,3 = Δx 2 y1,4 − Δx 2 y1,3 + Δx 2 y1,2 + f (y1,3 ) + u(y1,3 )

K 2K K

···
(2.244)
···
ẏ1,N −1 = Δx 2 y1,N − Δx 2 y1,N −1 + Δx
K 2K K
2 y1,N −2 + f (y1,N −1 )
+u(y1,N −1 )
ẏ1,N = Δx K
2 φ N +1 − Δx 2 y1,N + Δx y1,N −1 + f (y1,N ) + u(y1,N )
2K K

The dynamical system described in Eq. (13.52) is a differentially flat one with flat
output defined as the vector ỹ = [y1,1 , y1,2 , . . . , y1,N ]. Indeed all state variables can
be written as functions of the flat output and its derivatives. Moreover, by defining
the new control inputs

v1 = Δx K
2 φ0 + f (y1,1 ) + u(y1,1 )
v2 = f (y1,2 ) + u(y1,2 )
v3 = f (y1,3 ) + u(y1,3 )
(2.245)
···
v N −1 = f (y1,N −1 ) + u(y1,N −1 )
v N = Δx K
2 φ N +1 + f (y1,N ) + u(y1,N )
98 2 Differential Flatness Theory and Flatness-Based Control

the following state-space description is obtained

⎛ 2K ⎞
⎛ ⎞ − Δx 2 ΔxK
2 0 0 ··· 0 0 0 0 ⎛ ⎞
ẏ1,1 ⎜ K − 2K K
0 ··· 0 0 0 0 ⎟ y1,1
⎜ ẏ1,2 ⎟ ⎜ Δx 2 Δx 2 Δx 2 ⎟⎜
⎟ ⎜ y1,2 ⎟
⎜ ⎟ ⎜ K
− Δx 2 Δx 2 · · ·
2K K ⎟
⎜ ··· ⎟ = ⎜ 0 0 0 0 0 ⎟⎜
⎟⎜ ··· ⎟ ⎟+
Δx 2
⎜ ⎟ ⎜
⎜ · · · · · · ··· ··· ··· ··· ··· ··· ··· ⎟⎝
⎝ ẏ1,N −1 ⎠ ⎜ ⎟ y1,N −1 ⎠
⎝ 0 0 0 0 ··· 0 K
− Δx
2K K ⎠
ẏ1,N Δx 2 2 Δx 2 y1,N
0 0 0 0 ··· 0 0 K
Δx 2
− Δx
2K
2

⎛ ⎞⎛ ⎞
1 0 0 ··· 0 0 v1
⎜ 0 1 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟⎜ ⎟
⎜ 0 0 1 · · · 0 0 ⎟ ⎜ v3 ⎟
⎜
+⎜ ⎟ ⎜ ⎟
⎟⎜ ⎟
⎜· · · · · · · · · · · · ⎟ ⎜ · · · ⎟
⎝ 0 0 0 · · · 1 0 ⎠ ⎝v N −1 ⎠
0 0 0 ··· 0 1 vN
(2.246)

Assuming that all measurements from the set of points x j j ∈ [1, 2, . . . , m] are
available, the associated observation (measurement) equation becomes
⎛ ⎞
⎛ ⎞ y1,1
⎛ ⎞ 1 0 0 ··· 0 0 ⎜
z1 ⎟
⎜ ⎟ ⎜ y1,2 ⎟
⎜ z 2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟ ⎜ y1,3 ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ ⎟ (2.247)
⎝ · · ·⎠ ⎜ ⎟⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎝ y1,N −1 ⎠
0 0 0 ··· 0 1
y1,N

Thus, in matrix form one has the following state-space description of the system

ỹ˙ = A ỹ + Bv
(2.248)
z̃ = C ỹ

Moreover, denoting a = Dx K
2 and b = − Dx 2 , the initial description of the system
2K

given in Eq. (13.54) is rewritten as follows

2.7 Distributed Parameter Systems and Their Transformation … 99
⎛ ⎞
⎛ ⎞ b a 0 0 0 ··· 0 0 0 0 0 0 ⎛ ⎞
ẏ1,1 ⎜ a b a 0 0 ··· 0 0 0 0 0 0 ⎟ y1,1
⎜ ẏ1,2 ⎟ ⎜ ⎟⎜ ⎟
⎜ ⎟ ⎜ 0 0 a b a · · · 0 0 0 0 0 0 ⎟ ⎜ y1,2 ⎟
⎜ ··· ⎟=⎜ ⎟⎜ ··· ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟
⎝ ẏ1,N −1 ⎠ ⎜· · · · · · · · · · · · · · · · · · · · ·⎟ ⎝ y1,N −1 ⎠
⎝ 0 0 0 0 0 0 0 ··· a b a 0 ⎠
ẏ1,N y1,N
0 0 0 0 0 0 0 ··· 0 0 a b
⎛ ⎞⎛ ⎞
1 0 0 ··· 0 0 v1
⎜ 0 1 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟⎜ ⎟
⎜ 0 0 1 · · · 0 0 ⎟ ⎜ v3 ⎟
+⎜ ⎟⎜ ⎟
⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟ (2.249)
⎜ ⎟⎜ ⎟
⎝ 0 0 0 · · · 1 0 ⎠ ⎝v N −1 ⎠
0 0 0 ··· 0 1 vN

2.7.2 Differential Flatness of the Nonlinear

Heat Diffusion PDE

The previously defined state vector of the PDE model is considered without the effect
of the external control input u(x, t).

∂φ ∂ 2φ
= K 2 + f (φ) (2.250)
∂t ∂x

The state vector is again Ỹ = [y1,1 , y1,2 , . . . , y1, i, . . . , y1,N −1 , y1,N ], where y1,1 =
V1 , y1,2 = V2 , . . . , y1,i = Vi , . . . , y1,N −1 = VN −1 and y1,N = VN . It will be shown
in a different manner that the state-space description of the nonlinear PDE dynamics,
which has as control input only the boundary condition φ0 is a differentially flat one.
One has
2K K K
ẏ1,1 = − 2 y1,1 + φ0 + f (y1,1 ) + y1,2 (2.251)
Δx Δx 2 Δx 2
2K K K
ẏ1,2 = − y1,2 + y1,1 + f (y1,2 ) + y1,3 (2.252)
Δx 2 Δx 2 Δx 2
2K K K
ẏ1,3 = − y1,3 + y1,2 + f (y1,3 ) + y1,4 (2.253)
Δx 2 Δx 2 Δx 2
··· ···
2K K K
ẏ1,i = − y1,i + y1,i−1 + f (y1,i ) + y1,i+1 (2.254)
Δx 2 Δx 2 Δx 2
··· ···
2K K K
ẏ1,N −1 = − y1,N −1 + y1,N −2 + f (y1,N −1 ) + y1,N (2.255)
Δx 2 Δx 2 Δx 2
100 2 Differential Flatness Theory and Flatness-Based Control

2K K K
ẏ1,N = − y1,N + y1,N −1 + f (y1,N ) + φ N +1 (2.256)
Δx 2 Δx 2 Δx 2
The flat output is considered to be the state variable y1,N , which is denoted as
y = y1,N . Next, it is shown that all state variables which stand also for virtual
control inputs of the system αi = y1,N −i , can be written as functions of the flat
output y = y1,N
From Eq. (13.62) and considering that φ N +1 is constant one obtains

y1,N −1 = α1 = 1
[ ẏ
K /Δx 2 1,N
+ Δx
2K
2 y1,N − 2Δx 2 φ N +1 − f (y1,N )]
K
(2.257)
⇒y1,N −1 = h 1 (y, ẏ, · · · )

and following a similar procedure, from Eq. (13.61) one gets

y1,N −2 = α2 = 1
[ ẏ
K /Δx 2 1,N −1
+ 2K
y
Δx 2 1,N −1
− 2Δx K
2 y1,N − f (y1,N −1 )]
(2.258)
⇒y1,N −2 = h 2 (y, ẏ, · · · )

Continuing in a similar manner, from Eq. (13.60) one obtains

y1,i−1 = α N −i+1 = K /Δx

1
2 [ ẏ1,i + Δx 2 y1,i − 2Δx 2 y1,i+1 − f (y1,i )]
2K K
(2.259)
⇒y1,i−1 = h N −i+1 (y, ẏ, · · · )

From Eq. (13.59) one obtains

y1,2 = α N −2 = 1
[ ẏ
K /Δx 2 1,3
+ Δx2K
2 y1,3 − Δx 2 y1,4 − f (y1,3 )]
K
(2.260)
⇒y1,2 = h N −2 (y, ẏ, · · · )

From Eq. (13.58) one obtains

y1,1 = α N −1 = 1
[ ẏ
K /Δx 2 1,2
+ Δx2K
2 y1,2 − Δx 2 y1,3 − f (y1,2 )]
K
(2.261)
⇒y1,1 = h N −1 (y, ẏ, · · · )

Finally, From Eq. (13.57) one obtains

φ0 = α N = 1
[ ẏ
+ Δx
K /Δx 2 1,1
2K
2 y1,1 − Δx 2 y2 − f (y1,1 )]
K
(2.262)
⇒φ0 = h N (y, ẏ, · · · )

The above procedure confirms that all state variables of the model

y1,1 y1,2 y1,3 · · ·

(2.263)
y1,i · · · y1,N −1 y1,N
2.7 Distributed Parameter Systems and Their Transformation … 101

and the control input which is the boundary condition φ0 can be written as functions of
the flat output y = y1,N and of the flat output’s derivatives. Consequently, differential
flatness for the heat PDE is shown.
Additionally, one can consider decomposition of the PDE state-space equation
into submodels, where at each submodel the virtual control input is αi = y1,N −i

ẏ1,i = − Δx
2K
2 y1,i +
K
y
Δx 2 1,i−1
+ f (y1,i ) + K
y
Δx 2 1,i+1 (2.264)
··· ···

and with local flat output y1,i one can confirm that all subsystems are differentially
flat.
Chapter 3
Nonlinear Adaptive Control Based
on Differential Flatness Theory

3.1 Introduction

As explained in Chap. 2, the assumption about complete knowledge of the system’s

dynamic model does not always hold. Therefore, there is need for flatness-based
control schemes, capable of functioning efficiently under parametric model uncer-
tainty and even without any prior knowledge about the system’s dynamics. To this
end, flatness-based adaptive fuzzy control methods are developed.
First the chapter analyzes flatness-based adaptive neuro-fuzzy control for single-
input nonlinear dynamical systems. Such systems can be written in the Brunovsky
form via a transformation of their state variables and control input. The resulting
control signal is shown to contain nonlinear elements, which in case of unknown
system parameters can be approximated using neuro-fuzzy networks. Using Lya-
punov stability analysis it is shown that one can compute an adaptation law for the
neurofuzzy approximators which assures stability of the single-input control loop.
Next the chapter proposes flatness-based adaptive neuro-fuzzy control for uncer-
tain multi-input multi-output (MIMO) nonlinear dynamical systems. The considered
control scheme, based on differential flatness theory, extends the class of MIMO
systems in which indirect adaptive fuzzy control can be applied. MIMO nonlin-
ear systems satisfying the differential flatness property can be written in the linear
canonical (Brunovsky) form via a transformation of their state variables and control
inputs. The resulting control signal is shown to contain again nonlinear elements,
which in case of unknown system parameters can be calculated using neuro-fuzzy
approximators. By applying again Lyapunov stability analysis it is shown that one
can find an adaptation law for the neuro-fuzzy approximators which assures stability
of the MIMO control loop.

© Springer International Publishing Switzerland 2015 103

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_3
104 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

3.2 Flatness-Based Adaptive Neuro-Fuzzy Control for SISO

Systems

3.2.1 Overview

As analyzed in Chap. 2, flatness-based control can be designed so as to cope effi-

ciently with model uncertainties and parametric changes in nonlinear dynamical
systems [427, 465, 495]. This chapter proposes flatness-based adaptive fuzzy con-
trol for nonlinear dynamical systems with unknown parameters. Actually, the pro-
posed flatness-based control method can work effectively even in the case of no-prior
knowledge about the system’s dynamics.
As explained in Chap. 2, all single-input systems are differentially flat and can be
transformed to the linear canonical (Brunovsky) form. After such a transformation,
the resulting control input is shown to contain nonlinear elements which depend
on the system’s parameters. If the parameters of the system are unknown, then the
nonlinear terms which appear in the control signal can be approximated with the
use of neuro-fuzzy networks. In the chapter it is shown that a suitable learning law
can be defined for the aforementioned neuro-fuzzy approximators so as to preserve
the closed-loop system stability. Lyapunov stability analysis proves also that the
proposed flatness-based adaptive fuzzy control scheme results in H∞ tracking per-
formance, in accordance with the results of [409, 410, 413].
Adaptive fuzzy control is suitable for compensation of model uncertainties and
external perturbations in nonlinear dynamical systems [7, 87, 96, 104, 168, 210,
302, 394, 407, 413, 427, 443, 504, 593]. The adaptive fuzzy control system based
on differential flatness theory extends the class of systems to which indirect adaptive
fuzzy control can be applied. Unlike other adaptive control schemes which are based
on several assumptions about the structure of the nonlinear system as well as about the
uncertainty characterizing the system’s model, the proposed adaptive fuzzy control
scheme based on differential flatness theory offers an exact solution to the design
of fuzzy controllers for unknown dynamical systems. The only requirement for the
design of the controller and for succeeding H∞ tracking performance for the control
loop is that there exists a solution for a Riccati equation associated with the linearized
error dynamics of the differentially flat model. This condition is shown to be true
for several nonlinear systems, thus providing a systematic approach to the design of
reliable controllers for such systems [426, 433].
3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 105

3.3 Flatness-Based Adaptive Fuzzy Control for SISO

Dynamical Systems

3.3.1 Transformation of SISO Nonlinear Systems

into a Canonical Form

A single-input differentially flat dynamical system is considered next:

ẋ = f s (x, t) + gs (x, t)(u + d̃), x ∈ R n , u ∈ R, d̃ ∈ R (3.1)

where f s (x, t), gs (x, t) are nonlinear vector fields defining the system’s dynamics,
u denotes the control input and d̃ denotes additive input disturbances. Knowing
that the system of Eq. (3.1) is differentially flat, the next step is to try to write it
into a Brunovsky form. It has been shown that, in general, transformation into the
Brunovsky (canonical) form can be succeeded for systems that admit static feedback
linearization [340]. Single input differentially flat systems, admit static feedback
linearization, and can be transformed into the Brunovksy form.
The selected flat output is denoted by y. Then, as analyzed in Sect. 3.4.2, for the
state variables xi of the system of Eq. (3.1) it holds that

xi = φi (y, ẏ, . . . , y (r −1) ), i = 1, . . . , n (3.2)

while for the control input it holds that

u = ψ(y, ẏ, . . . , y (r −1) , y (r ) ) (3.3)

Introducing the new state variables y1 = y and yi = y (i−1) , i = 2, . . . , n, the initial

system of Eq. (3.1) can be written in the Brunovsky form:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 0 1 0 ··· 0 y1 0
⎜ ẏ2 ⎟ ⎜ 0 0 1 · · · 0 ⎟ ⎜ y2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ · · · ⎟ ⎜· · · · · · · · · · · · · · ·⎟ ⎜ · · · ⎟ ⎜· · ·⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ · · · ⎟ ⎜· · · · · · · · · · · · · · ·⎟ ⎜ · · · ⎟ + ⎜· · ·⎟ v (3.4)
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ẏn−1 ⎠ ⎝ 0 0 0 · · · 1 ⎠ ⎝ yn−1 ⎠ ⎝ 0 ⎠
ẏn 0 0 0 ··· 0 yn 1

where v = f (x, t) + g(x, t)(u + d̃) is the control input for the linearized model, and
d̃ denotes additive input disturbances. Thus one can use

y (n) = f (x, t) + g(x, t)(u + d̃) (3.5)

where f (x, t), g(x, t) are unknown nonlinear functions, while as mentioned above,
d̃ is an unknown additive disturbance. It is possible to make the system’s state vector
106 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

x follow a given bounded reference trajectory xd . In the presence of model uncertain-

ties and external disturbances, denoted by wd , successful tracking of the reference
trajectory is provided by the H∞ criterion [308, 413, 524]:
T T
0 e T Qedt ≤ ρ 2 0 wd T wd dt (3.6)

where ρ is the attenuation level and corresponds to the maximum singular value of
the transfer function G(s) of the linearized model associated to Eqs. (3.4) and (3.5).
The concept of H∞ control comes from the need to compensate for the worst
type of disturbances that may affect a nonlinear dynamical system. This was initially
formulated for linear dynamics described with the use of transfer functions. The
H∞ norm of a linear system with transfer function G(s), is denoted by ||G||∞ and
is defined as ||G||∞ = supω σmax [G( jω)] where σmax is the system’s maximum
singular value [407, 410, 413]. The term sup denotes the supremum or least upper
bound of the function σmax [G( j (ω)], and thus the H∞ norm of G(s) is the maximum
value of σmax [G( j (ω)] over all frequencies ω. H∞ norm has a physically meaningful
interpretation when considering the system y(s) = G(s)u(s). When this system is
driven with a unit sinusoidal input at a specific frequency, σmax |G( jω)| is the largest
possible output for the corresponding sinusoidal input. Thus, the H∞ norm is the
largest possible amplification over all frequencies of a sinusoidal input.

3.3.2 Adaptive Control Law for SISO Nonlinear Systems

For the measurable state vector x of the system of Eqs. (3.4) and (3.5), and for
uncertain functions f (x, t) and g(x, t) an appropriate control law is
1 (n)
u= [y − fˆ(x, t) − K T e + u c ] (3.7)
ĝ(x, t) d

with e = [e, ė, ë, . . . , e(n−1) ]T and e = y − yd , K T = [kn , kn−1 , . . . , k1 ], such

that the polynomial e(n) + k1 e(n−1) + k2 e(n−2) + · · · + kn e is Hurwitz. The term u c
denotes the supervisory (supplementary) control input that is used for unmodeled
dynamics and external perturbations. The control law of Eq. (3.7) results in

e(n) = −K T e + u c + [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + g(x, t)d̃

(3.8)

where the supervisory control term u c aims at the compensation of the approximation
error

w = [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u (3.9)

3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 107

as well as of the additive disturbance term d1 = g(x, t)d̃. The above relation
can be written in a state-equation form. The state vector is taken to be e T =
[e, ė, . . . , e(n−1) ], which after some operations yields

ė = (A − B K T )e + Bu c + B{[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d1 }

(3.10)
where
⎛ ⎞ ⎛ ⎞
0 1 0 ··· ··· 0 0
⎜ 0 0 1 ··· ··· 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟
⎜· · · · · · · · · · · · · · · · · ·⎟ ⎜· · ·⎟
A=⎜ ⎜ ⎟ ⎜ ⎟
· · · · · · · · · · · · · · · · · · ⎟ , B = ⎜· · ·⎟ (3.11)
⎜ ⎟ ⎜ ⎟
⎝ 0 0 0 ··· ··· 1 ⎠ ⎝0⎠
0 0 0 ··· ··· 0 1

and K = [kn , kn−1 , . . . , k2 , k1 ]T . This is again a canonical form for the dynamics of
the tracking error. As explained above, the control signal u c is an auxiliary control
term, used for the compensation of d̃ and w, which can be selected according to H∞
control theory:

u c = − r1 B T Pe (3.12)

3.3.3 Approximators of SISO System Unknown Dynamics

The approximation of functions f (x, t) and g(x, t) of Eq. (3.5) can be carried out
with neuro-fuzzy networks (Fig. 3.1). The estimation of f (x, t) and g(x, t) can be
written as [543, 544]:

fˆ(x|θ f ) = θ Tf φ(x), ĝ(x|θg ) = θgT φ(x), (3.13)

where φ(x) are kernel functions with elements

n
i=1 μ A (xi )
l
φ l (x) =
N n i l l = 1, 2, . . . , N (3.14)
l=1 i=1 μ A (xi )
i

It is assumed that the weights θ f and θg vary in the bounded areas Mθ f and Mθg
which are defined as

Mθ f = {θ f ∈ R h : ||θ f || ≤ m θ f },
(3.15)
Mθg = {θg ∈ R h : ||θg || ≤ m θg }
108 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

Fig. 3.1 Neuro-fuzzy approximator: G i Gaussian basis function, Ni normalization unit

with m θ f and m θg positive constants. The values of θ f and θg that give optimal
approximation are:

θ ∗f = arg minθ f ∈Mθ [supx∈Ux | f (x) − fˆ(x|θ f )|]

f (3.16)
θg∗ = arg minθg ∈Mθg [supx∈Ux |g(x) − ĝ(x|θg )|]

The approximation error of f (x, t) and g(x, t) is given by

w = [ f (x, t) − fˆ(x, |θ ∗f )] + [g(x, t) − ĝ(x|θ ∗f )]u ⇒

(3.17)
w = {[ fˆ(x|θ f ) − fˆ(x|θ ∗f )] + [ f (x, t) − fˆ(x|θ f )]}+
+{[ĝ(x|θg ) − ĝ(x|θg∗ )] + [g(x, t) − ĝ(x|θg )]}u

where: (i) fˆ(x|θ ∗f ) is the approximation of f for the best estimation θ ∗f of the weights’
vector θ f , (ii) ĝ(x|θg∗ ) is the approximation of g for the best estimation θg∗ of the
weights’ vector θg .
The approximation error w can be decomposed into wa and wb , where

wa = [ fˆ(x|θ f ) − fˆ(x|θ ∗f )] + [ĝ(x|θg ) − ĝ(x|θg∗ )]u

(3.18)
wb = [ fˆ(x|θ ∗f ) − f (x, t)] + [ĝ(x|θg∗ ) − g(x, t)]u

Finally, the following two parameters are defined as

θ̃ f = θ f − θ ∗f
(3.19)
θ̃g = θg − θg∗ .

A difference between the neuro-fuzzy approximator given in Fig. 3.1 and a radial
basis functions (RBF) neural network is the normalization layer that appears between
the Gaussian basis functions layer and the weights output layer. After normalization,
3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 109

the sum (over the complete set of rules) of the fuzzy membership values of each
input pattern becomes equal to 1.

3.3.4 Lyapunov Stability Analysis for SISO Dynamical Systems

The adaptation law of the weights θ f and θg as well as of the supervisory control term
u c is derived by the requirement for negative derivative of the quadratic Lyapunov
function
1 T 1 T 1 T
V = e Pe + θ̃ θ̃ f + θ̃ θ̃g (3.20)
2 2γ1 f 2γ2 g

Differentiating Eq. (3.20) gives

1 T ˙ 1 T ˙
V̇ = 21 ė T Pe + 21 e T P ė + γ1 θ̃ f θ̃ f + γ2 θ̃g θ̃g ⇒ (3.21)

Next, substituting Eq. (3.17) into Eq. (3.21) gives

V̇ = 21 e T {(A − B K T )T P + P(A − B K T )}e+

+B T Pe(u c + w + d1 ) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙ g
(3.22)

Assumption 1: For given positive definite matrix Q and coefficients r and ρ there
exists a positive definite matrix P, which is the solution of the following matrix
equation:

T 2 1
(A − B K T ) P + P(A − B K T ) − PB( − 2 )B T P + Q = 0 (3.23)
r ρ

Substituting Eq. (3.23) into V̇ yields after some operations

V̇ = − 21 e T Qe + 21 e T PB( r2 − 1
ρ2
)B T Pe + B T Pe(− r1 e T PB)+
(3.24)
+B T Pe(w + d1 ) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g

It holds that

θ̃˙ f = θ̇ f − θ˙∗f = θ̇ f
(3.25)
θ̃˙g = θ̇g − θ˙g∗ = θ̇g

The following weight adaptation laws are considered [543]:


−γ1 e T PBφ(x) i f ||θ f || < m θ f
θ̇ f = (3.26)
0 ||θ f || ≥ m θ f
110 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

−γ2 e T PBφ(x)u c i f ||θg || < m θg
θ̇g = (3.27)
0 ||θg || ≥ m θg

θ̇ f and θ̇g are set to 0, when

||θ f || ≥ m θ f , ||θg || ≥ m θg . (3.28)

The update of θ f stands for a gradient algorithm on the cost function 21 ( f − fˆ)2 .
The update of θg is also of the gradient type, while u c implicitly tunes the adaptation
gain γ2 . Substituting Eqs. (3.26) and (3.27) into V̇ finally gives

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PB(w + d1 )−
−e T PB(θ f − θ ∗f )T φ(x) − e T PB(θg − θg∗ )T φ(x)u c ⇒
(3.29)

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PB(w + d1 ) + e T PBwα .

Denoting w1 = w + d1 + wα one gets

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PBw1 or equivalently,
(3.30)
V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + 21 e T PBw1 + 21 w1T B T Pe

Lemma: The following inequality holds:

1 T
2 e PBw1 + 21 w1T B T Pe − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (3.31)

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ2a2 + 1 2
ρ2
b − 2ab ≥ 0 ⇒ 21 ρ 2 a 2 + 2ρ1 2 b2 − ab ≥ 0 ⇒
(3.32)
ab − 1 2
2ρ 2
b ≤ 21 ρ 2 a 2 ⇒ 21 ab + 21 ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2

The following substitutions are carried out: a = w1 and b = e T PB and the previous
relation becomes
1 T
2 w1 B T Pe + 21 e T PBw1 − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (3.33)

The previous inequality is used in V̇ , and thus one arrives at

1 1
V̇ ≤ − e T Qe + ρ 2 w1T w1 (3.34)
2 2
3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 111

The attenuation coefficient ρ can be chosen such that the right part of Eq. (3.34) is
always upper bounded by 0. For instance, it suffices at any iteration of the control
algorithm to have

− 21 e T Qe + 21 ρ 2 ||w||21 ≤0⇒
− 21 ||e||2Q + 21 ρ 2 ||w||21 ≤0⇒ (3.35)
||e||2
ρ 2 ≤ ||w||Q2
1

Without knowledge of the uncertainty and disturbances term ||w||1 a sufficiently

small value of ρ can assure that the above inequality holds and thus that the control
loop stability can be assured. Actually, ρ should be given the least value that permits
to obtain a solution for the Riccati equation of Eq. (3.23).
Equation (3.34) can be used to show that the H∞ performance criterion is satisfied.
The integration of V̇ from 0 to T gives
T T T
0 V̇ (t)dt ≤ − 21 0 ||e||2 dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
(3.36)
T T
2V (T ) + 0 ||e|| Q dt
2 ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt

Moreover, if there exists a positive constant Mw > 0 such that

∞
0 ||w1 ||2 dt ≤ Mw (3.37)

then one gets

∞
0 ||e||2Q dt ≤ 2V (0) + ρ 2 Mw (3.38)
∞
Thus, the integral 0 ||e||2Q dt is bounded. Moreover, V (T ) is bounded and from the
definition of the Lyapunov function V in Eq. (3.20) it becomes clear that e(t) will be
also bounded since e(t) ∈ Ωe = {e|e T Pe≤2V (0) + ρ 2 Mw }.
According to the above and with the use of Barbalat’s Lemma one obtains the
asymptotic elimination of the tracking error, that is, lim t→∞ e(t) = 0 (Fig. 3.2).

3.3.5 Simulation Tests

3.3.5.1 Control of an Inverted Pendulum Model

The performance of the proposed flatness-based adaptive fuzzy controller was first
tested in the benchmark problem of cart-pole balancing. The derivation of the state
equations of the cart-pole system stems from the solution of the associated Euler-
Lagrange equation. The model presented here follows [543, 544] and considers only
the dynamics of the pole (inverted pendulum). Denoting by θ the angle of the pole
112 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

Fig. 3.2 The cart-pole balancing problem

(with respect to the vertical axis), m the mass of the pole, M the mass of the cart,
and l the length of the pole, one gets

θ̈ = f (x, t) + g(x, t)u (3.39)

where the nonlinear functions f (x, t) and g(x, t) are given by

mlx22 sin(x1 )cos(x1 )−(M+m)gsin(x1 )
f (x, t) = mlcos 2 (x1 )−2l(M+m)
and
(3.40)
cos(x1 )
g(x, t) = mlcos 2 (x1 )−2l(M+m)

Setting x1 = θ and x2 = θ̇ the state equation of the cart-pole system is obtained as

   
ẋ1 01 x1 0
= + v (3.41)
ẋ2 00 x2 1

where v = f (x, t) + g(x, t)u + d̃. Equation (3.41) describes a system in the
Brunovksy form. The flat output is taken to be y = x1 . It can be easily proven
that all state variables of the system of Eq. (3.41) and the associated control input can
be written as functions of the flat output y and its derivatives. Indeed, it holds that

x1 = y
x2 = ẏ (3.42)
u = g(y,
1
ẏ) [ ÿ − f (y, ẏ)]
3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 113

The following parameter values were chosen: m = 0.1 kg, M = 1.0 kgr, l = 0.5 m.
The basis functions used in the estimation of f (x, t) and g(x, t) were μ A j (x̂) =
x̂−c j 2
e( σ ) , j = 1, . . . , 5. Since there are two inputs x1 and ẋ1 and each fuzzy input
variable consists of 5 fuzzy sets there will be 25 fuzzy rules of the form:

R l : IF x is Al1 AND ẋ is Al2 THEN fˆl is bl and

25 (3.43)
ˆl 2 μl (xi )
l=1 f
fˆ(x, t) =
i=1 Ai

25 2
l=1 i=1 μ A (xi )
l
i

(l)
The centers ci , i = 1, 2 and the variances v(l) of each rule are as follows:

(l) (l)
Rule c1 c2 v(l)
R (1) −0.3 −0.03 3
R (2) −0.3 −0.01 3
R (3) −0.3 0.00 3
R (4) −0.3 0.01 3
R (5) −0.3 0.03 3
R (6) −0.1 −0.03 3
··· ··· ··· ···
··· ··· ··· ···
R (25) 0.3 0.03 3

The estimation ĝ(x, t) was derived in a similar way. A measurable state vector was
assumed and used for the training of the neuro-fuzzy approximators. The positive
definite matrix P was found from the solution of the algebraic Riccati equation (3.23),
where Q was chosen to be a multiple of the identity matrix I2 .
Two different setpoints were studied: (i) a sinusoidal signal of amplitude 0.1 and
period T = 15 s, (ii) a seesaw setpoint of amplitude 0.15 and period T = 15 s.
At the beginning of the second half of the simulation time an additive sinusoidal
disturbance of amplitude A = 1.0 and period T = 7.5 s was applied to the system.
The approximations fˆ and ĝ were used in the control law, given by Eq. (3.7). The
position and velocity variations for the sinusoidal setpoint are depicted in Fig. 3.3a, b,
respectively.
The performance of the proposed flatness-based adaptive fuzzy control is also
tested in the tracking of a seesaw setpoint. The associated position and velocity vari-
ation are demonstrated in Fig. 3.4a, b, respectively. The control signal in the case of
tracking of a sinusoidal setpoint is shown in Fig. 3.5a, while the control signal when
tracking a seesaw setpoint is shown in Fig. 3.5b. Finally, the approximation of func-
tion g(x, t) in the case of tracking of a sinusoidal setpoint is shown in Fig. 3.6a (and
is marked as a dashed line), while when tracking a seesaw setpoint the approximated
function g(x, t) is shown in Fig. 3.6b.
114 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

(a) 1.5 (b) 1

1
0.5
position x1 of the pendulum

velocity x2 of the pendulum

0.5
0

−0.5
−0.5

−1
−1

−1.5 −1.5
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 3.3 a Tracking of a sinusoidal position setpoint (red line) by the angle of the pendulum.
b Tracking of a sinusoidal velocity setpoint (red line) by the angular velocity of the pendulum

(a) 1.2 (b) 1.5

1
1
position x1 of the pendulum

velocity x2 of the pendulum

0.8
0.5
0.6

0.4 0

0.2 −0.5

0
−1
−0.2
−1.5
−0.4

−0.6 −2
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 3.4 a Tracking of a seesaw position setpoint (red line) by the angle of the pendulum. b Tracking
of a seesaw velocity setpoint (red line) by the angular velocity of the pendulum

Comparing the proposed flatness-based adaptive fuzzy control method with other
adaptive fuzzy control approaches and with the analysis on neural adaptive control
methods given in the relevant bibliography (e.g., [167, 168, 205]), the following
can be noted: (i) the transformation of the initial nonlinear system into the lin-
earized Brunovksy (canonical) form does not require the computation of partial
derivatives or Lie derivatives, (ii) there is no need to make restrictive assumptions
about the number of truncated higher order terms in the linearization of the system’s
nonlinear model or about a bounded error in the linearization of the output of the
neural/fuzzy approximators, (iii) the number of adaptable parameters involved in the
training of the neural/fuzzy approximators remains small and there is no need to use
an excessive number of neural/fuzzy approximators to produce the control signal,
3.3 Flatness-Based Adaptive Fuzzy Control for SISO Dynamical Systems 115

(a) 50 (b) 60

50
40

control signal of the pendulum

control signal of the pendulum

40

30
30

20 20

10
10

0
−10

−10 −20
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 3.5 a Control input of the pendulum when tracking a sinusoidal setpoint. b Control input of
the pendulum when tracking a seesaw setpoint

(a) 0 (b) 0

−0.1 −0.1
real vs estimated function g

real vs estimated function g

−0.2 −0.2

−0.3 −0.3

−0.4 −0.4

−0.5 −0.5

−0.6 −0.6

−0.7 −0.7
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 3.6 a Approximation of function g(x, t) of the pendulum model when tracking a sinusoidal
setpoint. b Approximation of function g(x, t) of the pendulum model when tracking a seesaw
setpoint

(iv) the tracking performance of the neuro-fuzzy control loop is evaluated with the use
of specific metrics (e.g., H∞ tracking performance), and (v) the proposed flatness-
based control method can also be extended to MIMO systems [426, 433].
It is also noted that the proposed flatness-based adaptive fuzzy control law assures
stability of the control loop and the asymptotic elimination of the tracking error.
Therefore, it can be ensured that θ̃g = θg − θg∗ →0 which means that ĝ(x)→g(x).
Provided that function g(x) does not become zero as long as x remains bounded, then
ĝ(x) will be also different from zero. For the cart-pole system described in Eqs. (3.39)
and (3.40) it holds that g(x) = 0 for x1 = θ = 0, however, for x1 ∈ (0, π ) this case is
116 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

not going to occur, even if ideally there is no function approximation error and ĝ(x)
coincides with g(x). In the generic case, to assure the avoidance of singularities in
the proposed control scheme one has to exclude singularity points from the reference
trajectory that the system’s state vector has to track.
Singularities may appear not only in the proposed adaptive fuzzy control scheme
but in all control systems that are based on static feedback linearization. For exam-
ple, linearization of the system through the use of a new control input of the form
v = f (x) + g(x)u means that u = g(x)−1 [v − f (x)], which does not exclude
the appearance of singularities. Therefore, singularities do not concern only the
proposed control method but the whole class of static feedback-based linearization
schemes. Modifications, as described in Chap. 2, can be introduced in the design of
the controller to prohibit the appearance of singularities, for example, a change of
coordinates that results in a new state-space representation that does not include any
points of singularity [340].

3.4 Flatness-Based Adaptive Fuzzy Control for MIMO

Systems

3.4.1 Overview

The results of Sect. 3.2 and can be extended toward adaptive fuzzy control based on
differential flatness theory for multi-input multi-output (MIMO) nonlinear dynamical
systems. There are numerous control problems for electromechanical systems with
MIMO nonlinear dynamics. The application of geometric techniques and mainly of
the feedback linearization method has enabled advancements in this research direc-
tion [428, 434, 524, 543, 544]. These techniques can be applied to nonlinear systems,
of uncertain or unknown dynamics. Based on the universal approximation theorem,
many stable adaptive fuzzy control schemes have been developed for unknown multi-
input multi-output (MIMO) dynamical systems [89, 91, 92, 289, 301, 501, 508, 526,
527]. The capability of neuro-fuzzy controllers to compensate for model parametric
uncertainties, external disturbances, as well as for incomplete measurement of the
system’s state vector has been analyzed in several studies [14, 110, 288, 484, 523,
524]. Moreover, results on neural and fuzzy control have been presented for MIMO
dynamical systems subjected to parametric changes and external perturbations, as
well as for systems in which model uncertainty may come from the inability to mea-
sure precisely the complete state vector. [101, 164, 302, 393, 527]. To improve the
robustness of these control schemes a supervisory control term is added to the control
signal, which can take the form of sliding mode control or H∞ control [92, 96, 291].
In this chapter an approach to the design of robust controllers for MIMO nonlinear
dynamical systems of unknown dynamics will be developed according to flatness-
based control theory. This approach extends the class of nonlinear systems to which
adaptive fuzzy control can be applied.
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 117

Flatness-based control is shown to be efficient for systems characterized by model

uncertainties and exogenous disturbances [189, 190]. In this chapter, the results of
Sect. 3.2 are generalized and applied to multi-input multi-output (MIMO) dynamics
systems. The chapter is concerned with differentially flat multi-input multi-output
(MIMO) dynamical systems which can be written in the Brunovksy (canonical) form.
Transformation into the Brunovksy form can be succeeded for systems that admit
static feedback linearization (i.e., a change of coordinates for both the system state
variables and the system’s control input). For such flat multi-input systems necessary
and sufficient conditions that allow an endogenous transformation into Brunovsky
coordinates will be also stated. In subsequent parts of this book the results on adaptive
fuzzy control based on differential flatness theory will be extended to MIMO systems
that admit only dynamic feedback linearization (Chaps. 6 and 10).
After transforming a MIMO system into canonical form, the resulting control
inputs are shown to contain nonlinear elements which depend on the system’s para-
meters. If the parameters of the system are unknown, the nonlinear terms that appear
in the control inputs can be approximated with the use of neuro-fuzzy networks. In
this chapter it is shown that a suitable learning law can be defined for the aforemen-
tioned neuro-fuzzy approximators so as to preserve the closed-loop MIMO system
stability. Lyapunov stability analysis proves also that the proposed flatness-based
adaptive fuzzy control scheme results in H∞ tracking performance, in accordance
to the results of [407, 408, 413, 433].

3.4.2 Differential Flatness for MIMO Nonlinear Dynamical

Systems

3.4.2.1 Conditions for Applying Differential Flatness Theory

The transformation of MIMO nonlinear systems into the canonical (Brunovsky) from,
through the application of differential flatness theory, are explained first. Next, a new
control method for such systems are developed, also in accordance with differential
flatness theory. It is shown that the proposed control and filtering methods can be
efficiently applied to the model of the nonlinear MIMO dynamical systems.
The MIMO nonlinear dynamical system can be written as a state-space model
of the form ẋ = f (x, u). Moreover, the previous state-space description can be
transformed to the form of an affine in-the-input system by adding an integrator to
each input [55, 286]. In the latter case one obtains

m
ẋ = f (x) + i=1 gi (x)u i (3.44)

The following definitions are now used [437]:

(i) Lie derivative: L f h(x) stands for the Lie derivative L f h(x) = (∇h) f and
the repeated Lie derivatives are recursively defined as L 0f h = h for i = 0,
L if h = L f L i−1 i−1
f h = ∇ L f h f for i = 1, 2, . . ..
118 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

(ii) Lie Bracket: ad if g stands for a Lie Bracket which is defined recursively as
ad if g = [ f, ad i−1
f g] with ad f g = g and ad f g = [ f, g] = ∇g f − ∇fg.
0

If the system of Eq. (3.44) can be linearized by a diffeomorphism z = φ(x) and a

static state feedback u = α(x) + β(x)v in the following form:

ż i, j = z i+1, j for 1≤ j≤m and 1≤i≤v j − 1

(3.45)
ż vi, j = v j

with mj=1 v j = n, then y j = z 1, j for 1≤ j≤m are the 0-flat outputs which can be
written as functions of only the elements of the state vector x. In accordance with the
linearization criteria provided in Chap. 1, and to provide conditions for transforming
the system of Eq. (3.44) into the canonical form described in Eq. (3.45) the following
theorem has been stated [55]:
Theorem: For nonlinear systems described by Eq. (3.44) the following variables
are defined: (i) G 0 = span[g1 , . . . , gm ], (ii) G 1 = span[g1 , . . . , gm , ad f g1 , . . . ,
j
ad f gm ], . . . (k) G k = span{ad f gi for 0≤ j≤k, 1≤i≤m}. Then, the linearization
problem for the system of Eq. (3.44) can be solved if and only if: (1) The dimension
of G i , i = 1, . . . , k is constant for x ∈ X ⊆R n and for 1≤i≤n −1, (2) The dimension
of G n−1 if of order n, (3) The distribution G k is involutive for each 1≤k≤n − 2.

3.4.2.2 Transformation of the Nonlinear MIMO Dynamics

into Canonical State-Space Form

It is assumed now that after defining the flat outputs of the previous state-space
description of the nonlinear MIMO dynamics and after expressing the system’s state
variables and control inputs as functions of the flat output and of the associated
derivatives, the system can be transformed into the Brunovsky canonical form:

ẋ1 = x2
···
ẋr1 −1 = xr1

p
ẋr1 = f 1 (x) + j=1 g1 j (x)u j + d1

ẋr1 +1 = xr1 +2
··· (3.46)
ẋ p−1 = x p
p
ẋ p = f p (x) + j=1 g p j (x)u j + d p

y1 = x1
···
y p = xn−r p +1

where x = [x1 , . . . , xn ]T is the state vector of the transformed system (according to

the differential flatness formulation), u = [u 1 , . . . , u p ]T is the set of control inputs,
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 119

y = [y1 , . . . , y p ]T is the output vector, f i are the drift functions and gi, j , i, j =
1, 2, . . . , p are smooth functions corresponding to the control input gains, while d j
is a variable associated to external disturbances. It holds that r1 + r2 + · · · + r p = n.
Having written the initial nonlinear system in the canonical (Brunovsky) form it
holds that
p
(r )
yi i = f i (x) + j=1 gi j (x)u j + d j (3.47)

Next, the following vectors and matrices can be defined: f (x) = [ f 1 (x), . . . ,
f n (x)]T , g(x) = [g1 (x), . . . , gn (x)]T , with gi (x) = [g1i (x), . . . , g pi (x)]T , A =
diag[A1 , . . . , A p ], and B = diag[B1 , . . . , B p ], C T = diag[C1 , . . . , C p ], d =
[d1 , . . . , d p ]T , where matrix A has the MIMO canonical form, i.e., with block-
diagonal elements
⎛ ⎞
0 1 ··· 0
⎜0 0 · · · 0 ⎟
⎜ ⎟
⎜ ⎟
Ai = ⎜ ... ... · · · ... ⎟
⎜ ⎟
⎝0 0 · · · 1 ⎠
(3.48)
0 0 · · · 0 r ×r
i i

 
BiT = 0 0 · · · 0 1 1×r
  i
Ci = 1 0 · · · 0 0 1×r
i

Thus, Eq. (3.47) can be written in state-space form

ẋ = Ax + Bv + B d̃
(3.49)
y = Cx

where the control input is written as v = f (x) + g(x)u. The system of Eqs. (3.48)
and (3.49) is in controller and observer canonical form.
The reference setpoints for the system’s outputs y1 , . . . , y p are denoted as
y1m , . . . , y pm , thus for the associated tracking errors it holds that

e1 = y1 − y1m
e2 = y2 − y2m
(3.50)
···
e p = y p − y pm

The error vector of the outputs of the transformed MIMO system is denoted as

E 1 = [e1 , . . . , e p ]T
ym = [y1m , . . . , y pm ]T
(3.51)
···
ym(r ) = [y1m
(r )
, . . . , y (r ) T
pm ]
120 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

(r )
where yim denotes the r th order derivative of the ith reference output of the MIMO
dynamical system. Thus, one can also define the following vectors: (i) a vector
containing the state variables of the system and the associated derivatives, (ii) a
vector containing the reference outputs of the system and the associated derivatives
r −1 T
x = [x1 , . . . , x1r1 −1 , . . . , x p , . . . , x pp ] (3.52)

r1 −1 r −1
Ym = [y1m , . . . , y1m , . . . , y pm , . . . , y pm
p
]T (3.53)

while in a similar manner one can define a vector containing the tracking error of the
system’s outputs and the associated derivatives
r −1 T
e = Ym − x = [e1 , . . . , e1r1 −1 , . . . , e p , . . . , e pp ] (3.54)

It is assumed that matrix g(x) is a nonsingular one, i.e., g −1 (x) exists and is bounded
for all x ∈ Ux , where Ux ⊂R n is a compact set. In any case, the problem of singular-
ities in matrix g(x) can be handled by adding to ĝ(x) a small perturbation ĝ which
results in avoidance of the singularity points and which is asymptotically eliminated
by the robustness of the control loop.
The objective of the adaptive fuzzy controller, denoted as u = u(x, e|θ ) is: (i)
all the signals involved in the controller’s design are bounded and it holds that
lim t→∞ e = 0, (ii) the H∞ tracking performance criterion is succeeded for a pre-
scribed attenuation level.
As in the case of adaptive fuzzy control for single-input dynamical systems
described in Sect. 3.2, successful tracking of the reference signal is denoted by the
H∞ criterion. This has already been formulated as [413, 524]:
T T
0 e T Qedt ≤ ρ 2 0 wd T wd dt (3.55)

where ρ is the attenuation level coefficient and denotes the aggregate effect of model
uncertainty and external perturbations.

3.4.3 Flatness-Based Adaptive Fuzzy Control for MIMO

Nonlinear Systems

3.4.3.1 Control Law for MIMO Nonlinear Systems

The control signal of the MIMO nonlinear system which has been transformed into
the Brunovsky form as described by Eq. (3.49) contains the unknown nonlinear func-
tions f (x) and g(x), which can be approximated by

fˆ(x|θ f ) = Φ f (x)θ f
(3.56)
ĝ(x|θg ) = Φg (x)θg
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 121

where

Φ f (x) = (ξ 1f (x), ξ 2f (x), . . . ξ nf (x))T (3.57)

with ξ if (x), ı = 1, . . . , n being the vector of kernel functions (e.g., normalized fuzzy
Gaussian membership functions), where

ξ if (x) = (φ i,1 i,2 i,N

f (x), φ f (x), . . . , φ f (x)) (3.58)

thus giving
⎛ ⎞
φ 1,1 1,2 1,N
f (x) φ f (x) · · · φ f (x)
⎜ 2,1 ⎟
⎜φ f (x) φ 2,2 2,N
f (x) · · · φ f (x)⎟
⎟
Φ f (x) = ⎜
⎜ ··· (3.59)
⎝ ··· ··· ··· ⎟ ⎠
φ n,1
f (x) φ n,2
f (x) · · · φ n,N
f (x)

while the weights vector is defined as

 
θ f T = θ 1f , θ 2f , · · · θ Nf (3.60)

j = 1, . . . , N is the number of basis functions used to approximate the components

of function f , which are denoted as i = 1, . . . , n. Thus, one obtains the relation of
Eq. (3.56), i.e. fˆ(x|θ f ) = Φ f (x)θ f .
In a similar manner, for the approximation of function g one has
 T
Φg (x) = ξg1 (x), ξg2 (x), · · · ξgN (x) (3.61)

with ξgi (x), ı = 1, . . . , N being the vector of kernel functions (e.g., normalized fuzzy
Gaussian membership functions), where
 
ξgi (x) = φgi,1 (x), φgi,2 (x), . . . , φgi,N (x) (3.62)

thus giving
⎛ ⎞
φg1,1 (x) φg1,2 (x) · · · φg1,N (x)
⎜ 2,1 2,2 2,N ⎟
⎜φ (x) φg (x) · · · φg (x)⎟
Φg (x) = ⎜ g ⎟ (3.63)
⎝ ··· ··· ··· ··· ⎠
φgn,1 (x) φgn,2 (x) · · · φgn,N (x)

while the weights vector is defined as

 p T
θg = θg1 , θg2 , . . . , θg (3.64)
122 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

where the components of matrix θg are defined as

 
j
θg = θgj1 , θgj2 , · · · θgjN (3.65)

j = 1, . . . , N is the number of basis functions used to approximate the components

of function g which are denoted as i = 1, . . . , n. Thus, one obtains about matrix
θg ∈ R N × p
⎛ 1 2 p ⎞
θg1 θg1 · · · θg1
⎜ 1 2 p ⎟
⎜ θ θ · · · θg2 ⎟
θg = ⎜ g2 g2 ⎟ (3.66)
⎝··· ··· ··· ···⎠
p
θg1N θg2N · · · θg N

It holds that
⎛ ⎞ ⎛ 1 p⎞
g1 g1 g12 · · · g1
⎜ g2 ⎟ ⎜ p⎟
⎟ ⎜ g2 g22 · · · g2 ⎟
1
g=⎜
⎝ · · ·⎠ = ⎜ ⎟ (3.67)
⎝· · · · · · · · · · · ·⎠
gn p
gn1 gn2 · · · gn

Using the above, one finally has the relation of Eq. (3.56), i.e., ĝ(x|θg ) = Φg (x)θg .
If the state variables of the system are available for measurement then a state-
feedback control law can be formulated as
(r )
u = ĝ −1 (x|θg )[− fˆ(x|θ f ) + ym + K T e + u c ] (3.68)

where fˆ(x|θ f ) and ĝ(x|θg ) are fuzzy models to approximate f (x) and g(x), respec-
tively. u c is a supervisory control term, e.g., H∞ control term which is used to com-
pensate for the effects of modeling inaccuracies and external disturbances. Using the
system’s state-space description of Eq. (3.49) the control term u c is defined as

u c = − r1 B T Pe (3.69)

Moreover, K T is the feedback gain matrix that assures that the characteristic poly-
nomial of matrix A − B K T will be a Hurwitz one.

3.4.4 Flatness-Based Control for a MIMO Robotic

Manipulator

Adaptive fuzzy control for MIMO nonlinear dynamical systems with the use
of differential flatness theory will be explained through an application example.
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 123

A 2-DOF rigid link robotic manipulator is considered. The dynamic model of the
robot is a differentially flat one. Indeed, it holds that
    
M11 M12 θ̈1 F1 (θ, θ̇ ) G 1 (θ ) T
+ + = 1 (3.70)
M21 M22 θ̈2 F2 (θ, θ̇ ) G 2 (θ ) T2

where M ∈ R 2×2 is the robot’s inertia matrix, F ∈ R 2×1 is the Coriolis and cen-
trifugal forces matrix and G ∈ R 2×1 is the gravitational effects matrix. The above
dynamic model is obtained with the use of the Euler-Lagrange method [160]. Equiv-
alently,
  −1 
θ̈1 M11 M12 F1 (θ, θ̇ )
=− −
θ̈2 M21 M22 F2 (θ, θ̇ )
 −1   −1  (3.71)
M11 M12 G 1 (θ ) M11 M12 T1
− +
M21 M22 G 2 (θ ) M21 M22 T2

Denoting the inverse of the inertia matrix as

 
M11 M12 N11 N12
= (3.72)
M21 M22 N21 N22

one obtains
  
θ̈1 N11 N12 F1 (θ, θ̇ )
=− −
θ̈2 N21 N22 F2 (θ, θ̇ )
    (3.73)
N11 N12 G 1 (θ ) N11 N12 T1
− +
N21 N22 G 2 (θ ) N21 N22 T2

which can be also written as

θ̈1 = −N11 F1 (θ, θ̇ ) − N12 F2 (θ, θ̇ )−

(3.74)
−N11 G 1 (θ ) − N12 G 2 (θ ) + N11 T1 + N12 T2

θ̈2 = −N21 F1 (θ, θ̇ ) − N22 F2 (θ, θ̇ )−

(3.75)
= N21 G 1 (θ ) − N22 G 2 (θ ) + N21 T1 + N22 T2

The following state variables are defined:

x1 = θ1 , x2 = θ̇1 , x3 = θ2 , x4 = θ̇2 (3.76)

It holds that

ẍ1 = f 1 (x) + g1 (x)u

(3.77)
ẍ3 = f 2 (x) + g2 (x)u
124 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

where

f 1 (x) = −N11 F1 (θ, θ̇ ) − N12 F2 (θ, θ̇ )−

−N11 G 1 (θ ) − N12 G 2 (θ ) ∈ R 1×1

g1 (x) = [N11 N12 ] ∈ R 1×2

(3.78)
f 2 (x) = −N21 F2 (θ, θ̇ ) − N22 F2 (θ, θ̇ )−
−N21 G 1 (θ ) − N22 G 2 (θ ) ∈ R 1×1

g2 (x) = [N21 N22 ] ∈ R 2×2

The flat output is defined as

y = [θ1 , θ2 ] = [x1 , x3 ] (3.79)

It holds that

ẋ1 = x2
ẋ2 = f 1 (x) + g1 (x)u
(3.80)
ẋ3 = x4
ẋ4 = f 2 (x) + g2 (x)u

therefore, all system state variables can be written as functions of the flat output y
and its derivatives. The same holds for the control input u

x1 = [1 0]y T x2 = [1 0] ẏ T
(3.81)
x3 = [0 1]y T x4 = [0 1] ẏ T

Moreover, from Eq. (3.80) it holds that

  
ẍ1 f 1 (x) g1 (x)
= + u i.e.
ẍ3 f 2 (x) g2 (x)
(3.82)
 −1   
g (x) ẍ1 f (x)
u= 1 − 1
g2 (x) ẍ3 f 2 (x)

Knowing that x = h(y, ẏ) one finally obtains

 −1   
g1 (h(y, ẏ)) [1 0] ÿ T f 1 (h(y, ẏ))
u= − (3.83)
g2 (h(y, ẏ)) [0 1] ÿ T f 2 (h(y, ẏ))

Therefore, the considered robotic system is a differentially flat one. Next, taking
into account also the effects of additive disturbances to the joints of the robotic
manipulator the dynamic model becomes
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 125

ẍ1 = f 1 (x, t) + g1 (x, t)u + d̃1

(3.84)
ẍ3 = f 2 (x, t) + g2 (x, t)u + d̃2
   
ẍ1 f 1 (x, t) g1 (x, t) d̃
= + u+ 1 (3.85)
ẍ3 f 2 (x, t) g2 (x, t) d̃2

The following control input is defined:

 −1  d   T  
ĝ (x, t) ẍ1 fˆ1 (x, t) K1 u
u= 1 · − ˆ − e + c1 (3.86)
ĝ2 (x, t) ẍ3d f 2 (x, t) K 2T u c2

where [u c1 u c2 ]T is a robust control term that is used for the compensation of the
model’s uncertainties as well as of the external disturbances and the vector of the
feedback gain is K iT = [k1i , k2i , . . . , kn−1
i , kni ].
Substituting Eq. (3.86) into Eq. (3.85) the closed-loop tracking error dynamics is
obtained
    −1
ẍ1 f (x, t) g (x, t) ĝ1 (x, t)
= 1 + 1 ·
ẍ3 f 2 (x, t) g2 (x, t) ĝ2 (x, t)
       (3.87)
ẍ1d fˆ1 (x, t) K 1T u c1 d̃1
− − e + +
ẍ3d fˆ2 (x, t) K 2T u c2 d̃2

Equation (3.87) can now be written as

 
ẍ1 f 1 (x, t)
= +
ẍ3 f 2 (x, t)
    −1
g1 (x, t) − ĝ1 (x, t) ĝ1 (x, t) ĝ1 (x, t)
+ + · (3.88)
 g2 (x, t) 
− ĝ2 (x, t)  ĝ

2 (x, t)

ĝ2 (x, t)

ẍ1d fˆ (x, t) K 1T u d̃
· d − ˆ1 − e + c1 + 1
ẍ3 f 2 (x, t) K 2T u c2 d̃2

and using Eq. (3.86) this results in

  
ë1 f (x, t) − fˆ1 (x, t) g1 (x, t) − ĝ1 (x, t)
= 1 + u−
ë3 f (x, t) − fˆ2 (x, t) g2 (x, t) − ĝ2 (x, t)
 2    (3.89)
K 1T u d̃
− e + c1 + 1
K 2T u c2 d̃2

The following description for the approximation error is defined:

 
f 1 (x, t) − fˆ1 (x, t) g1 (x, t) − ĝ1 (x, t)
w= + u (3.90)
f 2 (x, t) − fˆ2 (x, t) g2 (x, t) − ĝ2 (x, t)
126 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

Moreover, the following matrices are defined:

⎛ ⎞ ⎛ ⎞
0100 00
⎜0 0 0 0⎟ ⎜1 0 ⎟
A=⎝ ⎜ ⎟ , B=⎝ ⎟ ⎜
0 0 0 1⎠ 0 0⎠
0000 01 (3.91)
 1 1 1 1
K1 K2 K3 K4
KT =
K 12 K 22 K 32 K 42

Using matrices A, B, K T , Eq. (3.89) is written in the following form:


f 1 (x, t) − fˆ1 (x, t)
ė = (A − B K )e + Bu c + B
T
ˆ +
  f 2 (x, t) − f 2 (x, t) (3.92)
g (x, t) − ĝ1 (x, t)
+ 1 u + d̃
g2 (x, t) − ĝ2 (x, t)

Next, the following approximators of the unknown system dynamics are defined:

fˆ (x|θ f ) x ∈ R 4×1 fˆ1 (x|θ f ) ∈ R 1×1
fˆ(x) = ˆ1 (3.93)
f 2 (x|θ f ) x ∈ R 4×1 fˆ2 (x|θ f ) ∈ R 1×1

with kernel functions

n
j=1 μ A j (x j )
i
i, j
φ f (x) =
N n (3.94)
i=1 j=1 μ A (x j )
i
j

where l = 1, 2 and μ Ai (x) is the ith membership function of the antecedent (IF) part
j
of the lth fuzzy rule. Similarly, the following approximators of the unknown system
dynamics are defined:

ĝ (x|θg ) x ∈ R 4×1 ĝ1 (x|θg ) ∈ R 1×2
ĝ(x) = 1 (3.95)
ĝ2 (x|θg ) x ∈ R 4×1 ĝ2 (x|θg ) ∈ R 1×2

The values of the weights that result in optimal approximation are

θ ∗f = arg minθ f ∈Mθ f [supx∈Ux ( f (x) − fˆ(x|θ f ))]

(3.96)
θg∗ = arg minθg ∈Mθg [supx∈Ux (g(x) − ĝ(x|θg ))]

where the variation ranges for the weights are defined as

Mθ f = {θ f ∈ R h : ||θ f ||≤m θ f }
(3.97)
Mθg = {θg ∈ R h : ||θg ||≤m θg }

For the value of the approximation error defined in Eq. (3.90) that corresponds to the
optimal values of the weights vectors θ ∗f and θg∗ one has
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 127
 
w= f (x, t) − fˆ(x|θ ∗f ) +
  −1 (3.98)
+ g(x, t) − ĝ(x|θg∗ ) ĝ(x, t) u

which is next written as

 
w = f (x, t) − fˆ(x|θ f ) + fˆ(x|θ f ) − fˆ(x|θ ∗f ) +
  −1 (3.99)
+ g(x, t) − ĝ(x|θg ) + ĝ(x|θg ) − ĝ(x|θg∗ ) ĝ(x, t) u

which can be also written in the following form:

 
w = wa + wb (3.100)

where
 −1
wa = {[ fˆ(x, θ f ) − fˆ(x|θ ∗f )] + [ĝ(x, θg ) − ĝ(x|θg∗ )]} · ĝ(x, t) u (3.101)

wb = {[ f (x, t) − fˆ(x|θ f )] + [g(x, t) − ĝ(x|θg )]}

 −1 (3.102)
ĝ(x, t) u

Moreover, the following weights error vectors are defined:

θ̃ f = θ f − θ ∗f
(3.103)
θ̃g = θg − θg∗

3.4.5 Lyapunov Stability Analysis for MIMO Nonlinear

Systems

3.4.5.1 Stability Proof for the Control Loop

The following quadratic Lyapunov function is defined:

1 T 1 T 1
V = e Pe + θ̃ f θ̃ f + tr [θ̃gT θ̃g ] (3.104)
2 2γ1 2γ2

Parameter γ1 is the learning rate used in the adaptation of the weights of the neu-
rofuzzy approximator for f (x), while parameter γ2 is the learning rate used in the
adaptation of the weights of the neurofuzzy approximation for g(x). It holds that

1 ˙T ˙T
V̇ = 21 ė T Pe + 21 e T P ė + γ1 θ̃ f θ̃ f + γ2 tr [θ̃g θ̃g ]
1 (3.105)
128 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

The tracking error dynamics is described by


f 1 (x, t) − fˆ1 (x, t)
ė = (A − B K T )e + Bu c + B ˆ +
  f 2 (x, t) − f 2 (x, t) (3.106)
g (x, t) − ĝ1 (x, t)
+ 1 u + d̃
g2 (x, t) − ĝ2 (x, t)

and defining the approximation error

 
f (x, t) − fˆ1 (x, t) g1 (x, t) − ĝ1 (x, t)
w= 1 + u (3.107)
f 2 (x, t) − fˆ2 (x, t) g2 (x, t) − ĝ2 (x, t)

the previous relation can be also written as

ė = (A − B K T )e + Bu c + B(w + d̃) (3.108)

From Eq. (3.105) one obtains

V̇ = 21 {e T (A − B K T )T + u cT B T +
+(w + d̃)T B T }Pe + 21 e T P{(A − B K T )e+ (3.109)
+Bu c + B(w + d̃)} + γ11 θ̃˙ Tf θ̃ f + γ12 tr [θ̃˙gT θ̃g ]

which in turn gives

V̇ = 21 e T {(A − B K T )T P + P(A − B K T )}e+

2 2e PBu c + 2 2B Pe(w + d̃)+
1 T 1 T
(3.110)
˙
+ γ11 θ̃ Tf θ̃ f + γ12 tr [θ̃˙gT θ̃g ]

Assumption 1: For given positive definite matrix Q there exists a positive definite
matrix P, which is the solution of the following matrix equation:

T
(A − B K T ) P + P(A − B K T ) − PB( r2 − 1
ρ2
)B T P+Q=0 (3.111)

Substituting Eqs. (3.111) and (3.69) into V̇ yields after some operations

V̇ = 21 e T {−Q + PB( r2 − 1
ρ2
)B T P}e+
(3.112)
e T PB{− r1 B T Pe} + B T P(w + d̃) + γ11 θ̃˙ Tf θ̃ f + γ12 tr [θ̃˙gT θ̃g ]
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 129

Therefore, it holds that

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PB(w + d̃)+
(3.113)
1 ˙T ˙T
γ1 θ̃ f θ̃ f + γ2 tr [θ̃g θ̃g ]
1

It also holds that

θ̃˙ f = θ̇ f − θ̇ ∗f = θ̇ f
(3.114)
θ̃˙g = θ̇g − θ̇g∗ = θ̇g

The following weights adaptation law is used:

θ̇ f = −γ1 Φ(x)T B T Pe
(3.115)
θ̇g = −γ2 Φ(x)T B T Peu T

where assuming N fuzzy rules and associated kernel functions the matrix dimensions
are θ f ∈ R N ×1 , θg ∈ R N ×2 , Φ(x) ∈ R 2×N , B ∈ R 4×2 , P ∈ R 4×4 and e ∈ R 4×1 .
Therefore, it holds that

V̇ = − 21 e T Qe − 2ρ1 2 e T PBB T Pe + e T PB(w + d̃)+

+ γ11 (−γ1 )e T PBΦ(x)(θ f − θ ∗f )+ (3.116)
+ γ12 (−γ2 )tr [ue T PBΦ(x)(θg − θg∗ )]

or

V̇ = − 21 e T Qe − 2ρ1 2 e T PBB T Pe + e T PB(w + d̃)+

+ γ11 (−γ1 )e T PBΦ(x)(θ f − θ ∗f )+ (3.117)
+ γ2 (−γ2 )tr [ue T PB(ĝ(x|θg ) − ĝ(x|θg∗ )]
1

Taking into account that u ∈ R 2×1 and e T PB(ĝ(x|θg ) − ĝ(x|θg∗ )) ∈ R 1×2 it holds
that

V̇ = − 21 e T Qe − 2ρ1 2 e T PBB T Pe + e T PB(w + d̃)+

+ γ11 (−γ1 )e T PBΦ(x)(θ f − θ ∗f )+ (3.118)
+ γ12 (−γ2 )tr [e T PB(ĝ(x|θg ) − ĝ(x|θg∗ ))u]

Since e T PB(ĝ(x|θg ) − ĝ(x|θg∗ ))u ∈ R 1×1 it holds that

tr (e T PB(ĝ(x|θg ) − ĝ(x|θg∗ )u) =

(3.119)
= e T PB(ĝ(x|θg ) − ĝ(x|θg∗ ))u
130 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

Therefore, one finally obtains

V̇ = − 21 e T Qe − 2ρ1 2 e T PBB T Pe + e T PB(w + d̃)+

+ γ11 (−γ1 )e T PBΦ(x)(θ f − θ ∗f )+ (3.120)
+ γ12 (−γ2 )e T PB(ĝ(x|θg ) − ĝ(x|θg∗ ))u

Next the following approximation error is defined:

wα = [ fˆ(x|θ f ) − fˆ(x|θ ∗f )] + [ĝ(x|θg ) − ĝ(x|θg∗ )]u (3.121)

Thus, one obtains

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe+
(3.122)
+e T PB(w + d̃) + e T P Bwα

Denoting the aggregate approximation error and disturbances vector as

w1 = w + d̃ + wα (3.123)

the derivative of the Lyapunov function becomes

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PBw1 (3.124)

which in turn is written as

V̇ = − 21 e T Qe − 2ρ1 2 e T PBB T Pe+

(3.125)
+ 21 e T PBw1 + 21 w1T B T Pe

Next, a Lemma is introduced:

Lemma: The following inequality holds:
1 T
2 e PBw1 + 21 w1T B T Pe − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (3.126)

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ2a2 + 1 2
ρ2
b − 2ab ≥ 0 ⇒ 21 ρ 2 a 2 + 2ρ1 2 b2 − ab ≥ 0 ⇒
(3.127)
ab − 1 2
2ρ 2
b ≤ 21 ρ 2 a 2 ⇒ 21 ab + 21 ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2

The following substitutions are carried out: a = w1 and b = e T PB and the previous
relation becomes
1 T
2 w1 B T Pe + 21 e T PBw1 − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (3.128)
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 131

The previous inequality is used in V̇ , and the right part of the associated inequality
is enforced
1 1
V̇ ≤ − e T Qe + ρ 2 w1T w1 (3.129)
2 2
As explained in the SISO systems control case, the attenuation coefficient ρ can be
chosen such that the right part of Eq. (3.129) is always upper bounded by 0. For
instance, it suffices at every iteration of the control algorithm to have

− 21 e T Qe + 21 ρ 2 ||w1 ||2 ≤ 0⇒ − 21 ||e||2Q + 21 ρ 2 ||w1 ||2 ≤ 0⇒

||e||2Q (3.130)
2 ρ ||w1 || ≤ 21 ||e||2Q + ⇒ρ 2 ≤
1 2 2
||w1 ||2

Again without knowledge of the uncertainties and disturbance term ||w1 || a suffi-
ciently small value of ρ can assure that the above inequality holds and thus that the
loop’s stability is assured. Actually, ρ should be given the least value which permits
to obtain a solution of the Riccati equation (3.111).
Equation (3.129) can be used to show that the H∞ performance criterion is satis-
fied. The integration of V̇ from 0 to T gives
T T T
0 V̇ (t)dt ≤ − 21 0 ||e||2 dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
(3.131)
T T
2V (T ) + 0 ||e|| Q dt
2 ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt

Moreover, if there exists a positive constant Mw > 0 such that

∞
0 ||w1 ||2 dt ≤ Mw (3.132)

one gets
∞
0 ||e||2Q dt ≤ 2V (0) + ρ 2 Mw (3.133)
∞
Thus, the integral 0 ||e||2Q dt is bounded and according to Barbalat’s Lemma one
obtains lim t→∞ e(t) = 0.
In the presented adaptive fuzzy control approach there is no prior assumption
about bounded weights of the neuro-fuzzy approximators. It is the Lyapunov stability
analysis that assures that the tracking error of the control loop will follow the H∞
tracking performance criterion and that the weights values will remain bounded.
The only assumption needed for the design of the controller and for succeeding H∞
tracking performance for the control loop is that there exists a solution for a Riccati
equation associated to the linearized error dynamics of the differentially flat model.
This assumption is quite reasonable for several nonlinear systems, thus providing a
systematic approach to the design of reliable controllers for such systems.
132 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

It is also noted that the training of the neuro-fuzzy networks and the approximation
of the nonlinear system dynamics is efficient if the persistence of excitation condi-
tion holds. The problem of persistence of excitation in the training of neuro-fuzzy
approximators has been studied in several research articles [98, 179, 542]. According
to [179], a Persistence of Excitation condition on the input variables of the neuro-
fuzzy model holds if the input patterns belong to domains around the network nodes
centers. In the case of a neuro-fuzzy approximator that comprises Gaussian basis
functions centered on a regular mesh, these domains cover part of the input domain
volume. Therefore, sufficient coverage of the input space by basis functions makes
the Persistence of Excitation condition hold.

3.4.5.2 The Role of Riccati Equation Coefficients in H∞ Control

Robustness

The linearized tracking error dynamics of Eq. (3.108) is considered again, i.e.,

ė = (A − B K T )e + Bu c + B(w + d̃)

The aim of H∞ control is to eliminate the impact of the modeling errors w =

[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u and the external disturbances d̃ which are
not white noise signals. This implies the minimization of the quadratic cost function
[259, 316]:
T
J (t) = 21 0 e T (t)e(t) + r u cT (t)u c (t)−
(3.134)
−ρ 2 (w + d̃)T (w + d̃)dt, r, ρ > 0

The weight r determines how much the control signal should be penalized and the
weight ρ determines how much the disturbances influence should be rewarded in the
sense of a mini-max differential game. The H∞ control law is u(t) = − r1 B T Pe(t)
where P is the positive definite symmetric matrix derived from the algebraic Riccati
equation Eq. (3.111).
The parameter ρ in Eq. (3.134) is an indication of the closed-loop system robust-
ness. If the values of ρ > 0 are excessively decreased with respect to r , then the
solution of the Riccati equation is no longer a positive definite matrix. Consequently,
there is a lower bound ρmin of ρ for which the H∞ control problem has a solution.
The acceptable values of ρ lie in the interval [ρmin , ∞). If ρmin is found and used in
the design of the H∞ controller, the closed-loop system will have increased robust-
ness. Unlike this, if a value ρ > ρmin is used, then an admissible stabilizing H∞
controller will be derived but it will be a suboptimal one. The Hamiltonian matrix
 
A −( r2 − ρ12 )B B T
H= (3.135)
−Q −A T
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 133

provides a criterion for the existence of a solution of the Riccati equation Eq. (3.111).
A necessary condition for the solution of the algebraic Riccati equation to be a positive
semidefinite symmetric matrix is that H has no imaginary eigenvalues [132].

3.4.6 Simulation Tests

The performance of the proposed flatness-based adaptive fuzzy MIMO controller was
tested in the nonlinear dynamic model of the 2-DOF rigid-link robotic manipulator
(Fig. 3.7). The control loop is depicted in Fig. 3.8. The differentially flat model of the
robot and its transformation to the Brunovksy form has been analyzed in Sect. 3.4.
The flat outputs were taken to be the robot’s joint angles y1 = x1 and y2 = x3 . It has
been proven that all state variables of the robotic model and the associated control
inputs, i.e., the torques applied by the motors to the links’ joints can be written as
functions of the flat output [y1 , y2 ] and of the associated derivatives.
The state feedback gain was K ∈ R 2×4 . The basis functions used in the estimation
x̂−c j 2
of f i (x, t), i = 1, 2 and gi j (x, t), i = 1, 2, j = 1, 2 were μ A j (x̂) = e( σ ) , j =
1, . . . , 3. Since there are four inputs x1 , ẋ1 and x2 , ẋ2 and the associated definition
sets (universe of discourse) consist of 3 fuzzy sets, for the approximation of functions
f i (x, t) i = 1, 2, there will be 81 fuzzy rules of the form:

R l : IF x1 is Al1 AND ẋ1 is Al2

(3.136)
AND x3 is Al3 AND ẋ3 is Al4 THEN fˆil is bl

81 4
ˆl i=1
l=1 f i μlA (xi ) (l)
and fˆi (x, t) =
81 4 i
. The centers ci , i = 1, . . . , 4 and the variances
l=1 i=1 μ Ai (xi )
l

v(l) of each rule are as follows:

Fig. 3.7 A 2-DOF rigid-link Y

robotic manipulator

θ2

θ1 + θ2

θ1
X
O
134 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

Fig. 3.8 Flatness-based adaptive-fuzzy control loop for the robotic manipulator

In the considered fuzzy rule-base there are four input parameters in the antecedent
parts of the fuzzy rules, i.e., x1 = θ1 , x2 = θ̇1 , x3 = θ2 and x4 = θ̇2 . Each parameter
is partitioned into 3 fuzzy sets. Therefore, as mentioned before, by taking all possible
combinations between the fuzzy sets one has 34 = 81 fuzzy rules. The finer the par-
tition of the input variables into fuzzy sets is, the more accurate the approximation of
the nonlinear system dynamics by the neuro-fuzzy model is expected to be (although
some of the rules of the fuzzy rule base may not be sufficiently activated due to
little coverage of the associated region of the patterns space by input data). However,
considering a large number of fuzzy sets for each input variable induces the curse
of dimensionality, which means that there is an excessive and rather unnecessary
increase in the number of the adaptable parameters that constitute the neuro-fuzzy
model (Table 3.1).
The estimation of the control input gain functions ĝi j (x, t) i = 1, 2 was derived in
a similar way. The time step was taken to be 0.01 s. In the beginning of the training of
the neuro-fuzzy approximators their weights were initialized to zero. Moreover, the
elements of the robot’s state vector were also initialized to zero. The positive definite
matrix P ∈ R 4×4 stems from the solution of the algebraic Riccati equation (3.111),
for Q also positive definite.
Two different setpoints were studied: (i) a sinusoidal signal of variable amplitude
and frequency, (ii) a seesaw setpoint of amplitude 0.30 and period T = 20 s. The
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 135

Table 3.1 Parameters of the fuzzy rule base

Rule c1(l) c2(l) c3(l) c4(l) v(l)
R (1) −1.0 −1.0 −1.0 −0.5 3
R (2) −1.0 −1.0 −1.0 0.0 3
R (3) −1.0 −1.0 −1.0 −1.0 3
R (4) −1.0 −1.0 0.0 −0.5 3
R (5) −1.0 −1.0 0.0 0.0 3
R (6) −1.0 −1.0 0.0 0.5 3
··· ··· ··· ··· ··· ···
··· ··· ··· ··· ··· ···
R (81) 1.0 1.0 1.0 0.5 3

approximations fˆ and ĝ were used in the derivation of the control law, given by
Eq. (3.115). To show the disturbance rejection capability of the proposed adaptive
fuzzy controller, at the beginning of the second half of the simulation time additive
sinusoidal disturbances of amplitude A = 0.5 and period T = 10 s were applied to
the robot’s joints. The external disturbances which appear as additive torques to the
robot’s joints can be due to a force F exerted on the manipulator’s end-effector (e.g.,
due to contact with a surface) and which through the relation T = J T F (J stands for
the Jacobian matrix of the robot’s kinematic model) generates torques on the joints
(Fig. 3.9).
In the simulation results that follow the position and velocity setpoints are noted
as continuous line while the position and velocity signals of the robot’s joints are
denoted as dashed lines. The position variation for the first joint of the robotic

(a) 2.5 (b) 30

2
20
1.5
state variable x2 of the robot
state variable x1 of the robot

1
10
0.5

0 0

−0.5
−10
−1

−1.5
−20
−2

−2.5 −30
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

t (sec) t (sec)

Fig. 3.9 a Tracking of sinusoidal position setpoint by joint 1 of the robot. b Tracking of sinusoidal
velocity setpoint by joint 1 of the robot
136 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

(a) 2.5 (b) 30

2
20
1.5
state variable x3 of the robot

state variable x4 of the robot

1
10
0.5

0 0

−0.5
−10
−1

−1.5
−20
−2

−2.5 −30
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40

t (sec) t (sec)

Fig. 3.10 a Tracking of sinusoidal position setpoint by joint 2 of the robot. b Tracking of sinusoidal
velocity setpoint by joint 2 of the robot

(a) 300 (b) 300

200 200
control signal of joint 1

control signal of joint 2

100 100

0 0

−100 −100

−200 −200

−300 −300
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 3.11 a Tracking of a sinusoidal setpoint: Control input to joint 1. b Tracking of a sinusoidal
setpoint: Control input to joint 2

manipulator when tracking a sinusoidal setpoint is depicted in Fig. 3.9a, while the
velocity variation is shown in Fig. 3.9b. For the second joint of the 2-DOF robot the
tracking of the position setpoint is depicted in Fig. 3.10a while the tracking of the
velocity setpoint is shown in Fig. 3.10b. The control inputs (motor torques) applied
to the first and second joints of the robotic manipulator are shown in Fig. 3.11a, b,
respectively.
The performance of the proposed flatness-based adaptive fuzzy control is also
tested in the tracking of a seesaw setpoint. It is shown that after suitable tuning
of the feedback gain matrix K defined in Eq. (3.91) precise tracking of the seesaw
setpoint can be succeeded, with fast elimination of the tracking error. The controller
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 137

(a) 2.5 (b) 30

2
20
1.5
state variable x1 of the robot

state variable x2 of the robot

1
10
0.5

0 0

−0.5
−10
−1

−1.5
−20
−2

−2.5 −30
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 3.12 a Tracking of a seesaw position setpoint by joint 1 of the robot. b Tracking of a seesaw
velocity setpoint by joint 1 of the robot

(a) 2.5 (b) 30

2
20
1.5
state variable x3 of the robot

state variable x4 of the robot

1
10
0.5

0 0

−0.5
−10
−1

−1.5
−20
−2

−2.5 −30
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 3.13 a Tracking of a seesaw position setpoint by joint 2 of the robot. b Tracking of a seesaw
velocity setpoint by joint 2 of the robot

is also suitable for compensating the effects of the additive torques disturbance that
is applied to the robot’s joints.
The position variation of the first joint is demonstrated in Fig. 3.12a, while the
velocity variation is shown in Fig. 3.12b. Similarly, the tracking of the position
reference setpoint for the second joint is depicted in Fig. 3.13a, while the tracking of
the associated velocity setpoint is given in Fig. 3.13b. The control signal in the case
of tracking a seesaw setpoint by the two joints of the robotic manipulator is shown
in Fig. 3.14a, b, respectively.
138 3 Nonlinear Adaptive Control Based on Differential Flatness Theory

(a) 300 (b) 300

200 200

control signal of joint 2

control signal of joint 1

100 100

0 0

−100 −100

−200 −200

−300 −300
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 3.14 a Tracking of a seesaw setpoint: Control input to joint 1. b Tracking of a seesaw setpoint:
Control input to joint 2

(a) 1.4 (b) 1.4

1.2 1.2
real vs estimated function g2b

real vs estimated function g2b

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 3.15 a Approximation of function g22 (x, t) when tracking a sinusoidal setpoint. b Approxi-
mation of function g22 (x, t) when tracking a seesaw setpoint

Finally, the approximation of function g22 (x, t) in the case of tracking of a sinu-
soidal setpoint is shown in Fig. 3.15a (and is marked as a dashed line), while when
tracking a seesaw setpoint the approximated function g22 (x, t) is shown in Fig. 3.15b.
The RMSE (root mean square error) of the examined control loop is also calculated
(assuming the same parameters of the controller) in the case of tracking of previous
setpoints (a) sinusoidal setpoint of variable amplitude/frequency and (b) seesaw
setpoint. The results are summarized in Table 3.2. It can be observed that the examined
controller, apart from elimination of tracking error for the parameters of the robot’s
state vector, succeeds also in good transient performance. Taking into account that
no prior knowledge about the robot’s dynamics has been used by the controller, and
3.4 Flatness-Based Adaptive Fuzzy Control for MIMO Systems 139

Table 3.2 RMSE of joints’ Parameter θ1 θ2

angles
RMSEa 0.0085 0.0028
RMSEb 0.0175 0.0145

at the first iterations of the control loop the controller undergoes self-tuning (learning
of values for specific gains) the transient characteristics of the control scheme can
be also deemed as quite satisfactory.
Chapter 4
Nonlinear Kalman Filtering Based
on Differential Flatness Theory

4.1 Introduction

A first type of uncertainties that flatness-based control has to cope with is the one
associated with parametric changes and external perturbations. This was analyzed
in Chap. 3, through flatness-based adaptive fuzzy control. A second type of uncer-
tainties in dynamical systems modeling and control has to do with the inability to
measure specific elements of the system’s state vector. To this end, the present chapter
introduces nonlinear Kalman Filtering based on differential flatness theory.
The chapter analyzes first the Derivative-free nonlinear Kalman Filter for SISO
and MIMO nonlinear dynamical systems. The considered nonlinear filtering scheme
which is based on differential flatness theory extends the class of systems to which
Kalman Filtering can be applied without the need for calculation of Jacobian matri-
ces. As explained, nonlinear systems satisfying the differential flatness property can
be written in the Brunovsky form via a transformation of their state variables and
control inputs. The Derivative-free nonlinear Kalman Filter consists of the Kalman
Filter recursion applied on the linearized equivalent of the system together with an
inverse transformation based on differential flatness theory that enables to retrieve
estimates for the state variables of the initial nonlinear model. In terms of accuracy
of the provided state estimation, the performance of the new filter is equivalent to
the one of the Unscented Kalman Filter and significantly improved comparing to
the Extended Kalman Filter. In terms of speed of computation, the Derivative-free
nonlinear Kalman Filter outperforms all other nonlinear estimation algorithms.
Aiming also at finding more efficient implementations of distributed nonlinear
filtering, in this chapter a derivative-free approach to Extended Information Filtering
is introduced and applied to state estimation-based control of nonlinear dynamical
systems, such as robots. First, the system undergoes a linearization transformation
and next state estimation is performed by applying the Kalman Filter recursion to the
linearized model. At a second level, the standard Information Filter is used to fuse the
state estimates obtained from local derivative-free Kalman filters running at the local

© Springer International Publishing Switzerland 2015 141

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_4
142 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

information processing nodes. This approach has significant advantages because

unlike the Extended Information Filter (i) it is not based on local linearization of the
nonlinear dynamics (ii) it does not assume truncation of higher order Taylor expan-
sion terms thus preserving the accuracy and robustness of the performed estimation,
(iii) it does not require the computation of Jacobian matrices. The Derivative-free
Extended Information Filter appears to be faster than the previously mentioned non-
linear filtering methods (i.e., Extended Information Filter, Unscented Information
Filter, and Distributed Particle Filter) while also providing very accurate (in terms
of variance) state estimates.

4.2 The Derivative-Free Nonlinear Kalman Filter

4.2.1 Overview

State estimation of nonlinear dynamical systems with the use of filters is a significant
topic in the area of control systems design, robotics, and mechatronics since it can
provide improved methods for sensorless control of electromechanical systems. For
nonlinear systems, and under the assumption of Gaussian noise, the Extended Kalman
Filter (EKF) is frequently applied for estimating the nonmeasurable state variables
through the processing of input and output sequences. This has been analyzed in [31,
51, 192, 229, 372, 417]. The Extended Kalman Filter is based on linearization of
the system using a first-order Taylor expansion [405, 408, 412, 422, 573]. Although
EKF is efficient in several estimation problems, it is characterized by cumulative
errors due to the local linearization assumption and this may affect the accuracy of
the state estimation or even risk the stability of the observer-based control loop.
Aiming at finding more efficient implementations of nonlinear Kalman Filtering,
in this chapter a derivative-free approach to Kalman filtering is introduced and applied
to state estimation-based control of a wide class of nonlinear electromechanical
systems. In the proposed derivative-free Kalman Filtering method the system is
first subject to a linearization transformation that is based on differential flatness
theory and next state estimation is performed by applying the standard Kalman
Filter recursion to the linearized model. Unlike EKF, the proposed method provides
estimates of the state vector of the nonlinear system without the need for derivatives
and Jacobian matrices calculation. By avoiding linearization approximations, the
proposed filtering method improves the accuracy of estimation of the system state
variables, and results in smooth control signal variations and in minimization of the
tracking error of the associated control loop. The chapter extends the results of [424,
429, 434] toward single-input single-output (SISO) and multi-input multi-output
(MIMO) nonlinear dynamical systems.
4.2 The Derivative-Free Nonlinear Kalman Filter 143

4.2.2 Extended Kalman Filtering for Nonlinear

Dynamical Systems

4.2.2.1 The Continuous-Time Kalman Filter for Linear Systems

First, the continuous-time dynamical system of Eq. (4.1) is assumed [408]:


ẋ(t) = Ax(t) + Bu(t) + w(t), t ≥ t0
(4.1)
z(t) = C x(t) + v(t), t ≥ t0

where x ∈ R m×1 is the system’s state vector, and z ∈ R p×1 is the system’s output.
Matrices A, B, and C can be time-varying and w(t), v(t) are uncorrelated white
Gaussian noises. The covariance matrix of the process noise w(t) is Q(t), while the
covariance matrix of the measurement noise is R(t). Then, the Kalman Filter is a
linear state observer which is given by
⎧
⎪ ˙
⎨x̂ = A x̂ + Bu + K [z − C x̂], x̂(t0 ) = 0
K (t) = PC T R −1 (4.2)
⎪
⎩
Ṗ = A P + P A T + Q − PC T R −1 CP

where x̂(t) is the optimal estimation of the state vector x(t) and P(t) is the covariance
matrix of the state vector estimation error with P(t0 ) = P0 . The Kalman Filter
consists of the system’s state equation plus a corrective term K [z−C x̂]. The selection
of gain K corresponds actually to the solution of an optimization problem. This is
expressed as the minimization of a quadratic cost functional and is performed through
the solution of a Riccati equation. In that case the observer’s gain K is calculated
by K = PC T R −1 considering an optimal control problem for the dual system
(A T , C T ), where the covariance matrix of the estimation error P is found by the
solution of a continuous-time Riccati equation of the form

Ṗ = AP + P A T + Q − PC T R −1 CP (4.3)

where matrices Q and R stand for the process and measurement noise covariance
matrices, respectively.
144 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

4.2.2.2 The Discrete-Time Kalman Filter for Linear

Dynamical Systems

In the discrete-time case, a dynamical system is assumed to be expressed in the form

of a discrete-time state model [408, 422]:

x(k + 1) = A(k)x(k) + L(k)u(k) + w(k)

(4.4)
z(k) = Cx(k) + v(k)

where the state x(k) is a m-vector, w(k) is a m-element process noise vector, and A
is a m × m real matrix. Moreover, the output measurement z(k) is a p-vector, C is
a p × m-matrix of real numbers, and v(k) is the measurement noise. It is assumed
that the process noise w(k) and the measurement noise v(k) are uncorrelated.
Now the problem of interest is to estimate the state x(k) based on the sequence
of output measurements z(1), z(2), . . . , z(k). The initial value of the state vector
x(0), and the initial value of the error covariance matrix P(0) is unknown and an
estimation of it is considered, i.e., x̂(0) is a guess of E[x(0)] and P̂(0) is a guess of
Cov[x(0)].
For the initialization of matrix P one can set P̂(0) = λI , with λ > 0. The
state vector x(k) has to be estimated taking into account x̂(0), P̂(0), and the output
measurements Z = [z(1), z(2), . . . , z(k)]T , i.e., x̂(k) = αn (x̂(0)), P̂(0), Z (k)).
This is a linear minimum mean squares estimation problem (LMMSE) formulated
as x̂(k + 1) = an+1 (x̂(k), z(k + 1)). The process and output noise are white and their
covariance matrices are given by: E[w(i)wT ( j)] = Qδ(i − j) and E[v(i)vT ( j)] =
Rδ(i − j).
Using the above, the discrete-time Kalman filter can be decomposed into two parts:
(i) time update (prediction stage), and (ii) measurement update (correction stage).
The first part employs an estimate of the state vector x(k) made before the output
measurement z(k) is available (a priori estimate). The second part estimates x(k)
after z(k) has become available (a posteriori estimate). The following measurement
processing stages are distinguished:
• When the set of measurements Z − = {z(1), . . . , z(k − 1)} is available. From
Z − an a priori estimation of x(k) is obtained which is denoted by x̂ − (k) and this
denotes the estimate of x(k) given Z − .
• When z(k) is available, the output measurements set becomes Z = {z(1), . . . , z(k)},
where x̂(k) and this denotes the estimate of x(k) given Z .
The associated estimation errors are defined by e− (k) = x(k) − x̂ − (k)= the a priori
error, and e(k) = x(k)− x̂(k)= the a posteriori error. The estimation error covariance
matrices associated with x̂ − (k) and x̂(k) are defined as P − (k) = Cov[e− (k)] =
E[e− (k)e− (k) ] and P(k) = Cov[e(k)] = E[e(k)e T (k)], respectively [229]. From
T

the definition of the trace of a matrix, the mean square error of the estimates can
be written as MSE(x̂ − (k)) = E[e− (k)e− (k) ] = tr(P − (k)) and MSE(x(k)) =
T

E[e(k)e T (k) = tr (P(k)).

4.2 The Derivative-Free Nonlinear Kalman Filter 145

Finally, the linear Kalman filter equations are:

measurement update:

K (k) = P − (k)C T [C·P − (k)C T + R]−1

x̂(k) = x̂ − (k) + K (k)[z(k) − C x̂ − (k)] (4.5)
P(k) = P − (k) − K (k)C P − (k)

time update:

P − (k + 1) = A(k)P(k)A T (k) + Q(k)

(4.6)
x̂ − (k + 1) = A(k)x̂(k) + L(k)u(k)

4.2.2.3 The Extended Kalman Filter

State estimation can be also performed for nonlinear dynamical systems using the
Extended Kalman Filter recursion. The following nonlinear state model is considered
[408, 422]:
x(k + 1) = φ(x(k)) + L(k)u(k) + w(k)
(4.7)
z(k) = γ (x(k)) + v(k)

where x ∈ R m×1 is the system’s state vector and z ∈ R p×1 is the system’s out-
put, while w(k) and v(k) are uncorrelated, Gaussian zero mean noise processes with
covariance matrices Q(k) and R(k), respectively. The operators φ(x) and γ (x) are the
vectors φ(x) = [φ1 (x), φ2 (x), . . ., φm (x)]T , and γ (x) = [γ1 (x), γ2 (x), . . . , γ p (x)]T ,
respectively. It is assumed that φ and γ are sufficiently smooth in x so that each
one has a valid series Taylor expansion. Following a linearization procedure, φ is
expanded into Taylor series about x̂:

φ(x(k)) = φ(x̂(k)) + Jφ (x̂(k))[x(k) − x̂(k)] + · · · (4.8)

where Jφ (x) is the Jacobian of φ calculated at x̂(k):

⎛ ∂φ1 ∂φ1 ∂φ1 ⎞
∂ x1 ∂ x2 ··· ∂ xm
⎜ ∂φ ∂φ ⎟
⎜ 2 2
··· ∂φ2 ⎟
∂φ ⎜ ∂ x1 ∂ x2 ∂ xm⎟
Jφ (x) = |x=x̂(k) = ⎜
⎜ .. .. ..
⎟
.. ⎟ (4.9)
∂x ⎜ .
⎝ . . . ⎟
⎠
∂φm ∂φm ∂φm
∂ x1 ∂ x2 ··· ∂ xm

Likewise, γ is expanded about x̂ − (k)

146 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

γ (x(k)) = γ (x̂ − (k)) + Jγ [x(k) − x̂ − (k)] + · · · (4.10)

where x̂ − (k) is the estimation of the state vector x(k) before measurement at the
kth instant to be received and x̂(k) is the updated estimation of the state vector after
measurement at the kth instant has been received. The Jacobian Jγ (x) is
⎛ ∂γ1 ∂γ1 ∂γ1 ⎞
∂ x1 ∂ x2 ··· ∂ xm
⎜ ∂γ ∂γ ⎟
⎜ 2 2 ··· ∂γ2 ⎟
∂γ ⎜ ∂ x1 ∂ x2 ∂ xm ⎟
Jγ (x) = |x=x̂ − (k) = ⎜
⎜ .. .. ..
⎟
.. ⎟ (4.11)
∂x ⎜ .
⎝ . . . ⎟
⎠
∂γ p ∂γ p ∂γ p
∂ x1 ∂ x2 ··· ∂ xm

The resulting expressions create first-order approximations of φ and γ . Thus the

linearized version of the system is obtained:

x(k + 1) = φ(x̂(k)) + Jφ (x̂(k))[x(k) − x̂(k)] + w(k)

(4.12)
z(k) = γ (x̂ − (k)) + Jγ (x̂ − (k))[x(k) − x̂ − (k)] + v(k)

Now, the EKF recursion is as follows: First, the time update is considered: by x̂(k)
the estimation of the state vector at time instant k is denoted. Given initial conditions
x̂ − (0) and P − (0) the recursion proceeds as:
• Measurement update. Acquire z(k) and compute:

K (k) = P − (k)JγT (x̂ − (k))·[Jγ (x̂ − (k))P − (k)JγT (x̂ − (k)) + R(k)]−1
x̂(k) = x̂ − (k) + K (k)[z(k) − γ (x̂ − (k))] (4.13)
P(k) = P − (k) − K (k)Jγ (x̂ − (k))P − (k)

• Time update. Compute:

P − (k + 1) = Jφ (x̂(k))P(k)JφT (x̂(k)) + Q(k)

(4.14)
x̂ − (k + 1) = φ(x̂(k)) + L(k)u(k)

The schematic diagram of the EKF loop is given in Fig. 4.1.

Indicative applications of the Extended Kalman Filter in control of electromechan-
ical systems have been presented in [102] where the EKF has been used in state
estimation-based control of electrohydraulic actuators, in [480] where the EKF has
been used as a disturbance observer resulting in improved control of a direct cur-
rent motor, in [298] where the EKF has been used in sensorless control of nonlinear
brushless DC motor models, in [307] where the EKF has been used for estimating
the electromagnetic torque of direct-torque controlled brushless DC motors and for
implementing a sensorless control scheme, and in [514] where the EKF has been
applied for rotor resistance estimation of a DC motor, for position and velocity esti-
mation and finally for state estimation-based control. Examples on nonlinear filtering
4.2 The Derivative-Free Nonlinear Kalman Filter 147

Fig. 4.1 Schematic diagram of the EKF loop

for electric machines tasks have been given in [138] where an EKF-based fault diag-
nosis scheme for actuators has been studied, in [578] where the EKF has been pro-
posed for residual generation and diagnosis of incipient faults in DC electric motors,
and in [13] where the EKF has been used in fault diagnosis of an electrohydraulic
actuation system.

4.2.2.4 Unscented Kalman Filter

Next, the functioning of the Unscented Kalman Filter (UKF) will be explained.
The UKF belongs to a wider class of nonlinear estimators known as Sigma-Point
Kalman Filters (SPKF) [224, 225, 231]. Unlike EKF, in the Unscented Kalman Filter
no analytical Jacobians of the system equations need to be calculated. Some basic
operations performed in the UKF algorithm are summarized as follows:

(1) Denoting the current state mean as x̂, √ a set of 2n + 1 sigma points is taken
from√the columns of the n × n matrix (n + √ λ)Pxx as follows: x = x̂, x =
0 i

x̂ +[ (n + λ)Pxx ]i , i = 1, . . . , n, x = x̂ −[ (n + λ)Pxx ]i , i = n+1, . . . , 2n,

i

and the associate weights are computed as:

(m) λ (c) λ
W0 = (n+λ) W0 = (n+λ)+(1−α 2 +b)
(4.15)
(m) (c)
Wi = 1
2(n+λ) Wi = 2(n+λ)1
148 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

where i = 1, 2, . . . , 2n and λ = α 2 (n + κ) − n is a scaling parameter, while

α, β, and κ are constant parameters. Matrix Pxx is the covariance matrix of the
state x.
(2) Transform each of the sigma points as z i = h(x i ) i = 0, . . . , 2n.
(3) Mean and covariance estimates for z can be computed as
2n (m) i
ẑ  i=0 Wi z
2n (c) i
(4.16)
Pzz = i=0 Wi (z − ẑ)(z
i − ẑ)T

(4) The cross-covariance of x and z is estimated as

2n (c) i
Pxz = i=0 Wi (x − x̂)(z i − ẑ)T (4.17)
√
The matrix square root of positive definite matrix Pxx means a matrix A = Pxx such
that Pxx = A A T and a possible way for calculation is Singular Value Decomposition
(SVD). As in the case of the Extended Kalman Filter, the Unscented Kalman Filter
also consists of prediction stage (time update) and correction stage (measurement
update) [225, 469].
Time update: Compute the predicted state mean x̂ − (k) and the predicted covariance
Pxx − (k), using the Unscented Transform (UT):

[x̂ − (k), Pxx

− (k)] = UT( f , x̂(k − 1), P (k − 1))
d xx
− (4.18)
Pxx (k) = Pxx (k − 1) + Q(k − 1)

Measurement update: Obtain the new output measurement z(k) and compute the pre-
dicted mean ẑ(k) and covariance of the measurement Pzz (k), and the cross-covariance
of the state and measurement Pxz (k)

[ẑ(k), Pzz (k), Pxz (k)] = U T (h d , x̂ − (k), Pxx

− (k))
(4.19)
Pzz (k) = Pzz (k) + R(k)

Then compute the filter gain K (k), the state mean x̂(k), and the covariance Pxx (k),
conditional to the measurement y(k)

K (k) = Px z (k)Pzz−1 (k)

x̂(k) = x̂ − (k) + K (k)[z(k) − ẑ(k)] (4.20)
Pxx (k) = − (k) −
Pxx K (k)Pzz (k)K (k)T

The filter starts from the initial mean m(0) and covariance Px x (0). The stages of
state vector estimation with the use of the Unscented Kalman Filter algorithm are
depicted in Fig. 4.2.
4.2 The Derivative-Free Nonlinear Kalman Filter 149

Fig. 4.2 Schematic diagram

of the Unscented Kalman
Filter loop

4.2.3 Derivative-Free Kalman Filtering to SISO

Nonlinear Systems

4.2.3.1 Conditions for Derivative-Free Kalman Filtering in SISO

Nonlinear Systems

Differential flatness theory will be used for arriving at a filtering method that will have
improved performance compared to the Extended Kalman Filter. It will be shown
that through a nonlinear transformation it is possible to design a state estimator for
a class of nonlinear systems, which can substitute for the Extended Kalman Filter.
The results will be towards derivative-free Kalman Filtering for nonlinear systems.
The following continuous-time nonlinear single-output system is considered [334,
335, 434].
p
ẋ = f (x) + q0 (x, u) + i=1 θi qi (x, u), or
ẋ = f (x) + q0 (x, u) + Q(x, u)θ x ∈ R n , u ∈ R m , θ ∈ R p (4.21)
z = h(x), z ∈ R

with qi : R n × R m → R n , 0 ≤ i ≤ p, f : R n →R n , h: R n →R, smooth functions,

h(x0 ) = 0, q0 (x, 0) = 0 for every x ∈ R n ; x is the state vector, u(x, t): R + →R m
is the control which is assumed to be known, θ is the parameter vector which is
supposed to be constant, and y is the scalar output.
The first main assumption on the class of systems considered is the linear depen-
dence on the parameter vector θ . The second main assumption requires that systems
of Eq. (4.21) are transformable by a parameter independent state-space change of
coordinates in R n
150 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

ζ = T (x), T (x0 ) = 0 (4.22)

into the system p

ζ̇ = Ac ζ + ψ0 (z, u) + i=1 θi ψi (z, u)⇒
ζ̇ = Ac ζ + ψ0 (z, u) + Ψ (z, u)θ (4.23)
z = Cc ζ

with ⎛ ⎞
0 1 0 ··· 0
⎜0 0 1 ··· 0⎟
⎜ ⎟
Ac = ⎜ . .. .. .. .. ⎟ (4.24)
⎝ .. . . . .⎠
0 0 0 ··· 0
 
Cc = 1 0 0 · · · 0 (4.25)

and ψi : R × R m →R n smooth functions for i = 0, . . . , p. The necessary and suffi-

cient conditions for the initial nonlinear system to be transformable into the form of
Eq. (4.23) have been given in [334, 335, 434] and are summarized in the following:
(i) rank{dh(x), d L f h(x), . . . , d Ln−1 h(x)} = n, ∀x ∈ R n (which implies local
f
observability). It is noted that L f h(x) stands for the Lie derivative L f h(x) =
(∇h) f and the repeated Lie derivatives are recursively defined as L0f h =
h for i = 0, Lif h = L f Li−1 i−1
f h = ∇ L f h f for i = 1, 2, . . ..
j
(ii) [adif g, ad f g] = 0, 0 ≤ i, j ≤ n − 1. It is noted that adif g stands for a Lie
Bracket which is defined recursively as adif g = [ f, adi−1
f g] with ad f g = g
0

and ad f g = [ f, g] = ∇g f − ∇ f g.
(iii) [qi , adif g] = 0, 0 ≤ i ≤ p, 0 ≤ j ≤ n − 2 ∀ u ∈ R m .
(iv) the vector fields adif g, 0 ≤ i ≤ n − 1 are complete, in which g is the vector
field satisfying ⎛ ⎞ ⎛ ⎞
dh 0
⎜ .. ⎟ ⎜ .. ⎟
<⎝ . ⎠ , g >= ⎝ . ⎠ (4.26)
d(Ln−1
f h) 1

∂h
where dh has been defined as dh = ∂x = [ ∂∂h ∂h ∂h
x1 , ∂ x2 , . . . , ∂ xn ]. Then for every
parameter vector θ , the system
p
ζ̂ = Ac ζ̂ + ψ0 (z, u) + i=1 θi ψi (z, u) + K (z − Cc ζ̂ )
(4.27)
x̂ = T −1 (ζ̂ )
4.2 The Derivative-Free Nonlinear Kalman Filter 151

is an asymptotic observer for a suitable choice of K provided that the state x(t) is
bounded, with estimation error dynamics
⎛ ⎞
−k1 1 0 ··· 0
⎜ −k2 0 1 ··· 0⎟
⎜ ⎟
ė = (Ac − K Cc )e = ⎜
⎜ ··· ··· ··· ··· · · ·⎟
⎟e (4.28)
⎝−kn−1 0 0 ··· 1⎠
−kn 0 0 ··· 0

The eigenvalues of Ac − K Cc can be arbitrarily placed by choosing the vector K ,

since they coincide with the roots of the polynomial s n + k1 s n−1 + · · · + kn .
From, Eq. (4.27) it can be noticed that to obtain estimates of the state vector of
the initial nonlinear system, knowledge of the inverse transformation T −1 (diffeo-
morphism) is needed. At a later part of the book (Chap. 7) it will be shown that the
inverse transformation needs the inverse of the Jacobian matrix J that is computed
for the transformed state vector of the system.

4.2.3.2 State Estimation for SISO Systems with the Derivative-Free

Nonlinear Kalman Filter

Since Eq. (4.27) provides an asymptotic observer for the initial nonlinear system of
Eq. (4.21) one can consider a case in which the observation error gain matrix K can
be provided by the Kalman Filter equations given initially in the continuous-time
KF formulation, or in discrete-time form by Eqs. (10.62) and (10.63). The following
single-input single-output nonlinear dynamical system is considered

x (n) = f (x, t) + g(x, t)u(x, t) (4.29)

where z = x is the system’s output, and f (x, t), g(x, t) are nonlinear functions. It
can be noticed that the system of Eq. (4.29) belongs to the general class of systems of
Eq. (4.21). Moreover, Eq. (4.29) is the canonical form one obtains after the application
of a differential flatness theory-based transformation. Assuming the transformation
ζi = x (i−1) , i = 1, . . . , n, and x (n) = f (x, t) + g(x, t)u(x, t) = v(ζ, t), i.e.,
ζ̇n = v(ζ, t), one obtains the linearized system of the form

ζ̇1 = ζ2
ζ̇2 = ζ3
··· ··· (4.30)
ζ̇n−1 = ζn
ζ̇n = v(ζ, t)
152 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

which in turn can be written in state-space equations of the canonical form, as

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ζ̇1 0 1 0 ··· 0 ζ1 0
⎜ ζ̇2 ⎟ ⎜ 0 0 1 ··· 0⎟ ⎜ ζ2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ · · · ⎟ = ⎜· · · ··· ··· ··· · · ·⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ · · · ⎟ + ⎜· · ·⎟ v(ζ, t) (4.31)
⎝ζ̇n−1 ⎠ ⎝ 0 0 0 ··· 1 ⎠ ⎝ζn−1 ⎠ ⎝ 0 ⎠
ζ̇n 0 0 0 ··· 0 ζn 1
 
z = 1 0 0 ··· 0 ζ (4.32)

The system of Eqs. (4.31) and (4.32) has been written in the form of Eq. (4.23),
which means that Eq. (4.27) is the associated asymptotic observer. Therefore, the
observation gain K appearing in Eq. (4.27) can be found using either linear observer
design methods (in that case the elements of the observation error gain matrix K
have fixed values), or the recursive calculation of the continuous-time Kalman Filter
gain described in Sect. 4.2.2.2. If the discrete-time Kalman Filter is to be used then
one has to apply the recursive formulas of Eqs. (4.5) and (4.6) on the discrete-time
equivalent of Eqs. (4.31) and (4.32).

4.2.4 Simulation Tests

4.2.4.1 DC Motor Model

The model of the linear DC motor shown in Fig. 4.3 is described by the set of equations

L I˙ = −ke ω − R I + V
(4.33)
J ω̇ = ke I − kd ω − Γd

with the following notations L: armature inductance, I : armature current, ke : motor

electrical constant, R: armature resistance, V : input voltage, taken as control input,
J : motor inertia, ω: rotor rotation speed, kd : mechanical dumping constant, Γd :
disturbance torque. Assuming Γ˙d = 0 and denoting the state vector as [x1 , x2 , x3 ]T =
[θ, θ̇ , θ̈ ]T , a linear model of the DC motor is obtained:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẋ1 0 1 0 x1 0
⎝ẋ2 ⎠ = ⎝0 0 1 ⎠ ⎝x2 ⎠ + ⎝ 0 ⎠ V (4.34)
ẋ3 −ke2 −kd R −J R−kd L x3 ke
0 JL JL JL

Next, control for a nonlinear DC motor model will be presented. Usually, the DC
motor model is considered to be linear by neglecting the effect of armature reaction
or by assuming that the compensating windings remove this effect. Introducing the
armature reaction leads to a nonlinear system. In that case the dynamic model of the
4.2 The Derivative-Free Nonlinear Kalman Filter 153

Fig. 4.3 Parameters of the DC motor model

DC motor model can be written as an affine in-the-input system [202]:

ẋ = f (x) + g(x)u (4.35)

with ẋ denoting the derivative of the motor’s state vector, x = [x1 , x2 , x3 ]T =

[θ, θ̇ , i α ] (θ is the position of the motor, θ̇ is the angular velocity of the motor,
and i α is the armature current). Functions f (x) and g(x) are vector field functions
defined as:
⎛ ⎞ ⎛ ⎞
x2 0
f (x) = ⎝k1 x2 + k2 x3 + k3 x32 + k4 T1 ⎠ , g(x) = ⎝ 0 ⎠ (4.36)
k5 x 2 + k6 x 2 x 3 + k7 x 3 k8

where k1 = −F/J , k2 = A/J , k3 = B/J , k4 = −1/J , k5 = −A/L, k6 = −B/L,

k7 = −R/L, k8 = −1/L, R, and L are the armature resistance and induction,
respectively, and J is the rotor’s inertia, while F is the friction. Now, considering
k4 T1 as disturbance, the state-space equation of the DC motor can be rewritten as

ẋ1 = x2
ẋ2 = k1 x2 + k2 x3 + k3 x32 + k4 T1 (4.37)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u

where T1 the load torque and u is the terminal voltage. From the second row of
Eq. (4.37) one obtains,

ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 x3 ẋ3 ⇒

(4.38)
ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 k5 x2 x3 + 2k3 k6 x2 x32 + 2k3 k7 x32 + 2k3 k8 x3 u
154 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

Thus the input–output relation can be written as

ẍ2 = f¯(x) + ḡ(x)u (4.39)

where, f¯(x) = k1 ẋ2 + k2 ẋ3 + 2k3 k5 x2 x3 + 2k3 k6 x2 x32 + 2k3 k7 x32 , and ḡ(x) =
2k3 k8 x3 . Taking k1 = 0 and considering the torque T1 as external disturbance, the
nonlinear DC motor model of Eq. (4.37) becomes

ẋ1 = x2
ẋ2 = k2 x3 + k3 x32 (4.40)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u

Then, setting the output to be y = x1 one can see that all state variables xi , i = 1, 2, 3
and the control input u can be expressed as functions of this output and its derivatives.
Indeed it holds
x1 = y
x2 = ẏ

−k2 + |k22 +4k3 ÿ| (4.41)
x3 = 2k3
u = ḡ(x)
1
[y (3) − f¯(x)]

The aforementioned system of Eq. (4.40) can be written in the Brunovsky (canonical)
form: ⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (4.42)
ẏ3 000 y3 1

where v = f¯(x, t) + ḡ(x, t)u. Denoting the state vector of the transformed descrip-
tion of the system as y = [y1 , y2 , y)3]T , the measurement (observation) equation
becomes
z 1 = y1 ⇒
(4.43)
z 1 = [1 0 0]y

(r )
Finally, using a control input of the form u = ḡ(x̂, t)−1 [− f¯(x̂, t)+ ym + K cT e +u c ]
it is possible to make the electric motor’s angle track any desirable setpoint.

4.2.4.2 Evaluation Experiments

The dynamic model of the DC motor given in Eq. (4.39) is considered. The control
law described above is used to make the motor track the desirable trajectory. It is
assumed that only encoder measurements are available (angle θ ), thus the state vector
x used in the control law is not directly measured and has to be estimated through
filtering. First, state estimation-based control is implemented with the use of the
4.2 The Derivative-Free Nonlinear Kalman Filter 155

Extended Kalman Filter. The associated covariance matrices Q and R were taken to
be diagonal, both having diagonal elements equal to 10−3 . The feedback controller’s
gain appearing was taken to be K T = [60, 30, 20]. The initial value of the system’s
state vector was x(0) = [0, 0, 0]T while the initial value of the estimated state vector
was x̂(0) = [1, 1, 1]T . The estimation error covariance matrix P ∈ 3×3 and the
EKF gain K ∈ 3×1 were used in Eqs. (4.5) and (4.6). The sampling period was
taken to be Ts = 0.01 s. The obtained results for the case of a seesaw and a sinusoidal
setpoint are shown in Figs. 4.4, 4.5, 4.6 and 4.7. The reference setpoint in denoted
by the red line, the state vector variables of the motor are denoted by the dashed blue
line, while the estimated state vector elements are described by the green line.
Next, state estimation-based control was implemented with the use of the
Unscented Kalman Filter. As explained the UKF does not require linearization of
the system’s dynamical model and does not incorporate the calculation of Jacobians.
Since linearization errors are avoided, the Unscented Kalman Filter results in more
robust state estimation thus also making more accurate the tracking of the reference
setpoints by the associated control loop. It can be also noticed that the variations of
the control signal are quite smooth. The associated simulation results are depicted
in Figs. 4.8, 4.9, 4.10 and 4.11.
Finally, in the case of state estimation-based control with the use of the derivative-
free nonlinear Kalman Filter (DKF), the estimation of the state vector x̂ is obtained
by applying derivative-free Kalman Filtering to the linearized model of Eqs. (4.44)
and (4.45), according to Eqs. (4.5) and (4.6). Starting from Eq. (4.39) and defining
ζ1 = θ , ζ2 = θ̇ , and ζ3 = θ̈ the state-space equations of the motor are written as:

(a) 1.5
EKF − x1 (b) 1.5
EKF − x1

1 1
position x of the motor (rad)

position x 1 of the motor (rad)

0.5 0.5

0 0
1

−0.5 −0.5

−1 −1

−1.5 −1.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.4 Variable x1 of the state vector in state estimation with use of the Extended Kalman Filter,
a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint
156 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 3
EKF − x2 (b) 3
EKF − x2

velocity x of the motor (rad/sec)

velocity x2 of the motor (rad/sec)

2 2

1 1

0 0

2
−1 −1

−2 −2

−3 −3
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.5 Variable x2 of the state vector in state estimation with use of the Extended Kalman Filter,
a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint

(a) 6
EKF − x3 (b) 6
EKF − x3
acceleration x of the motor (rad/sec2)
acceleration x of the motor (rad/sec2)

4 4

2 2

0 0

−2 −2
3

−4 −4

−6 −6

−8 −8
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.6 Variable x3 of the state vector in estimation performed with use of the Extended Kalman
Filter, a when tracking a seesaw setpoint, b when tracking of a sinusoidal setpoint

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ζ̇1 010 ζ1 0
⎝ζ̇2 ⎠ = ⎝0 0 1⎠ ⎝ζ2 ⎠ + ⎝0⎠ v(ζ, t) (4.44)
ζ̇3 000 ζ3 1
 
z= 100 ζ (4.45)

where v(ζ, t) = f¯(x, t) + ḡ(x, t)u(x, t). To perform state estimation-based (sensor-
less) control only measurements of the rotor’s angle θ were used. As in the EKF case
the associated covariance matrices Q and R were taken to be diagonal, both having
4.2 The Derivative-Free Nonlinear Kalman Filter 157

(a)
control signal Vin of the motor (v.u.)
400
EKF − control input (b) 400 EKF − control input

control signal Vin of the motor (v.u.)

300 300

200 200

100 100

0 0

−100 −100

−200 −200

−300 −300

−400 −400
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.7 Control signal of the Extended Kalman Filter-based control loop, a when tracking a seesaw
setpoint, b when tracking a sinusoidal setpoint

(a) 1.5
UKF − x1 (b) 1.5
UKF − x1
position x1 of the motor (rad)

1 1
position x of the motor (rad)

0.5 0.5

0 0
1

−0.5 −0.5

−1 −1

−1.5 −1.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.8 Variable x1 of the state vector in state estimation with use of the Unscented Kalman Filter,
a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint

diagonal elements equal to 10−3 . The tracking performance of the derivative-free

nonlinear Kalman Filter-based control loop, in the case of a seesaw and a sinusoidal
setpoint are depicted in Figs. 4.12, 4.13, 4.14 and 4.15.
Moreover, results on the accuracy of the performed estimation are presented in
Tables 4.1 and 4.2. Table 4.1 shows the accuracy of the setpoint tracking in state
estimation-based control implemented with the use of the considered nonlinear filters.
Table 4.2 shows the accuracy of estimation when the nonlinear filters were used only
for reconstructing the motor’s state vector out of output measurements, while control
was implemented with feedback of the complete state vector. The measure used for
evaluating the accuracy of the estimation performed by the nonlinear filters, as well
158 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a)
velocity x2 of the motor (rad/sec)
3
UKF − x2 (b) 3
UKF − x2

velocity x2 of the motor (rad/sec)

2 2

1 1

0 0

−1 −1

−2 −2

−3 −3
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.9 Variable x2 of the state vector in state estimation with use of the Unscented Kalman Filter,
a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint

(a) UKF − x3 (b) UKF − x3

6 6
acceleration x of the motor (rad/sec2)

acceleration x of the motor (rad/sec 2)

4 4

2 2

0 0

−2 −2
3

−4 −4

−6 −6

−8 −8
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.10 Variable x3 of the state vector in estimation was performed with use of the Unscented
Kalman Filter, a when tracking a seesaw setpoint, b when tracking of a sinusoidal setpoint

as the accuracy of tracking of the state estimation-based control loops was the Mean
Square Error (MSE). Alternatively, the Cramer–Rao Lower Bound (CRLB) could
have been considered [574].
It can be observed that comparing to EKF the derivative-free nonlinear Kalman
Filter results in smaller estimation errors for the system’s state variables and con-
sequently the accuracy of tracking of the reference setpoints is also improved. The
ranges of variation of the control signal are also kept small. Using the derivative-
free nonlinear Kalman Filter the performance of the state estimation-based control
is comparable to the one obtained when using the Unscented Kalman Filter for
4.2 The Derivative-Free Nonlinear Kalman Filter 159

(a) UKF − control input (b) UKF − control input

400 400
control signal Vin of the motor (v.u.)

control signal Vin of the motor (v.u.)

300 300

200 200

100 100

0 0

−100 −100

−200 −200

−300 −300

−400 −400
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.11 Control signal of the Unscented Kalman Filter-based control loop, a when tracking a
seesaw setpoint, b when tracking a sinusoidal setpoint

(a) 1.5
DKF − x1 (b) 1.5
DKF − x1

1 1
position x of the motor (rad)
position x of the motor (rad)

0.5 0.5

0 0
1
1

−0.5 −0.5

−1 −1

−1.5 −1.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.12 Variable x1 of the state vector in state estimation with use of the derivative-free Kalman
Filter, a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint

estimating the nonmeasurable state vector parameters. It can be also noted that in
terms of computation cost the proposed derivative-free nonlinear Kalman Filtering
(DKF) approach has some advantages comparing to the Unscented Kalman Filter:
since the DKF follows the fast recursion of the standard Kalman Filter it does not
need to update sigma points at each iteration, to propagate the nonlinear estimation
through these sigma points and to calculate cross-covariance matrices.
The performance of the nonlinear filtering schemes was also tested in state
estimation-based control for the tracking of a setpoint of variable amplitude, fre-
quency, and phase. The associated results for the Extended Kalman Filter have
160 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a)
velocity x 2 of the motor (rad/sec)
3
DKF − x2 (b) 3
DKF − x2

velocity x2 of the motor (rad/sec)

2 2

1 1

0 0

−1 −1

−2 −2

−3 −3
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.13 Variable x2 of the state vector in state estimation with use of the derivative-free Kalman
Filter, a when tracking a seesaw setpoint, b when tracking a sinusoidal setpoint

(a) 6
DKF − x3 (b) 6
DKF − x3
acceleration x of the motor (rad/sec 2)

acceleration x3 of the motor (rad/sec )

2

4 4

2 2

0 0

−2 −2
3

−4 −4

−6 −6

−8 −8
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.14 Variable x3 of the state vector in estimation was performed with use of the derivative-free
Kalman Filter, a when tracking a seesaw setpoint, b when tracking of a sinusoidal setpoint

been given in Figs. 4.16 and 4.17, for the Unscented Kalman Filter have been given
in Figs. 4.18 and 4.19, while for derivative-free Kalman filter have been given in
Figs. 4.20 and 4.21. Again the reference setpoint is denoted by the red line, the state
vector variables of the motor are denoted by the dashed blue line, while the estimated
state vector elements are described by the green line. The improved performance suc-
ceeded by the derivative-free nonlinear Kalman Filter (DKF) can be noticed once
more.
4.2 The Derivative-Free Nonlinear Kalman Filter 161

(a) DKF − control input (b) DKF − control input

400 400
control signal Vin of the motor (v.u.)

control signal Vin of the motor (v.u.)

300 300

200 200

100 100

0 0

−100 −100

−200 −200

−300 −300

−400 −400
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.15 Control signal of the derivative-free Kalman Filter-based control loop, a when tracking
a seesaw setpoint, b when tracking a sinusoidal setpoint

Table 4.1 MSE for state x1 x2 x3

estimation-based control
EKF 4.104e − 2 6.416e − 2 2.579e − 1
UKF 4.034e − 2 6.258e − 2 2.546e − 1
DKF 3.978e − 2 6.069e − 2 2.451e − 1

Table 4.2 MSE for nonlinear x1 x2 x3

filters
EKF 1.103e − 4 3.223e − 4 2.748e − 3
UKF 1.277e − 6 5.399e − 4 1.839e − 3
DKF 1.566e − 6 1.974e − 5 2.960e − 5

Using the derivative-free nonlinear Kalman Filter, process and measurement

noises affecting the motor’s model are taken into account in the filter’s recursion
and an optimal state estimate can be obtained despite the effects of these noises.
This holds if the uncertainties described as process noise are bounded. Moreover,
to compensate for additive external disturbances and model uncertainties one can
consider results on disturbance observers within a Kalman Filter framework [427,
570].
The proposed derivative-free Kalman Filtering approach can contribute to the
development of improved industrial control and fault diagnosis methods. First, the
transformation of the initial nonlinear system to the observer canonical form and the
previously analyzed design of the state estimator, assures stability of the nonlinear
filter and asymptotic elimination of the estimation error. Second, the derivative-free
162 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 1.5
EKF − x1 (b) EKF − x2
3

velocity x2 of the motor (rad/sec)

1 2
position x of the motor (rad)

0.5 1

0 0
1

−0.5 −1

−1 −2

−1.5 −3
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.16 State estimation-based control with the use of the Extended Kalman Filter for tracking a
sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x1 , b state variable
x2

(a) 6
EKF − x3 (b) EKF − control input
400
acceleration x of the motor (rad/sec2)

control signal Vin of the motor (v.u.)

4 300

200
2

100
0
0
−2
3

−100
−4
−200

−6 −300

−8 −400
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.17 State estimation-based control with the use of the Extended Kalman Filter for tracking a
sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x3 , b control input

Kalman Filtering approach and the transformation to the observer canonical form
make possible the application of smoothing algorithms (initially designed for linear
systems) that compensate for the effects of delayed sensor measurements in the
control loop. This can be particularly useful in the case of networked control and
fault diagnosis schemes where communication delays and packet drops can affect
the accuracy of the state estimation.
4.2 The Derivative-Free Nonlinear Kalman Filter 163

(a) UKF − x1 (b) UKF − x2

1.5 3

velocity x of the motor (rad/sec)

1 2
position x of the motor (rad)

0.5 1

0 0
1

2
−0.5 −1

−1 −2

−1.5 −3
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.18 State estimation-based control with the use of the Unscented Kalman Filter for tracking a
sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x1 , b state variable
x2

(a) UKF − x3 (b) UKF − control input

6 400
acceleration x of the motor (rad/sec 2)

control signal Vin of the motor (v.u.)

4 300

200
2
100
0
0
−2
3

−100
−4
−200

−6 −300

−8 −400
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.19 State estimation-based control with the use of the Unscented Kalman Filter for tracking
a sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x3 , b control input

4.2.5 Derivative-Free Kalman Filtering for MIMO

Nonlinear Systems

The results of Chap. 3 about transformation of MIMO nonlinear systems into the
canonical form will be generalized toward Derivative-free Kalman Filtering for
MIMO nonlinear dynamical systems. The proposed method for derivative-free
Kalman Filtering for MIMO nonlinear systems will be analyzed through an applica-
tion example. A 2-DOF rigid-link robotic manipulator is considered. The dynamic
164 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 1.5
DKF − x1 (b) 3
DKF − x2

velocity x of the motor (rad/sec)

1 2
position x of the motor (rad)

0.5 1

0 0
1

2
−0.5 −1

−1 −2

−1.5 −3
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.20 State estimation-based control with the use of the derivative-free nonlinear Kalman Filter
for tracking a sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x1 ,
b state variable x2

(a) DKF − x3 (b) DKF − control input

6 400
acceleration x of the motor (rad/sec 2)

control signal Vin of the motor (v.u.)

4 300

200
2

100
0
0
−2
3

−100
−4
−200

−6 −300

−8 −400
0 10 20 30 40 50 60 0 10 20 30 40 50 60
t (sec) t (sec)

Fig. 4.21 State estimation-based control with the use of the derivative-free nonlinear Kalman Filter
for tracking a sinusoidal setpoint of variable amplitude, frequency, and phase: a state variable x3 ,
b control input

model of the robot has been analyzed in Chap. 3 and it has been shown that (i) it satis-
fies differential flatness properties, (ii) it can be transformed into the linear canonical
form.
As previously shown this robotic system is a differentially flat one. Next, consid-
ering also the effects of additive disturbances to the joints of the robotic manipulator
the dynamic model becomes
4.2 The Derivative-Free Nonlinear Kalman Filter 165

ẍ1 = f 1 (x, t) + g1 (x, t)u + d̃1

(4.46)
ẍ3 = f 2 (x, t) + g2 (x, t)u + d̃2
       
ẍ1 f (x, t) g (x, t) d̃
= 1 + 1 u+ 1 (4.47)
ẍ3 f 2 (x, t) g2 (x, t) d̃2

Moreover, considering that the model’s structure and parameters are known and
that the additive input disturbance d̃i , i = 1, 2, . . . are also known, the following
feedback linearizing control input is defined
 −1        
g1 (x̂, t) ẍ1d f 1 (x̂, t) K 1T u c1
u= − − e+ (4.48)
g2 (x̂, t) ẍ3d f 2 (x̂, t) K 2T u c2

where [u c1 u c2 ]T is a supervisory control term that is used for the compensation

of the model’s uncertainties as well as of the external disturbances and K iT =
[k1i , k2i , . . . , kn−1
i , kni ] are the rows of the error feedback gain matrix [407, 410].
It is also noted that to perform efficient state estimation under such model uncer-
tainties and external disturbances one can consider results on disturbance observers
within a Kalman Filter framework [427].
Finally, the differentially flat robotic model is written in the Brunovsky (canonical)
form. Considering the state vector x ∈ R 4×1 , with the state variables defined in
Eq. (3.76), the following matrices are defined
⎛ ⎞ ⎛ ⎞
0100 00  
⎜0 0 0 0⎟ ⎜1 0 ⎟ 1000
A=⎝ ⎜ ⎟ ⎜ ⎟
, B = ⎝ ⎠, C = (4.49)
0 0 0 1⎠ 00 0010
0000 01

Using the matrices of Eq. (4.49) one obtains the Brunovsky form of the MIMO robot
model dynamics
ẋ = Ax + Bv
(4.50)
y = Cx

where the new input v is given by

     
f 1 (x, t) g (x, t) d̃
v= + 1 u+ 1 (4.51)
f 2 (x, t) g2 (x, t) d̃2

Prior to performing discrete-time Kalman Filtering for the model of Eq. (4.50) matri-
ces A, B, and C of Eq. (4.49) are turned into discrete-time ones using common dis-
cretization methods. These discrete-time matrices are denoted as Ad , Bd , and Cd ,
respectively.
For the robotic model of Eqs. (4.50) and (4.51) state estimation can be performed
using the standard Kalman Filter recursion, as described in Eqs. (4.5) and (4.6).
166 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

4.2.6 Simulation Tests

The performance of the proposed derivative-free nonlinear Kalman Filter was tested
in the nonlinear state estimation-based control for a 2-DOF rigid-link robotic manip-
ulator (Fig. 3.7). The differentially flat model of the robot and its transformation to
the Brunovksy form has been analyzed in Sect. 4.2.5. The flat outputs were taken to
be the robot’s joint angles y1 = x1 and y2 = x3 . It has been proven that all state
variables of the robotic model and the associated control inputs, i.e., the torques
applied by the motors to the links’ joints can be written as functions of the flat output
[y1 , y2 ] and of the associated derivatives.
The position and velocity variations for the sinusoidal setpoint for the first joint of
the robotic manipulator are depicted in Fig. 4.22. For the second joint of the 2-DOF
robot the tracking of the position and velocity setpoints is depicted in Fig. 4.23. The
control inputs (motor torques) applied to the first and second joint of the robotic
manipulator are shown in Fig. 4.24.
The performance of the state estimation-based control has been also tested in the
tracking of a seesaw setpoint. The position and velocity variations of the first joint
are demonstrated in Fig. 4.25. Similarly, the tracking of the position and velocity
reference setpoints for the second joint are depicted in Fig. 4.26. The control signal
in the case of tracking of a tracking a seesaw setpoint by the two joints of the robotic
manipulator is shown in Fig. 4.27. The estimated state variables were denoted as
green line whereas the real state variables were denoted as blue lines.

(a) 1.5 (b) 2

1.5
1
state variable x of the robot

state variable x of the robot

1
0.5
0.5
1

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.22 Derivative-free Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed
line) by the angle of the first joint of the robot (continuous line), b tracking of a sinusoidal velocity
setpoint (dashed line) by the angular velocity of the first joint of the robot (continuous line)
4.2 The Derivative-Free Nonlinear Kalman Filter 167

(a) 1.5 (b) 2

1.5
1

state variable x4 of the robot

state variable x3 of the robot

1
0.5
0.5

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.23 Derivative-free Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed
line) by the angle of the second joint of the robot (continuous line), b tracking of a sinusoidal
velocity setpoint (dashed line) by the angular velocity of the second joint of the robot (continuous
line)

(a) 60 (b) 26
24
55
22
control signal of joint 1

control signal of joint 2

50
20

45 18

16
40
14
35
12

30 10
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.24 Tracking of a sinusoidal setpoint by the 2-DOF robotic manipulator with the use of the
derivative-free Kalman Filter: a control input (motor torque) applied to the first joint, b control
input (motor torque) applied to the second joint

Moreover, the derivative-free nonlinear Kalman Filter was tested against the
Extended Kalman Filter which is among the most commonly used estimation meth-
ods for nonlinear dynamical systems. In the latter case, the position and velocity
variations for the sinusoidal setpoint for the first joint of the robotic manipulator are
depicted in Fig. 4.28. For the second joint of the 2-DOF robot the tracking of the
168 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 0.6 (b) 2

0.5 1.5

state variable x 2 of the robot

state variable x1 of the robot

0.4 1

0.3 0.5

0.2 0

0.1 −0.5

0 −1

−0.1 −1.5

−0.2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.25 Derivative-free Kalman Filtering: a tracking of a seesaw position setpoint (dashed line)
by the angle of the first joint of the robot (continuous line), b tracking of a seesaw velocity setpoint
(dashed line) by the angular velocity of the first joint of the robot (continuous line)

(a) 0.6 (b) 2

0.5 1.5
state variable x of the robot
state variable x of the robot

0.4 1

0.3 0.5
4
3

0.2 0

0.1 −0.5

0 −1

−0.1 −1.5

−0.2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.26 Derivative-free Kalman Filtering: a tracking of a seesaw position setpoint (dashed line)
by the angle of the second joint of the robot (continuous line), b tracking of a seesaw velocity
setpoint (dashed line) by the angular velocity of the second joint of the robot (continuous line)

position and velocity setpoints is depicted in Fig. 4.29. The control inputs (motor
torques) applied to the first and second joint of the robotic manipulator are shown in
Fig. 4.30.
The performance of the state estimation-based control, with the use of the
Extended Kalman Filter is also shown in the tracking of a seesaw setpoint. The
position and velocity variations of the first joint are demonstrated in Fig. 4.31. Simi-
larly, the tracking of the position and velocity reference setpoints for the second joint
4.2 The Derivative-Free Nonlinear Kalman Filter 169

(a) 60 (b) 30
55
25
50
control signal of joint 1

control signal of joint 2

20
45

40 15

35
10
30
5
25

20 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.27 Tracking of a seesaw setpoint by the 2-DOF robotic manipulator with the use of the
derivative-free Kalman Filter: a control input (motor torque) applied to the first joint, b control
input (motor torque) applied to the second joint

(a) 1.5 (b) 2

1.5
1
state variable x of the robot
state variable x of the robot

1
0.5
0.5
2
1

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.28 Extended Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed line) by
the angle of the first joint of the robot (continuous line), b tracking of a sinusoidal velocity setpoint
(dashed line) by the angular velocity of the first joint of the robot (continuous line)

are depicted in Fig. 4.32. The control signal in the case of tracking a seesaw setpoint
by the two joints of the robotic manipulator is shown in Fig. 4.33.
Comparing the estimation performed by the derivative-free MIMO nonlinear
Kalman Filter with the one performed by the Extended Kalman Filter it can be
noticed that the derivative-free filtering approach results in more accurate state
170 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 1.5 (b) 2

1.5
1
state variable x3 of the robot

state variable x4 of the robot

1
0.5
0.5

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.29 Extended Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed line)
by the angle of the second joint of the robot (continuous line), b tracking of a sinusoidal velocity
setpoint (dashed line) by the angular velocity of the second joint of the robot (continuous line)

(a) 60 (b) 30
59 29

58 28
control signal of joint 1

control signal of joint 2

57 27

56 26

55 25

54 24

53 23

52 22

51 21

50 20
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.30 Tracking of a sinusoidal setpoint by the 2-DOF robotic manipulator with the use of the
Extended Kalman Filter: a control input (motor torque) applied to the first joint, b control input
(motor torque) applied to the second joint

estimates. The precision of the provided estimation is the one depicted in Tables 4.1
and 4.2 of Sect. 4.2.4 for the case of the DC motor. Moreover, comparing the associ-
ated state estimation-based control loop that was based on the derivative-free MIMO
nonlinear Kalman Filter to the control that was based on the Extended Kalman
4.2 The Derivative-Free Nonlinear Kalman Filter 171

(a) 0.7 (b) 2

0.6 1.5
state variable x of the robot

state variable x of the robot

1
0.5
0.5
0.4
1

2
0
0.3
−0.5
0.2
−1

0.1 −1.5

0 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.31 Extended Kalman Filtering: a tracking of a seesaw position setpoint (dashed line) by
the angle of the first joint of the robot (continuous line), b tracking of a seesaw velocity setpoint
(dashed line) by the angular velocity of the first joint of the robot (continuous line)

(a) 0.7 (b) 2

0.6 1.5
state variable x of the robot

state variable x of the robot

1
0.5

0.5
0.4
3

0
0.3
−0.5
0.2
−1

0.1 −1.5

0 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.32 Extended Kalman Filtering: a tracking of a seesaw position setpoint (dashed line) by
the angle of the second joint of the robot (continuous line), b tracking of a seesaw velocity setpoint
(dashed line) by the angular velocity of the second joint of the robot (continuous line)

Filter it was observed that the first control scheme was significantly more robust and
capable of tracking with better accuracy the desirable trajectories. These findings
show the suitability of the considered derivative-free MIMO nonlinear Kalman Fil-
ter for tracking, control, and fault diagnosis tasks.
172 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 60 (b) 35

50 30
control signal of joint 1

control signal of joint 2

25
40

20
30
15

20
10

10
5

0 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.33 Tracking of a seesaw setpoint by the 2-DOF robotic manipulator with the Extended
Kalman Filter: a control input (motor torque) applied to the first joint, b control input (motor
torque) applied to the second joint

4.3 The Derivative-Free Distributed Nonlinear

Kalman Filter

4.3.1 Overview

This part of the chapter extends the use of the Derivative-free nonlinear Kalman
Filter toward distributed state estimation-based control of nonlinear MIMO systems.
Up to now an established approach for performing distributed state estimation has
been the Extended Information Fitter [427]. In the Extended Information Filter the
local filters do not exchange raw measurements but send to an aggregation filter
their local information matrices (local inverse covariance matrices which can be also
associated to Fisher Information Matrices) and their associated local information
state vectors (products of the local information matrices with the local state vectors)
[264, 414]. The Extended Information Filter performs fusion of state estimates from
local distributed Extended Kalman Filters which in turn are based on the assump-
tion of linearization of the nonlinear system dynamics by first-order Taylor series
expansion and truncation of the higher order linearization terms [329, 369, 402].
Moreover, the Extended Kalman Filter requires the computation of Jacobians which
in the case of high order nonlinear dynamical systems can be a cumbersome proce-
dure. This approach introduces cumulative errors to the state estimation performed
by the local Extended Kalman Filter recursion which is finally transferred to the mas-
ter filter where the aggregate state estimate of the controlled system is computed.
Consequently, these local estimation errors may result in the deterioration of the
performance of the associated control loop or even risk its stability [415, 427].
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter 173

To overcome the aforementioned weaknesses of the Extended Information Filter, a

differential flatness theory-based approach to distributed nonlinear Kalman Filtering
has been proposed [424, 434]. Distributed filtering is now based on a derivative-
free implementation of Kalman Filtering which is shown to be applicable to MIMO
nonlinear dynamical systems [430, 437]. In the proposed derivative-free Kalman
Filtering method the system is first subject to a linearization transformation that is
based on differential flatness theory [429, 495]. Next state estimation is performed by
applying the standard Kalman Filter recursion to the linearized model. Unlike EKF,
the proposed method provides estimates of the state vector of the nonlinear system
without the need for derivatives and Jacobians calculation. By avoiding linearization
approximations, the proposed filtering method improves the accuracy of estimation
of the system state variables, and results in smooth control signal variations and in
minimization of the tracking error of the associated control loop. The chapter extends
the results of [429, 437] towards a networked control scheme that is distorted by
random delays and measurement packets drop.
Finally, the chapter shows that the aggregate state vector produced by a derivative-
free distributed nonlinear Kalman Filter, suitably modified to compensate for com-
munication delays and packet drops, can be used for sensorless control and robotic
visual servoing (see Fig. 5.24). The problem of visual servoing over a network of
synchronized cameras has been previously studied in [478]. In this chapter, visual
servoing over a cameras network is considered for the nonlinear dynamic model of a
multi-DOF robotic manipulator (the considered robotic model stands for a nonlinear
MIMO dynamical system). The position of the robot’s end effector in the carte-
sian space (and equivalently the angles of the robotic links) is measured through m
cameras. In turn, these measurements are processed by m-distributed derivative-free
Kalman Filters thus providing m different estimates of the robot’s state vector. Next,
the local state estimates are fused with the use of the standard Information Filter.
After all, the aggregate estimation of the state vector is used in a control loop which
enables the robotic link to perform trajectory tracking.

4.3.2 Overview of the Extended Information Filter

4.3.2.1 Fusing Estimations from Local Distributed Filters

Again, the discrete-time nonlinear system of Eqs. (4.46) and (4.47) is considered. The
Extended Information Filter (EIF) performs fusion of the local state vector estimates
which are provided by local Extended Kalman Filters, using the Information matrix
and the Information state vector [264]. The Information Matrix is the inverse of the
state vector covariance matrix, and can be also associated to the Fisher Information
matrix [414]. The Information state vector is the product between the Information
matrix and the local state vector estimate
174 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

Y(k) = P −1 (k) = I (k)

−1 (4.52)
ŷ(k) = P − (k) x̂(k) = Y(k)x̂(k)

The update equation for the Information Matrix and the Information state vector are
given by

Y (k) = P − (k)−1 + JγT (k)R −1 (k)Jγ (k) = Y− (k) + I (k)

ŷ(k) = ŷ− (k) + JγT R(k)−1 [z(k) − γ (x(k)) + Jγ (k)x̂ − (k)] = ŷ− (k) + i(k),
(4.53)

where I (k) = JγT (k)R ( k)−1 Jγ (k) is the associated information matrix and, i(k) =
JγT (k)R ( k)−1 [(z(k) − γ (x(k))) + Jγ x̂ − (k)] is the information state contribution.
The predicted information state vector and Information matrix are obtained from

ŷ− (k) = P − (k)−1 x̂ − (k)

−1 (4.54)
Y− (k) = P − (k) = [Jφ (k)P − (k)Jφ (k)T + Q(k)]−1

It is assumed that an observation vector z i (k) is available for the N different sensor
sites (vision nodes) i = 1, 2, . . . , N and each vision node observes the robot accord-
ing to the local observation model, expressed by z i (k) = γ (x(k)) + vi (k), i =
1, 2, . . . , N , where the local noise vector vi (k)∼N (0, R i ) is assumed to be white
Gaussian and uncorrelated between sensors. The variance of a composite observa-
tion noise vector vk is expressed in terms of the block diagonal matrix R(k) =
diag[R(k)1 , . . . , R N (k)]T . The information contribution can be expressed by a lin-
ear combination of each local information state contribution i i and the associated
information matrix I i at the ith sensor site
N iT −1 i −
i(k) = i=1 Jγ (k)R (k) [z (k) − γ (x(k)) + Jγ (k) x̂ (k)]
i i i
N i T −1 i
(4.55)
I (k) = i=1 Jγ (k)R (k) Jγ (k).
i

Thus, the update equations for fusing the local state estimates is
N
ŷ(k) = ŷ− (k) + iT −1 i −
i=1 Jγ (k)R (k) [z (k) − γ (x(k)) + Jγ (k) x̂ (k)]
i i i
N i T (4.56)
Y(k) = Y− (k) + i=1 Jγ (k)R i (k)−1 Jγi (k)

It is noted that in the Extended Information Filter an aggregation (master) fusion

filter produces a global estimate by using the local sensor information provided by
each local filter. As in the case of the Extended Kalman Filter the local filters which
constitute the Extended information Filter can be written in terms of time update and
measurement update equations.
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter 175

Measurement update: Acquire z(k) and compute

Y (k) = P − (k)−1 + JγT (k)R(k)−1 Jγ (k)

or Y (k) = Y − (k) + I (k)
where I (k) = JγT (k)R −1 (k)Jγ (k), and, (4.57)
ŷ(k) = ŷ − (k) + JγT (k)R(k)−1 [z(k) − γ (x̂(k)) + Jγ x̂ − (k)]
or ŷ(k) = ŷ − (k) + i(k)

Time update: Compute

−1
Y − (k + 1) = P − (k + 1) = [Jφ (k)P(k)Jφ (k)T + Q(k)]−1
−1 −
(4.58)
and y − (k + 1) = P − (k + 1) x̂ (k + 1)

If the derivative-free nonlinear Kalman Filter is used in place of the Extended Kalman
Filter then in the EIF equations the following matrix substitutions should be per-
formed: Jφ (k)→Ad , Jγ (k)→Cd , where matrices Ad and Cd are the discrete-time
equivalents of matrices Ac and Cc which have been defined in Eq. (4.49). Matrices
Ad and Cd can be computed using established discretization methods, as already
explained in Chap. 3. Moreover, the covariance matrices P(k) and P − (k) are the
ones obtained from the linear Kalman Filter update equations given in Sect. 4.2.2.

4.3.2.2 Calculation of the Aggregate State Estimation

The outputs of the local filters are treated as measurements which are fed into the
aggregation fusion filter (see Fig. 4.34) [264]. Then each local filter is expressed by
its respective error covariance and estimate in terms of information contributions and
is described by

−1
Pi −1 (k) = Pi− (k) + JγT (k)R(k)−1 Jγ (k)x̂i (k) =
= Pi (k)(Pi− (k)−1 x̂i− (k)) + JγT (k)R(k)−1 [z i (k) − γ i (x(k)) + Jγi (k)x̂i− (k)]
(4.59)

The global estimate and the associated error covariance for the aggregate fusion filter
can be rewritten in terms of the computed estimates and covariances from the local
filters using the relations
176 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

Fig. 4.34 Fusion of the distributed state estimates with the use of the Extended Information Filter

JγT (k)R(k)−1 Jγ (k) = Pi (k)−1 − Pi− (k)−1 and

JγT (k)R(k)−1 [z i (k) − γ i (x(k)) + Jγi (k)x̂ − (k)] = (4.60)
= Pi (k)−1 x̂ i (k) − Pi (k)−1 x̂ i (k − 1)

For the general case of N local filters i = 1, . . . , N , the distributed filtering archi-
tecture is described by the following equations
N −1 − P − (k)−1 ]
P(k)−1 = P − (k)−1 + i=1 [Pi (k) i
N (4.61)
x̂(k) = P(k)[P − (k)−1 x̂ − (k) + i=1 (Pi (k)−1 x̂i (k) − Pi− (k)−1 x̂i− (k))]

The global state update equation in the above distributed filter can be written in
terms of the information state vector and of the information N matrix, i.e., ŷ(k) =
ŷ − (k) + i=1
N
( ŷi (k) − ŷi− (k)), and Ŷ (k) = Ŷ − (k) + i=1 (Ŷi (k) − Ŷi− (k)). From
Eq. (4.61) it can be seen that if a local filter (processing station) fails, then the local
covariance matrices and the local state estimates provided by the rest of the filters
will enable an accurate computation of the monitored system’s state vector.
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter 177

4.3.2.3 Derivative-Free Extended Information Filtering

As mentioned above, for the system of Eqs. (4.46) and (4.47), state estimation is
possible by applying the standard Kalman Filter. The system is first turned into
discrete-time form using common discretization methods and then the recursion of
the linear Kalman Filter described in Eqs. (4.5) and (4.6) is applied.
If the derivative-free Kalman Filter is used in place of the Extended Kalman Filter
then in the EIF equations the following matrix substitutions should be performed:
Jφ (k)→Ad , Jγ (k)→Cd , where matrices Ad and Cd are the discrete-time equivalents
of matrices A and C which have been defined Eq. (4.49) and which appear also in
the measurement and time update of the standard Kalman Filter recursion. Matrices
Ad and Cd can be computed using established discretization methods. Moreover, the
covariance matrices P(k) and P − (k) are the ones obtained from the linear Kalman
Filter update equations given in Sect. 4.2.2.

4.3.3 Distributed Filtering for Sensorless Control

4.3.3.1 Visual Servoing Over a Network of Synchronized Cameras

The example to be presented describes the control of a planar robot with the use of
a position-based visual servo that comprises multiple fixed cameras. The chapter’s
approach relies on neither position nor velocity sensors, and directly sets the motor
control current using only visual feedback. Direct visual servoing is implemented
using a distributed filtering scheme which permits to fuse the estimates of the robot’s
state vector computed by local filters, each one associated to a camera in the cameras
network (see Fig. 4.35). The cameras’ network can be based on multiple cameras
connected to a computer with a frame grabber to form a vision node (see Figs. 4.35
and 4.36).
The master computer communicates video synchronization information over the
network to each vision node. Typical sources of measurement noise include charge-
coupled device (CCD) noise, analog-to-digital (A/D) noise, and finite word-length
effects. Under ideal conditions, the effective noise variance from these sources should
remain relatively constant. Occlusions can be also considered as a noise source.
Finally, communication delays and packet drops in the transmission of measure-
ments from the vision sensors to the information processing nodes induce additional
disturbances which should be compensated by the virtual servoing control loop.
178 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

Fig. 4.35 Distributed cameras network and distributed information processing units for visual
servoing

4.3.3.2 Distributed Filtering for Fusion of the Robot’s State Estimates

Fusion of the local state estimates which are provided by filters running on the vision
nodes can improve the accuracy and robustness of the performed state estimation,
thus also improving the performance of the robot’s control loop [126, 478, 506,
507, 594]. Under the assumption of Gaussian noise, a possible approach for fusing
the state estimates from the distributed local filters is the derivative-free distributed
nonlinear Kalman Filter. As explained in Sect. 4.3.2, this filter (that is also named
derivative-free Extended Information Filter) provides an aggregate state estimate by
weighting the state vectors produced by local Kalman Filters with the inverse of the
associated estimation error covariance matrices.
Visual servoing over the previously described cameras network is considered for
the nonlinear dynamic model of a two-link robotic manipulator. The robot can be
programmed to execute a manufacturing task, such as disassembly or welding [532].
The position of the robot’s end effector in the cartesian space (and consequently
the angle for the robotic link) is measured by the aforementioned m distributed
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter 179

Fig. 4.36 Visual servoing based on fusion of state estimates provided by local derivative-free
nonlinear Kalman Filters

cameras. The proposed multi-camera based robotic control loop can be also useful
in other vision-based industrial robotic applications where the vision is occluded or
heavily disturbed by noise sources, e.g., cutting. In such applications there is need to
fuse measurements from multiple cameras so as to obtain redundancy in the visual
information and permit the robot to complete safely and within the specified accuracy
constraints its assigned tasks [365, 594].

4.3.4 Simulation Tests

The performance of the proposed derivative-free distributed nonlinear Kalman Filter

was tested in the benchmark problem of state estimation-based control for a 2-DOF
rigid-link robotic manipulator (Fig. 3.7). The differentially flat model of the robot and
its transformation to the Brunovksy form have been analyzed in subsection Chap. 3.
It was assumed that only measurements of the angle of the robot’s joints could be
obtained through the robot’s encoders. The reference setpoint for each joint was a
sinusoidal signal of amplitude 1.0 and period T = 10 s. At the beginning of the
180 4 Nonlinear Kalman Filtering Based on Differential Flatness Theory

(a) 1.5 (b) 2

1.5
1
state variable x1 of the robot

state variable x2 of the robot

1
0.5
0.5

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.37 Derivative-free Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed
line) by the angle of the first joint of the robot (continuous line), b tracking of a sinusoidal velocity
setpoint (dashed line) by the angular velocity of the first joint of the robot (continuous line)

second half of the simulation time an additive sinusoidal disturbance of amplitude

A = 0.5 and period T = 10 s was applied to the system. The estimated state variables
were denoted as green lines whereas the real state variables were denoted as blue
lines. The approximations f (x̂) and g(x̂) were used in the derivation of the control
law, given by Eq. (3.86).

(a) 1.5 (b) 2

1.5
1
state variable x of the robot

state variable x4 of the robot

1
0.5
0.5
3

0 0

−0.5
−0.5
−1
−1
−1.5

−1.5 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.38 Derivative-free Kalman Filtering: a tracking of a sinusoidal position setpoint (dashed
line) by the angle of the second joint of the robot (continuous line), b tracking of a sinusoidal
velocity setpoint (dashed line) by the angular velocity of the second joint of the robot (continuous
line)
4.3 The Derivative-Free Distributed Nonlinear Kalman Filter 181

(a) 60 (b) 26
24
55
22
control signal of joint 1

control signal of joint 2

50
20

45 18

16
40
14
35
12

30 10
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 4.39 Tracking of a sinusoidal setpoint by the 2-DOF robotic manipulator with the use of the
derivative-free Kalman Filter: a control input (motor torque) applied to the first joint, b control
input (motor torque) applied to the second joint

The position and velocity estimates for the first joint of the robot when tracking a
sinusoidal setpoint are depicted in Fig. 4.37. For the second joint of the 2-DOF robot,
and in case of tracking again a sinusoidal setpoint, the position and angular velocity
estimates are depicted in Fig. 4.38. The control inputs (motor torques) applied to the
first and second joint of the robotic manipulator are shown in Fig. 4.39.
Chapter 5
Differential Flatness Theory
and Industrial Robotics

5.1 Overview

In this chapter, the application of differential flatness theory to the design of

controllers and state estimators for industrial robotic manipulators is analyzed. First,
an adaptive neuro-fuzzy controller is designed for a class of underactuated nonlinear
robotic manipulators, under the constraint that the system’s model is unknown. The
control algorithm aims at satisfying the H-infinity tracking performance criterion,
which means that the influence of the modeling errors and the external disturbances
on the tracking error is attenuated to an arbitrary desirable level. After transforming
the robotic system into the canonical form, the resulting control inputs are shown to
contain nonlinear elements which depend on the system’s parameters. The nonlinear
terms which appear in the control inputs are approximated with the use of neuro-fuzzy
networks. It is shown that a suitable learning law can be defined for the aforemen-
tioned neuro-fuzzy approximators so as to preserve the closed-loop system stability.
With the use of Lyapunov stability analysis it is proven that the proposed adaptive
fuzzy control scheme results in H-infinity tracking performance. The efficiency of
the proposed adaptive fuzzy control scheme is checked in the case of a 2-DOF planar
robotic manipulator that has the structure of a closed-chain mechanism.
Next, the chapter proposes a solution to the problem of observer-based adaptive
fuzzy control for MIMO nonlinear dynamical systems, such as multi-DOF indus-
trial robotic manipulators. A flatness-based adaptive fuzzy controller is designed
for a class of nonlinear systems, under the constraint that only the system’s output
is measured and that the system’s model is unknown. This is a control problem of
elevated difficulty because uncertainty exists not only about the system’s dynamic
model but also about the system’s state vector measurements. The control algorithm
aims at satisfying the H-infinity tracking performance criterion, which means that
improved robustness is provided against modeling errors and external disturbances.
By applying differential flatness theory, the MIMO robotic system is transformed
into the canonical form. The resulting control inputs are shown to contain nonlinear

© Springer International Publishing Switzerland 2015 183

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_5
184 5 Differential Flatness Theory and Industrial Robotics

elements which depend on the system’s parameters. The nonlinear terms which
appear in the control inputs of the robot are approximated with the use of neuro-fuzzy
networks. Moreover, since only the system’s output is measurable the complete state
vector has to be reconstructed with the use of a state observer. It is shown that a suit-
able learning law can be defined for the aforementioned neuro-fuzzy approximators
so as to preserve the closed-loop system stability. With the use of Lyapunov stability
analysis, it is proven that the proposed observer-based adaptive neuro-fuzzy control
scheme results in H-inifnity tracking performance.
The chapter also shows that apart from adaptive control for the robotic manipula-
tors, it is possible to develop Kalman Filter-based control with the use of differential
flatness theory. The Derivative-free nonlinear Kalman Filter is used for developing
a robust controller which can be applied to underactuated MIMO robotic systems.
The control problem for underactuated robots is nontrivial and becomes further com-
plicated if the robot are subjected to model uncertainties and external disturbances.
Using differential flatness theory, it is shown that the model of a closed-chain 2-DOF
robotic manipulator can be transformed to linear canonical form. For the linearized
equivalent of the robotic system, it is shown that a state feedback controller can be
designed. Since certain elements of the state vector of the linearized system cannot
be measured directly, it is proposed to estimate them with the use of the differential
flatness theory-based method for state estimation (analyzed in Chap. 4) which is the
Derivative-free nonlinear Kalman Filter. Moreover, by redesigning the Kalman Filter
as a disturbance observer, it is shown that one can estimate simultaneously exter-
nal disturbances terms that affect the robotic model or disturbance terms which are
associated with parametric uncertainty. The efficiency of the proposed Kalman Filter-
based control scheme is checked in the case of a 2-DOF planar robotic manipulator
that has the structure of a closed-chain mechanism.
Another section of the chapter analyzes a distributed filtering scheme for robotic
visual servoing capable of compensating for modeling uncertainties and external
disturbances. The visual servoing robotic scheme that was introduced in Chap. 4 is
revisited. This time, the following disturbances are assumed to affect the control
loop: (i) Contact forces at the manipulator’s end effector, (ii) missing measurements
about the robot’s motion characteristics (the measurements transmission link used for
network-based control exhibits delays or data losses). First, it is assumed that during
the compliance task performed by the robot unknown or varying forces are exerted
on its end effector. The derivative-free Extended Information Filter (derivative-free
distributed nonlinear Kalman Filter) is suitably modified through the use of concepts
from disturbance observers theory to compensate for the effect of these perturbations.
Subsequently, the chapter examines the use of the Derivative-free nonlinear
Kalman Filter in the compensation of delayed measurements and data losses in
robotic visual servoing. An analysis is given about the problem of distributed non-
linear filtering over a communication/sensors network, and the use of the estimated
state vector in a control loop. Distributed filtering is now based on a derivative-free
implementation of nonlinear Kalman Filtering. It is shown how the proposed distrib-
uted filtering method can succeed compensation of random delays and packet drops
5.1 Overview 185

which may appear during the transmission of measurements and of state vector esti-
mates, thus assuring a reliable performance of the distributed filtering-based control
scheme. Evaluation tests confirm the improved performance of the Derivative-free
distributed nonlinear Kalman Filter against the Extended Information Filter.

5.2 Adaptive Fuzzy Control of Underactuated

MIMO Robots

5.2.1 Overview

Control of underactuated robots is an important problem for industrial systems tech-

nology. The design of robotic mechanisms that can be controlled with a smaller
number of actuators of actuators than their degrees of freedom enables to reduce
cost and weight of robots and to succeed robustness in the case of actuators’ fail-
ures. The control problem of underactuated robotic manipulators has been studied
in several research articles during the last years [159, 201, 261, 262, 325, 366, 572,
597]. In [159], the property of differential flatness for a class of planar underactuated
open-chain robots having a specific inertia distribution, but driven by only a small
number of actuators has been analyzed. In [596], it was shown that closed-chain
underactuated robots satisfy differential flatness properties and this enables their
transformation into a linearized form for which the design of a feedback controller
becomes easier. In [572], energy-based control for underactuated robotic manipula-
tors has been proposed. In [261, 262] a Lyapunov-based approach to the design of
efficient control for underactuated robots is proposed. In [366], passive velocity field
control and decoupling vector field has been applied to the control of underactuated
mechanical systems. In [325], the problem of point-to-point control for underac-
tuated robotic manipulators has been presented. In [201], an open-loop vibrational
control for an underactuated mechanical system has been studied.
After transformation of the underactuated robotic model to the linear canonical
form, the resulting control input for the underactuated robotic mechanism is shown
to contain nonlinear elements which depend on the system’s parameters. If the para-
meters of the robot are unknown, then the nonlinear terms which appear in the control
signal can be approximated with the use of neuro-fuzzy networks. In this section, it
is shown that a suitable learning law can be defined for the aforementioned neuro-
fuzzy approximators so as to preserve the closed-loop system stability. Lyapunov
stability analysis proves also that the proposed flatness-based adaptive fuzzy control
scheme results in H∞ tracking performance, in accordance to the results of [405,
410, 413, 433].
It is shown that the only assumption needed for the design of the controller and
for succeeding H∞ tracking performance for the control loop is that there exists a
solution for a Riccati equation associated to the linearized error dynamics of the
186 5 Differential Flatness Theory and Industrial Robotics

differentially flat model. This condition holds for underactuated robotic systems,
thus providing a systematic approach to the design of reliable controllers for such
systems [426, 433].

5.2.2 Dynamic Model of the Closed-Chain 2-DOF

Robotic System

5.2.2.1 The 2-DOF Planar Underactuated Robot

The considered closed-chain 2-DOF robotic system depicted in Fig. 5.1 consists of
four bodies: (i) Bodies 1 and 2 are two sliders with masses m1 and m2 , respectively
[159, 597]. Body 3 is connected with a revolute joint to body 1 and has mass m3 ,
length l3 while its moment of inertia is I3 . Similarly body 4 is connected to body 2
with a revolute joint, has mass m4 , length l4 while the associated moment of inertia
is I4 . The motion of the system takes place in the 2D xy plane depicted in Fig. 5.1
while its dynamics is subjected to gravity.
The state variables for the robotic system are as follows q = [q1 , q2 , q3 , q4 ]T : q1
is the displacement of mass m1 along the x-axis, q3 is the turn angle of body 3 round
the revolute joint A. q2 is the displacement of m2 along the x-axis and q4 is the turn
angle of body 4 round the revolute joint B. The following geometric constraints hold:

l3 sin(q3 ) = l4 sin(q4 )
(5.1)
q1 + l3 cos(q3 ) = q2 + l4 cos(q4 )

The control inputs exerted on the robotic model are the horizontal force F1 that
causes the displacement of mass m1 along the x-axis, the torque T3 that causes the
rotation of the link with length l3 round the revolute joint A, the horizontal force

Fig. 5.1 A two-part

underactuated robotic system
constituting a closed-chain
mechanism
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 187

F2 that causes the displacement of mass m2 along the x-axis and the torque T4 that
causes the rotation of the link with length l4 round the revolute joint B. Thus, in the
most generic case the input vector can be u = [ f1 , T3 , f2 , T4 ]T . According the Euler-
Lagrange analysis, the dynamic model of the robotic manipulator is obtained [597]

A(q)q̈ + h(q, q̇) = u (5.2)

where ⎛ ⎞
a11 a12 0 0
⎜a21 a22 0 0 ⎟
A(q) = ⎜
⎝ 0
⎟ (5.3)
0 a23 a24 ⎠
0 0 a43 a44

with a11 = m1 + m3 , a12 = a21 = −m3 lc3 sin(q3 ), a22 = m3 lc23 + I3 , a33 = m2 + m4 ,
a34 = a43 = −m4 lc4 sin(q4 ), a44 = m4 lc24 + I4 , and
⎛ ⎞
−m3 lc3 cos(q3 )q̇32
⎜ m3 glc cos(q3 ) ⎟
h(q, q̇) = ⎜ 3
⎝−m4 lc4 cos(q4 )q̇2 ⎠
⎟ (5.4)
4
m4 glc4 cos(q4 )

5.2.2.2 Proof of the Robot’s Dynamic Model Using

the Euler-Lagrange Method

The closed-chain robotic system of Fig. 5.1, presented initially in Sect. 5.2.2 is con-
sidered again. This comprises four masses, out of which mass m1 and mass m2 slide
along the x-axis, while masses m3 and m4 associated with links l3 and l4 perform
both translational and rotational motion in the xy-plane.
For mass m1 one has that the potential energy is P1 = 0 while the kinetic energy
is K1 = 21 m1 q̇12 . For mass m2 one has again that the potential energy is P2 = 0 while
the kinetic energy is K2 = 21 m2 q̇22 .
For mass m3 the potential energy is given by

P3 = m3 glc3 sin(q3 ) (5.5)

The kinetic energy of mass m3 (link l3 ) is due to both translational motion and
rotational motion. Denoting by I3 the moment of inertia for the rotation of the link
l3 , one has the kinetic energy that is associated with the rotational motion is K3,1 =
1 2
2 I3 q̇3 . The kinetic energy of link l3 that is associated with the translational motion
is denoted as K3,2 and is given by

K3,2 = 21 m3 Ċ3 Ċ3T (5.6)

188 5 Differential Flatness Theory and Industrial Robotics

Ċ3 is the velocity of the center of gravity of mass (link) 3 in cartesian coordinates.
Using that the position of the center of gravity of link 3 is the vector

C3 = [q1 + lc3 cos(q3 ), lc3 sin(q3 )] (5.7)

its velocity is computed as follows

Ċ3 = [q̇1 − lc3 sin(q3 )q̇3 , lc3 cos(q3 )q̇3 ] (5.8)

Thus, using Eq. (5.8) the kinetic energy of link 3 that is due to its translational motion
is given by

K3,2 = 21 m3 Ċ3 Ċ3T ⇒

K3,2 = 21 m3 [(q̇1 − lc3 sin(q3 )q̇3 )2 + (lc3 cos(q3 )q̇3 )2 ]⇒ (5.9)
K3,2 = 2 m3 [q̇1
1 2 + lc23 q̇32 − 2lc3 sin(q3 )q̇1 q̇3 ]

Using K3,1 and K3,2 one has that the aggregate kinetic energy of mass (link) m3 is

K3 = 21 I3 q̇32 + 21 m3 [q̇12 + lc23 q̇32 − 2lc3 sin(q3 )q̇1 q̇3 ] (5.10)

For mass m4 the potential energy is given by

P4 = m4 glc4 sin(q4 ) (5.11)

The kinetic energy of mass m4 (link l4 ) is due to both translational motion and
rotational motion. Denoting by I4 the moment of inertia for the rotation of the link
l4 , one has the kinetic energy that is associated with the rotational motion is K4,1 =
1 2
2 I4 q̇4 . The kinetic energy of link l4 that is associated with the translational motion
is denoted as K4,2 and is given by

K4,2 = 21 m4 Ċ4 Ċ4T (5.12)

Ċ4 is the velocity of the center of gravity of mass (link) 4 in cartesian coordinates.
Using that the position of the center of gravity of link 4 is the vector

C4 = [q2 + lc4 cos(q4 ), lc4 sin(q4 )] (5.13)

its velocity is computed as follows

Ċ4 = [q̇2 − lc4 sin(q4 )q̇4 , lc4 cos(q4 )q̇4 ] (5.14)

Thus, using Eq. (5.14) the kinetic energy of link 4 that is due to its translational
motion is given by
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 189

K4,2 = 21 m4 Ċ4 Ċ4T ⇒

K4,2 = 2 m4 [(q̇2 − lc4 sin(q4 )q̇4 )2 + (lc4 cos(q4 )q̇4 )2 ]⇒
1
(5.15)
K4,2 = 21 m4 [q̇42 + lc24 q̇42 − 2lc4 sin(q4 )q̇2 q̇4 ]

Using K4,1 and K4,2 one has that the aggregate kinetic energy of mass (link) m4 is

K4 = 21 I4 q̇42 + 21 m4 [q̇22 + lc24 q̇42 − 2lc4 sin(q4 )q̇2 q̇4 ] (5.16)

The Lagrangian of the considered robotic system (closed kinematic mechanism) is

given by
 
L = 4i=1 Ki − 4i=1 Pi ⇒

L = 21 m1 q̇12 + 21 m2 q̇22 +
+ 21 I3 q̇32 + 21 m3 [q̇12 + lc23 q̇32 − 2lc3 sin(q3 )q̇1 q̇3 ]+ (5.17)
+ 21 I4 q̇42 + 2 m4 [q̇2
1 2 + lc24 q̇42 − 2lc4 sin(q4 )q̇2 q̇4 ]+
−m3 glc3 sin(q3 ) − m4 glc4 sin(q4 )

Next, it can be computed that

∂L
∂ q̇1 = m1 q̇1 + m3 q̇1 + m3 lc3 sin(q3 )q̇3 (5.18)

and by differentiating Eq. (5.18) with respect to time gives

∂ ∂L
∂t ∂ q̇1 = (m1 + m3 )q̈1 − m3 lc3 cos(q3 )q̇32 − m3 lc3 sin(q3 )q̈3 (5.19)

while one also obtains

∂L
∂q1 =0 (5.20)

In a similar manner, one computes

∂L
∂ q̇2 = m2 q̇2 + m4 q̇2 − m4 lc4 sin(q4 )q̇4 (5.21)

and by differentiating Eq. (5.21) with respect to time gives

∂ ∂L
∂t ∂ q̇2 = (m2 + m4 )q̈2 − m4 lc4 cos(q4 )q̇42 − m4 lc4 sin(q4 )q̈4 (5.22)

while one also obtains

∂L
∂q2 =0 (5.23)

In an equivalent way one finds

∂L
∂ q̇3 = I3 q̇3 + m3 lc23 q̇3 − m3 lc3 sin(q3 )q̇1 (5.24)

and by differentiating Eq. (5.24) with respect to time gives

190 5 Differential Flatness Theory and Industrial Robotics

∂ ∂L
∂t ∂ q̇3 = I3 q̈3 + m3 lc23 q̈3 − m3 lc3 cos(q3 )q̇1 q̇3 − m3 lc3 sin(q3 )q̈1 (5.25)

while one also obtains

∂L
∂q3 = −m3 lc3 cos(q3 )q̇1 q̇3 − m3 glc3 cos(q3 ) (5.26)

Following the same procedure one gets

∂L
∂ q̇4 = I4 q̇4 + m4 lc24 q̇4 − m4 lc4 sin(q4 )q̇2 (5.27)

and by differentiating Eq. (5.27) with respect to time gives

∂ ∂L
∂t ∂ q̇4 = I4 q̈4 + m4 lc24 q̈4 − m4 lc4 cos(q4 )q̇2 q̇4 − m4 lc4 sin(q4 )q̈2 (5.28)

while one also obtains

∂L
∂q4 = −m4 lc4 cos(q4 )q̇2 q̇4 − m4 glc4 cos(q4 ) (5.29)

Next, by using the previous equations and by applying the Euler-Lagrange principle
one gets
∂ ∂L ∂L
∂t ∂ q̇1 − ∂q1 = F1 ⇒ (5.30)
(m1 + m3 )q̈1 − m3 lc3 sin(q3)q̈3 − m3 lc3 cos(q3 )q̇32 = F1

∂ ∂L ∂L
− ∂q
∂t ∂ q̇2 = F2 ⇒
2 (5.31)
(m2 + m4 )q̈2 − m4 lc4 sin(q4)q̈4 − m4 lc4 cos(q4 )q̇42 = F2

∂ ∂L ∂L
− ∂q
∂t ∂ q̇3 3
= T1 ⇒
−m3 lc3 sin(q3 )q̈1 + (I3 + m3 lc23 )q̈3 − m3 lc3 cos(q3 )q̇1 q̇3 +
+m3 lc3 cos(q3 )q̇1 q̇3 + m3 glc3 cos(q3 ) = T1 ⇒ (5.32)

−m3 lc3 sin(q3 )q̈1 + (I3 + m3 lc23 )q̈3 + m3 glc3 cos(q3 ) = T1

∂ ∂L ∂L
− ∂q
∂t ∂ q̇4 4
= T2 ⇒
−m4 lc4 sin(q4 )q̈2 + (I4 + m4 lc24 )q̈4 − m4 lc4 cos(q4 )q̇2 q̇4 +
+m4 lc4 cos(q4 )q̇2 q̇4 + m4 glc4 cos(q4 ) = T2 ⇒ (5.33)

−m4 lc4 sin(q4 )q̈2 + (I4 + m4 lc24 )q̈4 + m4 glc4 cos(q4 ) = T2

By considering the state vector q̃ = [q1 , q3 , q2 , q4 ]T and by grouping together

Eqs. (5.30), (5.32) and (5.31), (5.33) one has

(m1 + m3 )q̈1 − m3 lc3 sin(q3)q̈3 − m3 lc3 cos(q3 )q̇32 = F1

(5.34)
−m3 lc3 sin(q3 )q̈1 + (I3 + m3 lc23 )q̈3 + m3 glc3 cos(q3 ) = T1
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 191

(m2 + m4 )q̈2 − m4 lc4 sin(q4)q̈4 − m4 lc4 cos(q4 )q̇42 = F2

(5.35)
−m4 lc4 sin(q4 )q̈2 + (I4 + m4 lc24 )q̈4 + m4 glc4 cos(q4 ) = T2

Equations (5.34) and (5.35) can now be written in matrix form

⎛ ⎞⎛ ⎞
(m1 + m3 ) −m3 lc3 sin(q3 ) 0 0 q̈1
⎜−m3 lc3 sin(q3 ) (I3 + m3 lc2 ) 0 0 ⎟ ⎜q̈2 ⎟
⎜ 3 ⎟⎜ ⎟+
⎝ 0 0 (m2 + m4 ) −m4 lc4 sin(q4 ) ⎠ ⎝q̈3 ⎠
0 0 −m4 lc4 sin(q4 ) (I4 + m4 lc24 ) q̈4
⎛ ⎞ ⎛ ⎞
−m3 lc3 cos(q3 )q̇32 F1
⎜ m3 glc cos(q3 ) ⎟ ⎜T1 ⎟
+⎜ ⎟ ⎜ ⎟
⎝−m4 lc4 cos(q4 )q̇2 ⎠ = ⎝F2 ⎠
3

4
m4 glc4 cos(q4 ) T2
(5.36)
Equation (5.36) can also take the compact matrix form

A(q̃)q̃¨ + h(q̃, q̃)

˙ = ũ (5.37)

This is the description of the robotic mechanism which has been given in Eqs. (5.2),
(5.3) and (5.4) .

5.2.2.3 Feedback Linearizability of the Robot’s Dynamic Model

Next, the case in which l3 = l4 is examined. Moreover, it is considered that the mass
m2 is connected to a spring with elasticity k2 . Finally, it is assumed that the only
inputs applied to the robotic model are u1 = f1 and u2 = T3 . Then the dynamic
model of the robot becomes [597]

A(ql )q̈ + h(ql , q̇l ) = [u1 , u2 ]T (5.38)

where ql = [q1 , q3 ]T and q = [q1 , q2 , q3 , q4 ]T . The inertia and Coriolis matrices

are defined as

M1 + M2 −2M2 l3 sin(q3 )
A(ql ) = (5.39)
−2M2 l3 sin(q3 ) I3 + I4 + 4M2 l32 sin2 (q3 )

−2M2 l3 q̇32 cos(q3 ) + k2 ld

h(ql , q̇l ) = (5.40)
k4 (q3 − π ) + 2l3 sin(q3 )(2M2 l3 q̇32 cos(q3 ) − k2 ld )

where M1 = m1 + m2 , M2 = m2 + m4 , ld = q1 + 2l3 cos(q3 ) − L. Denoting

x = [q1 , q3 , q̇1 , q̇3 ]T , the robot’s dynamic model can be written in the following
state-space form:
ẋ = f (x) + g1 (x)u1 + g2 (x)u2 (5.41)
192 5 Differential Flatness Theory and Industrial Robotics

where
⎛ ⎞
q̇1
⎜ q̇3 ⎟
⎜ −k l (I +I )+2l M (k (π −q )sin(q )+(I +I )q̇2 cos(q )) ⎟
⎜ 2d 3 4 3 2 4 3 3 3 4 3 3 ⎟
f (x) = ⎜
⎜ M2 (I3 +I4 )+M1 (I3 +I4 +4M2 l32 sin2 (q3 )) ⎟
⎟ (5.42)
⎜ ⎟
⎝ ⎠
k4 (M1 +M2 )(π −q3 )+2I3 M1 sin(q3 )(k2 ld −2l3 M2 q̇3 cos(q3 ))
2

M2 (I3 +I4 )+M1 (I3 +I4 +4M2 l32 sin2 (q3 ))

⎛ ⎞
0
⎜ 0 ⎟
⎜ ⎟
⎜ I3 +I4 +4M2 l32 sin2 (q3 ) ⎟
g1 (x) = ⎜ ⎟
⎜ (I3 +I4 )M2 +M1 (I3 +I4 +4M2 l3 sin (q3 )) ⎟
2 2 (5.43)
⎜ ⎟
⎝ ⎠
2M2 l3 sin(q3 )
(I3 +I4 )M2 +M1 (I3 +I4 +4M2 l32 sin2 (q3 ))

⎛ ⎞
0
⎜ 0 ⎟
⎜ 2M2 l3 sin(q3 ) ⎟
⎜ ⎟
g2 (x) = ⎜ (I3 +I4 )M2 +M1 (I3 +I4 +4M2 l32 sin2 (q3 )) ⎟ (5.44)
⎜ ⎟
⎝ ⎠
M1 +M2
(I3 +I4 )M2 +M1 (I3 +I4 +4M2 l32 sin2 (q3 ))

The robotic system is underactuated when only one of the control inputs is enabled.
It can be proven that when the only input is u1 = F1 then the robotic system is not
static feedback linearizable.
Next, the case in which the only control input is u2 = T3 is examined. When
k2 = 0 and k4 = 0 then the robotic model admits static feedback linearization.
Equivalently, with the use of the feedback linearization criteria stated in Chap. 1
this means that (i) the distribution D3 = [g(x), adf (x) g(x), adf2(x) g(x), adf3(x) g(x)]
has full rank and (ii) the vector fields D0 = [g(x)], D1 = [g(x), adf (x) g(x)] and
D2 = [g(x), adf (x) g(x), adf2(x) g(x)] are involutive.

5.2.3 Linearization of the Closed-Chain 2-DOF Robotic

System Using Lie Algebra Theory

After having proven that the closed-chain robotic manipulator admits static feedback
linearization, its transformation to a linear form will be attempted using suitably
chosen feedback linearizing outputs. The following linearizing output is defined first

z1 = y = h1 (q) = (M1 + M2 )q1 + 2M2 l3 cos(q3 ) (5.45)

5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 193

Applying Lie-algebra theory, it holds that [238]

⎛ ⎞
f1
 ⎜ ⎟
∂h1 ∂h1 ∂h1 ∂h1 ⎜f2 ⎟
z2 = Lf h1 = ∂q1 ∂q3 ∂ q̇1 ∂ q̇3 ⎝ ⎠ ⇒
f3
f4 ⎛ ⎞
f1
 ⎜f2 ⎟ (5.46)
z2 = Lf h1 = (M1 + M2 −2M2 l3 sin(q3 ) 0 0) ⎜ ⎟
⎝ f3 ⎠ ⇒
f4
z2 = Lf h1 = (M1 + M2 )f1 − 2M2 l3 sin(q3 )f2 ⇒
z2 = Lf h1 = (M1 + M2 )q̇1 − 2M2 l3 sin(q3 )q̇3

Similarly, it holds
⎛ ⎞
f1
  ⎜f ⎟
∂z2 ∂z2 ∂z2 ∂z2 ⎜ 2 ⎟
z3 = Lf2 h1 = ∂q1 ∂q3 ∂ q̇1 ∂ q̇3 ⎝f ⎠ ⇒
3
f4 ⎛ ⎞
f1 (5.47)
 ⎜ f2 ⎟
⎜
z3 = Lf h1 = 0 −2M2 l3 cos(q3 )q̇3 (M1 + M2 ) −2M2 l3 sin(q3 ) ⎝ ⎠ ⇒
2 ⎟
f3
f4
z3 = Lf2 h1 = (M1 + M2 )f3 − 2M2 l3 sin(q3 )f4 − 2M2 l3 cos(q3 )q̇3 f2

It holds that

q̈1 = f3 + ga3 u1 + gb3 u2 ⇒q̈1 = f3 + gb3 u2 ⇒f3 = q̈1 − gb3 u2

(5.48)
q̈3 = f4 + ga4 u1 + gb4 u2 ⇒q̈4 = f4 + gb4 u2 ⇒f4 = q̈3 − gb4 u2

Therefore, it holds

z3 = (M1 + M2 )f3 − 2M2 l3 sin(q3 )f4 − 2M2 l3 cos(q̇3 )f2 ⇒

z3 = (M1 + M2 )q̈1 − (M1 + M2 )gb2 u2 − 2M2 l3 sin(q3 )q̈3 + (5.49)
+2M2 l3 sin(q3 )gb4 u2 − 2M2 l3 cos(q3 )q̇32

or equivalently
z3 = (M1 + M2 )q̈1 − 2M2 l3 sin(q3 )q̈3

(M1 +M2 )2M2 l3 sin(q3 )

− (I u2
3 +I4 )M2 +M1 (I3 +I4 +4M2 l3 sin (q3 ))
2 2

(5.50)
(M1 +M2 )2M2 l3 sin(q3 )
+ (I u2
3 +I4 )M2 +M1 (I3 +I4 +4M2 l3 sin (q3 ))
2 2

−2M2 l3 cos(q3 )q̇32

194 5 Differential Flatness Theory and Industrial Robotics

Consequently, it holds that

z3 = (M1 + M2 )q̈1 − 2M2 l3 sin(q3 )q̈3 − 2M2 l3 q̇32 cos(q3 ) (5.51)

Using Eqs. (5.38)–(5.40) one obtains that

(M1 + M2 )q̈1 − 2M2 l3 sin(q3 )q̈3 = 2M2 l3 q̇32 cos(q3 ) − k2 ld + u1 (5.52)

with u1 = 0 due to underactuation. Therefore, it holds

z3 = 2M2 l3 q̇32 cos(q3 ) − k2 ld − 2M2 l3 q̇32 cos(q3 )⇒

(5.53)
z3 = −k2 ld ⇒z3 = −k2 (q1 + 2l3 cos(q3 ) − L)

Similarly, one has

⎛ ⎞
f1
  ⎜f ⎟
∂z3 ∂z3 ∂z3 ∂z3 ⎜ 2 ⎟
z4 = Lf3 h1 (q) = ∂q1 ∂q3 ∂ q̇1 ∂ q̇3 ⎝f ⎠ ⇒
3
f4
⎛ ⎞ (5.54)
f1
 ⎜f2 ⎟
z4 = −k2 2k2 l3 sin(q3 ) 0 0 ⎜ ⎟
⎝ f3 ⎠ ⇒
f4
z4 = −k2 f1 + 2k2 l3 sin(q3 )f2 ⇒z4 = −k2 q̇1 + 2k2 l3 sin(q3 )q̇3

Moreover, it holds that

ż4 = Lf4 h1 + Lga Lf3 h1 u1 + Lgb Lf3 h1 u2 (5.55)

Therefore,
⎛ ⎞
f1
  f ⎟
∂z4 ∂z4 ∂z4 ∂z4 ⎜
Lf h1 = ∂q1 ∂q3 ∂ q̇1 ∂ q̇3 ⎜
4 2⎟
⎝ f3 ⎠ ⇒
f4 ⎛ ⎞
f1
 ⎜f2 ⎟ (5.56)
Lf h1 = 0 2k2 l3 cos(q3 )q̇3 −k2 2k2 l3 sin(q3 ) ⎜
4 ⎟
⎝ f3 ⎠ ⇒
f4
Lf4 h1 = 2k2 l3 cos(q3 )q̇3 f2 − k2 f3 + 2k2 l3 sin(q3 )f4 ⇒
Lf4 h1 = 2k2 l3 cos(q3 )q̇32 − k2 f3 + 2k2 l3 sin(q3 )f4
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 195

It holds that u1 = 0 and

⎛ ⎞
gb1
  g ⎟
∂z4 ∂z4 ∂z4 ∂z4 ⎜ ⎜ b2 ⎟
Lgb Lf3 h1 = ∂q1 ∂q3 ∂ q̇1 ∂ q̇3 ⎝g ⎠ ⇒
b3
gb4 ⎛ ⎞ (5.57)
gb1
 ⎜gb2 ⎟
Lgb Lf3 h1 = 0 2k3 l3 cos(q3 )q̇3 −k2 2k2 l3 sin(q3 ) ⎜ ⎟
⎝gb3 ⎠
gb4

which can be also written as

2M2 l3 sin(q3 )
Lgb Lf3 h1 = −k2 (I +
3 +I4 )M2 +M1 (I3 +I4 +4M2 l3 sin (q3 ))
2 2
M1 +M2
+2k2 l3 sin(q3 ) (I3 +I4 )M2 +M1 (I3 +I4 +4M2 l32 sin2 (q3 ))
⇒
(5.58)
2k2 l3 sin(q3 )M1
Lgb Lf3 h1 = (I3 +I4 )M2 +M1 (I3 +I4 +4M2 l32 sin2 (q3 ))

Using next the relation

ż4 = Lf(4) h1 + Lga Lf3 h1 u1 + Lgb Lf3 h1 u2 = v (5.59)

and that ż1 = z2 , ż2 = z3 , ż3 = z4 one has that the robotic model can be written in
the linear canonical (Brunovsky) form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż1 0 1 0 0 z1 0
⎜ż2 ⎟ ⎜0 0 1 0⎟ ⎜ z2 ⎟ ⎜ 0 ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟v (5.60)
⎝ż3 ⎠ ⎝0 0 0 1 ⎠ ⎝ z3 ⎠ ⎝ 0 ⎠
ż4 0 0 0 0 z4 1

For the linearized system, a suitable feedback control law is

v = ż4d − k1 (z4 − z4d ) − k2 (z3 − z3d ) − k3 (z2 − z2d ) − k4 (z1 − z1d ) (5.61)

5.2.4 Differential Flatness of the Underactuated Manipulator

5.2.4.1 Differential Flatness of the Closed-Chain 2-DOF Robotic System

Next, a linearized equivalent description for the robot’s kinematic chain will be
obtained with the use of differential flatness theory. The following flat output is
chosen
y = (M1 + M2 )q1 + 2M2 l3 cos(q3 ) (5.62)
196 5 Differential Flatness Theory and Industrial Robotics

It holds that

ẏ = (M1 + M2 )q̇1 − 2M2 l3 sin(q3 )q̇3 ⇒

(5.63)
ÿ = (M1 + M2 )q̈1 = 2M2 l3 sin(q3 )q̈3 − 2M2 l3 cos(q3 )q̇32

Using Eqs. (5.2)–(5.4) it holds that

(M1 + M2 )q̈1 − 2M2 l3 sin(q3 )q̈3 = 2M2 l3 q̇32 cos(q3 ) − k2 ld + u1 (5.64)

where due to underactuation one has u1 = F1 = 0. Therefore, it holds

ÿ = −k2 ld ⇒ÿ = −k2 (q1 + 2l3 cos(q3 ) − L) (5.65)

Consequently
y(3) = −k2 [q̇1 − 2l3 sin(q3 )q̇3 ] (5.66)

and, respectively

y(4) = −k2 q̈1 + 2k2 l3 cos(q3 )q̇32 + 2k2 l3 sin(q3 )q̈3 (5.67)

From Eq. (12.112) describing the flat output and from the equations of its higher
order derivatives one has a set of equations which can be solved with respect to the
state variables q1 , q3 , q̇1 and q̇3 . It holds that
k2 y−M2 ÿ+M2 k2 L
q1 = k 2 M1 (5.68)

2 q1 −k2 L
q3 = cos−1 ( ÿ+k2k 2 l3
) (5.69)

k2 ẏ+M2 y(3)
q̇1 = k 2 M1
(5.70)

(M1 +M2 )q̇1 −ẏ

q̇3 = 1
sin(q3 ) 2M2 l3 (5.71)

Having expressed the elements of the state vector q as functions of the flat output and
its derivatives and knowing that q̈1 = f3 +ga3 u1 +gb3 u2 and q̈3 = f4 +ga4 u1 +gb4 u2 ,
it holds that the control inputs u1 and u2 can be also written as functions of the flat
output and its derivatives. Therefore, the robotic system stands for a differentially
flat model.

5.2.4.2 Design of a Flatness-Based Controller for the Closed-Chain

2-DOF Robotic System

From the relation q̇ = f (q) + ga (q)u1 + gb (q)u2 and the associated relations about
f (q), ga q and gb (q) it holds
5.2 Adaptive Fuzzy Control of Underactuated MIMO Robots 197

q̈1 = f3 + ga3 u1 + gb3 u2 ⇒q̈1 = f3 + gb3 u2 ⇒

−k2 ld (I3 +I4 )+2l3 M2 [k4 (π −q3 )sin(q3 )+(I3 +I4 )q̇32 cos(q3 )]
q̈1 = M2 (I3 +I4 )+M1 (I3 +I4 +4M2 l32 sin2 (q3 ))
+ (5.72)
2M2 l3 sin(q3 )
+ (I +I )M +M (I +I +4M l2 sin2 (q )) u2
3 4 2 1 3 4 2 3 3

Equivalently, one has

q̈3 = f4 + ga4 u1 + gb4 u2 ⇒q̈3 = f3 + gb4 u2 ⇒

k4 (M1 +M2 )(π −q3 )+2l3 M1 sin(q3 )[K2 ld −2l3 M2 q̇32 cos(q3 )]
q̈3 = M2 (I3 +I4 )+M1 (I3 +I4 +4M2 l32 sin2 (q3 ))
+ (5.73)
1 +M2
+ (I +I )M +M M u2
1 (I3 +I4 +4M2 l3 sin (q3 ))
2 2
3 4 2

Consequently, it holds

y(4) = −k2 q̈1 + 2k2 l3 cos(q3 )q̇32 + 2k2 l3 sin(q3 )q̈3 ⇒

y(4)
= −k2 (f3 + gb3 )u2 + 2k2 l3 cos(q3 )q̇32 + 2k2 l3 sin(q3 )(f4 + gb4 u2 )⇒
(4)
y = 2k2 l3 cos(q3 )q̇32 − k2 f3 + 2k2 l3 sin(q3 )f4 + [−k2 gb3 + 2k2 l3 sin(q3 )gb4 ]u2 ⇒
y(4) = fv + gv u⇒y(4) = v
(5.74)
where v = fv + gv u2 with

fv = 2k2 l3 cos(q3 )q̇32 − k2 f3 + 2k2 l3 sin(q3 )f4

(5.75)
gv = −k2 gb3 + 2k2 l3 sin(q3 )gb4

The following new state variables are defined z1 = y, z2 = ẏ, z3 = ÿ and z4 =

y(3) . For the new state variables, a description of the system in the linear canonical
(Brunovsky) form is obtained
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż1 0 1 0 0 z1 0
⎜ż2 ⎟ ⎜0 0 1 0⎟ ⎜ z2 ⎟ ⎜ 0 ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟v (5.76)
⎝ż3 ⎠ ⎝0 0 0 1 ⎠ ⎝ z3 ⎠ ⎝ 0 ⎠
ż4 0 0 0 0 z4 1

Considering that the linear displacement of the left part of the kinematic chain q1 is
measurable, and that the same holds for the turn angle q3 of joint A one has that the
flat output y = (M1 + M2 )q1 + 2M2 l3 cos(q3 ) is also a measurable variable.
Using the description of system in the linear canonical form, the appropriate
control law is

v = ż4d − k1 (z4 − z4d ) − k2 (z3 − z3d ) − k3 (z2 − z2d ) − k4 (z1 − z1d ) (5.77)
198 5 Differential Flatness Theory and Industrial Robotics

Moreover, using that the transformed control input that is applied to the robot is
v = fv + gv u2 , one can compute the real control input u2 = gv−1 (v − fv ).

5.2.5 Flatness-Based Adaptive Fuzzy Control

for the Underactuated Robot

5.2.5.1 Design of an Indirect Adaptive Controller

After transforming the dynamic model of the robotic mechanism into the linear
canonical form, the design of a state feedback controller becomes possible. To cope
with the lack of knowledge about the model’s parameters adaptive fuzzy control will
be implemented. The design of the adaptive fuzzy control scheme for the robotic
model that is transformed to the canonical (Brunovsky) form has been explained
in Chap. 3. Moreover, the stages of the associated Lyapunov stability analysis have
been given, and it has been explained that the proposed adaptive fuzzy control method
succeeds H∞ tracking performance for the closed-loop.
The unknown dynamics of the robot is modeled with the use of neuro-fuzzy
approximators. These have as inputs the state variables of the robotic model xi , i =
1, . . . , 4. Thus one arrives at the approximation of the unknown dynamics of the
robot fˆ (x) and ĝ(x), which in turn are used for the computation of the control signal
u.

5.2.6 Simulation Tests

The efficiency of the proposed adaptive fuzzy control scheme for the underactuated
robotic manipulator was evaluated in the tracking of various setpoints for state vari-
able q1 i.e., the linear displacement of mass M1 of the mechanism and state variable
q3 , i.e., the rotation angle of joint A. As it can be observed in Figs. 5.2, 5.3 and 5.4,
the proposed control scheme enabled accurate tracking of the reference setpoints
by the robot’s state variables. Indicative results about the estimation of unknown
functions fv (x, t) and gv (x, t) in the robot’s dynamics by the neuro-fuzzy approxi-
mators is shown in Fig. 5.5. By including an additional term in the control loop that
was based on the disturbances estimation, it became possible to compensate for the
disturbances effects.
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 199

(a) 20
x1 − xd1 3 (b) 20 2.5

x2 − xd2

x1 − xd1

x2 − xd2
2.5
10 10 2
2
0 0 1.5
1.5
−10 1 −10 1
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)
6 0.6 6 0.3
x3 − xd3

x4 − xd4

x3 − xd3

x4 − xd4
4 0.4 4 0.2
2 0.2 2 0.1
0 0 0 0
−2 −0.2 −2 −0.1
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)

Fig. 5.2 Convergence of the robot’s state variables: a setpoint 1, b setpoint 2

(a) 15 2.5 (b) 20 3

x1 − xd1
x1 − xd1

x2 − xd2

x2 − xd2
10 2.5
2 10
5 2
1.5 0
0 1.5
−5 1 −10 1
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)
6 0.3 6 0.6
x3 − xd3

x4 − xd4

x3 − xd3

x4 − xd4

4 0.2 4 0.4
2 0.1 2 0.2
0 0 0 0
−2 −0.1 −2 −0.2
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)

Fig. 5.3 Convergence of the robot’s state variables: a to setpoint 3, b to setpoint 4

5.3 Observer-Based Adaptive Fuzzy Control

of MIMO Robots

5.3.1 Overview

The chapter also proposes a solution to the problem of observer-based adaptive

fuzzy control for MIMO nonlinear dynamical systems (e.g., robotic manipulators).
The objective of this section is to design an adaptive fuzzy controller for a class of
nonlinear systems, under the constraint that only the system’s output is measured
and that the system’s model is unknown. The control algorithm aims at satisfying the
H∞ tracking performance criterion, which means that the influence of the modeling
200 5 Differential Flatness Theory and Industrial Robotics

(a) 20
x1 − xd1 3 (b) 30 3

x2 − xd2

x1 − xd1

x2 − xd2
2.5 20 2.5
10
2 10 2
0
1.5 0 1.5
−10 1 −10 1
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)
6 x4 − xd4 0.6 6 0.6

x3 − xd3

x4 − xd4
x3 − xd3

4 0.4 4 0.4
2 0.2 2 0.2
0 0 0 0
−2 −0.2 −2 −0.2
0 20 40 60 0 20 40 60 0 20 40 60 0 20 40 60
time (sec) time (sec) time (sec) time (sec)

Fig. 5.4 Convergence of the robot’s state variables: a to setpoint 5, b to setpoint 6

(a) 100 50
80 45
real vs estimated function g
real vs estimated function f

60 40
40 35
20 30
0 25
−20 20
−40 15
−60 10
−80 5
−100 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
time (sec) time (sec)
(b) 100 50
80 45
real vs estimated function g
real vs estimated function f

60 40
40 35
20 30
0 25
−20 20
−40 15
−60 10
−80 5
−100 0
0 10 20 30 40 50 60 0 10 20 30 40 50 60
time (sec) time (sec)

Fig. 5.5 Estimation of functions fv (x, t) and gv (x, t) of the robot dynamics by neuro-fuzzy networks
when tracking: a setpoint 1, b setpoint 2
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 201

errors and the external disturbances on the tracking error is attenuated to an arbitrary
desirable level. Differential flatness theory is used to transform the MIMO robotic
system into the canonical form. The resulting control inputs are shown to contain
nonlinear elements which depend on the system’s parameters. Since the parameters
of the system are unknown, then the nonlinear terms which appear in the control
inputs have to be approximated with the use of neuro-fuzzy networks. Moreover,
since only the system’s output is measurable the complete state vector has to be
reconstructed with the use of a state observer. In the current section, it is shown that
a suitable learning law can be defined for the aforementioned neuro-fuzzy approxi-
mators so as to preserve the closed-loop system stability. Lyapunov stability analysis
also proves that the proposed observer-based adaptive fuzzy control scheme results
in H∞ tracking performance, in accordance to the results of [282, 407, 410, 413].
For the design of the observer-based adaptive fuzzy controller, one has to solve
two Riccati equations, where the first one is associated with the controller and the
second one is associated with the observer. Parameters that affect the closed-loop
robustness are: (i) The feedback gain vector K, (ii) the observer’s gain vector Ko ,
and (iii) the positive definite matrices P1 and P2 which come from the solution of
the two algebraic Riccati equations and which weigh the above-mentioned observer
and controller terms. The proposed control architecture guarantees that, the output
of the closed-loop system will asymptotically track the desired trajectory and that
H∞ performance will be achieved.

5.3.2 Estimation of the Robot’s State Vector

In the case of a system with non-completely measurable state vector, the structure
and the stability properties of a MIMO adaptive fuzzy controller based on differential
flatness theory have been explained in Sect. 3.4. Now, the problem of controller is
more demanding because one cannot use the complete state vector of the system in the
feedback loop. With the application of differential flatness theory, in Chap. 3 it was
shown that the 2-DOF robotic model can be decomposed into the trivial subsystems
of Eq. (3.47), that is
(r ) p
yi i = fi (x) + j=1 gij (x)uj + dj (5.78)

which can be also written in the vector form

y(r) = f (x) + g(x)u + d (5.79)

As noted, the control of the robotic system already depicted in Fig. 3.7 described by
Eq. (3.70) and by its equivalent canonical form of Eq. (3.49) becomes more compli-
cated when the state vector x is not directly measurable and has to be reconstructed
through a state observer. The following definitions are used
202 5 Differential Flatness Theory and Industrial Robotics

• error of the state vector e = x − xm

• error of the estimated state vector ê = x̂ − xm
• observation error ẽ = e − ê = (x − xm ) − (x̂ − xm )
When an observer is used to reconstruct the state vector, the control law of Eq. (3.68)
is written as
u = ĝ−1 (x̂|θg )[−fˆ (x̂|θf ) + ym
(r)
− K T ê + uc ] (5.80)

Applying Eq. (5.80) to the nonlinear system described by Eq. (5.79), results into

(r)
y(r) = f (x) + g(x)ĝ−1 (x̂)[−fˆ (x̂) + ym − K T ê + uc ] + d⇒
(r)
= f (x) + [g(x) − ĝ(x̂) + ĝ(x̂)]ĝ (x̂)[−fˆ (x̂) + ym − K T ê + uc ] + d⇒
y(r) −1
(r)
y(r) = [f (x) − fˆ (x̂)] + [g(x) − ĝ(x̂)]u + ym − K T ê + uc + d
(5.81)
(r)
It holds that e = x − xm ⇒ y(r) = e(r) + ym . Substituting y(r) in the above equation
gives
(r) (r)
e(r) + ym = ym − K T ê + uc + [f (x) − fˆ (x̂)]+
(5.82)
+[g(x) − ĝ(x̂)]u + d

and equivalently

ė = Ae − BK T ê + Buc + B{[f (x) − fˆ (x̂)]+

(5.83)
+[g(x) − ĝ(x̂)]u + d̃}

e1 = C T e (5.84)

where e = [e1 , e2 , . . . , ep ]T with ei = [ei , ėi , ëi , . . . , eiri −1 ]T , i = 1, 2, . . . , p and

equivalently ê = [ê1 , ê2 , . . . , êp ]T with êi = [êi , ėˆ i , ëˆ i , . . . , êiri −1 ]T , i = 1, 2, . . . , p.
Matrices A, B and C have been defined in Eq. (3.48).
A state observer is designed according to Eqs. (5.83) and (5.84) and is given by
[524]:
ê˙ = Aê − BK T ê + Ko [e1 − C T ê] (5.85)

ê1 = C T ê (5.86)

The feedback gain matrix is denoted as K ∈ Rn×p . The observation gain matrix is
denoted as Ko ∈ Rn×p and its elements are selected so as to assure the asymptotic
elimination of the observation error.
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 203

5.3.3 Application of Flatness-Based Adaptive Fuzzy Control

The dynamic model of the 2-DOF robotic manipulator was first analyzed in Chap. 3.
The considered 2-DOF robotic model consists of two rigid links which are rotated
by DC motors, as shown in Fig. 5.6. It has been proven that, the considered robotic
system is a differentially flat one.
Next, taking into account also the effects of additive disturbances to the joints of
the robotic manipulator the dynamic model becomes

ẍ1 = f1 (x, t) + g1 (x, t)u + d1

(5.87)
ẍ3 = f2 (x, t) + g2 (x, t)u + d2

ẍ1 f1 (x, t) g1 (x, t) d

= + u+ 1 (5.88)
ẍ3 f2 (x, t) g2 (x, t) d2

It was also proven that by applying the following control input

−1  d     

ĝ (x, t) ẍ1 fˆ1 (x, t) K1T u
u= 1 − ˆ − e + c1 (5.89)
ĝ2 (x, t) ẍ3d f2 (x, t) K2T uc2

where [uc1 uc2 ]T is a robust control term that is used for the compensation of the
model’s uncertainties as well as of the external disturbances and using the feedback
gain vector KiT = [k1i , k2i , . . . , kn−1
i , kni ], the closed-loop tracking error dynamics
becomes

Fig. 5.6 Rigid-link 2-DOF robotic manipulator and joints’ actuation with the use of DC motors
204 5 Differential Flatness Theory and Industrial Robotics
 
f1 (x, t) − fˆ1 (x, t)
ė = (A − BK T )e + Buc +B +
f2 (x, t) − fˆ2 (x, t)
(5.90)


g1 (x, t) − ĝ1 (x, t)
+ u + d̃
g2 (x, t) − ĝ2 (x, t)

where, matrices A, B, K T have been defined as

⎛ ⎞ ⎛ ⎞
0100 0 0
⎜0 0 0 0 ⎟ ⎜1 0⎟
A=⎜ ⎟ ⎜
⎝0 0 0 1 ⎠ , B = ⎝0
⎟
0⎠
0000 0 1 (5.91)

K11 K21 K31 K41

KT =
K12 K22 K32 K42

When the estimated state vector x̂ is used in the feedback control loop, equivalently
to Eq. (10.182) one has
 
f1 (x, t) − fˆ1 (x̂, t)
ė = Ae − BK T ê + Buc + B +
f2 (x, t) − fˆ2 (x̂, t)
(5.92)


g1 (x, t) − ĝ1 (x̂, t)
+ u + d̃
g2 (x, t) − ĝ2 (x̂, t)

and considering that the approximation error w is now denoted as

 

f1 (x, t) − fˆ1 (x̂, t) g1 (x, t) − ĝ1 (x̂, t)

w= + u (5.93)
f2 (x, t) − fˆ2 (x̂, t) g2 (x, t) − ĝ2 (x̂, t)

Equation (5.92) can be also written as

ė = Ae − BK T ê + Buc + Bw + Bd̃ (5.94)

The associated state observer will be described again by Eqs. (5.85) and (5.86).

5.3.4 Dynamics of the Observation Error

For the transformed state vector of the robot, the observation error is defined as
ẽ = e − ê = x − x̂. Substructing Eq. (5.85) from Eq. (5.83) as well as Eq. (5.86) from
Eq. (5.84) one gets
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 205

ė − ê˙ = A(e − ê) + Buc + B{[f (x, t) − fˆ (x̂, t)]+

+[g(x, t) − ĝ(x̂, t)]u + d̃} − Ko C T (e − ê)

e1 − ê1 = C T (e − ê)

or equivalently

ẽ˙ = Aẽ + Buc + B{[f (x, t) − fˆ (x̂, t)] + [g(x, t) − ĝ(x̂, t)]u + d̃} − Ko C T ẽ

ẽ1 = C T ẽ

which can be written as

ẽ˙ = (A − Ko C T )ẽ + Buc + B{[f (x, t) − fˆ (x̂, t)] + [g(x, t) − ĝ(x̂, t)]u + d̃}
(5.95)

ẽ1 = C T ẽ (5.96)

or equivalently, it can be written as

ẽ˙ = (A − Ko C T )ẽ + Buc + Bw + d̃ (5.97)

ẽ1 = C T ẽ (5.98)

5.3.5 Approximation of the System’s Unknown Dynamics

Next, the following approximators of the system’s unknown dynamics are defined
 
ˆ 4×1 fˆ (x̂|θ ) ∈ R1×1
ˆf (x̂) = f1 (x̂|θf ) x̂ ∈ R 1 f
(5.99)
fˆ2 (x̂|θf ) x̂ ∈ R4×1 fˆ2 (x̂|θf ) ∈ R1×1

with kernel functions n

j=1 μAj (x̂j )
i
i,j
φf (x̂) = N n (5.100)
i=1 j=1 μA (x̂j )
i
j

where l = 1, 2, x̂ is the estimate of the state vector and μAi (x̂) is the ith membership
j
function of the antecedent (IF) part of the lth fuzzy rule. Similarly, the following
approximators of the unknown system dynamics are defined with the use of the
estimated state vector x̂

ĝ1 (x̂|θg ) x̂ ∈ R4×1 ĝ1 (x̂|θg ) ∈ R1×2

ĝ(x̂) = (5.101)
ĝ2 (x̂|θg ) x̂ ∈ R4×1 ĝ2 (x̂|θg ) ∈ R1×2
206 5 Differential Flatness Theory and Industrial Robotics

The values of the weights that result in optimal approximation are

θf∗ = arg minθf ∈Mθf [supx̂∈Ux̂ (f (x) − fˆ (x̂|θf ))]

(5.102)
θg∗ = arg minθg ∈Mθg [supx̂∈Ux̂ (g(x) − ĝ(x̂|θg ))]

where the variation ranges for the weights are defined as

Mθf = {θf ∈ Rh : ||θf ||≤mθf }

(5.103)
Mθg = {θg ∈ Rh : ||θg ||≤mθg }

Using Eq. (5.93), the value of the approximation error that corresponds to the optimal
values of the weights vectors θf∗ and θg∗ one has
  
w = f (x, t) − fˆ (x̂|θf∗ ) + g(x, t) − ĝ(x̂|θg∗ ) u (5.104)

which is next written as

 
w = f (x, t) − fˆ (x̂|θf ) + fˆ (x̂|θf ) − fˆ (x̂|θf∗ ) +
 (5.105)
+ g(x, t) − ĝ(x̂|θg ) + ĝ(x̂|θg ) − ĝ(x̂|θg∗ ) u

which can be also written in the following form


w = wa + wb (5.106)

where
wa = {[f (x, t) − fˆ (x̂|θf )] + [g(x, t) − ĝ(x̂|θg )]}u (5.107)

wb = {[fˆ (x̂|θf ) − fˆ (x̂|θf∗ )] + [ĝ(x̂, θg ) − ĝ(x̂|θg∗ )]}u (5.108)

Moreover, the following weights error vectors are defined

θ̃f = θf − θf∗
(5.109)
θ̃g = θg − θg∗

5.3.6 Lyapunov Stability Analysis

5.3.6.1 Design of the Lyapunov Function

In the case of observer-based adaptive fuzzy control with the use of differential
flatness theory, the adaptation law of the neuro-fuzzy approximators weights θf
and θg as well as the equation of the supervisory control term uc are derived from
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 207

the requirement for negative definiteness of a suitably chosen Lyapunov function.

Extending the results presented in Sect. 3.4, the Lyapunov function is now defined
as
V = 21 êT P1 ê + 21 ẽT P2 ẽ + 2γ11 θ̃fT θ̃f + 2γ12 tr[θ̃gT θ̃g ] (5.110)

The selection of the Lyapunov function is based on the following principle of indirect
adaptive control ê : limt→∞ x̂(t) = xd (t) and ẽ : limt→∞ x̂(t) = x(t). This yields
limt→∞ x(t) = xd (t). Substituting Eqs. (5.85), (5.86) and (5.95), (5.96) into Eq.
(5.110) and differentiating results into

V̇ = 21 ê˙ T P1 ê + 21 êT P1 ê˙ + 21 ẽ˙ T P2 ẽ + 21 ẽT P2 ẽ+

˙
(5.111)
+ γ11 θ̃˙fT θ̃f + ˙T
γ2 tr[θ̃ g θ̃g ]
1
⇒

V̇ = 21 {(A − BK T )ê + Ko C T ẽ}T P1 ê + 21 êT P1 {(A − BK T )ê + Ko C T ẽ}+

+ 21 {(A − Ko C T )ẽ + Buc + Bd̃ + Bw}T P2 ẽ+
(5.112)
+ 21 ẽT P2 {(A − Ko C T )ẽ + Buc + Bd̃ + Bw}+

+ γ11 θ̃˙fT θ̃f + ˙T

γ2 tr[θ̃ g θ̃g ]
1
⇒

V̇ = 21 {êT (A − BK T )T + ẽT CKoT }P1 ê + 21 êT P1 {(A − BK T )ê + Ko C T ẽ}+

+ 21 {ẽT (A − Ko C T )T + ucT BT + wT BT + d̃ T BT }P2 ẽ+ (5.113)
1 ˙T ˙T
2 ẽ P2 {(A − Ko C )ẽ + Buc + Bw + Bd̃} + γ1 θ̃f θ̃f + γ2 tr[θ̃ g θ̃g ] ⇒
1 T T 1

V̇ = 21 êT (A − BK T )T P1 ê + 21 ẽT CKoT P1 ê+

+ 21 êT P1 (A − BK T )ê + 21 êT P1 Ko C T ẽ+
+ 21 ẽT (A − Ko C T )T P2 ẽ + 21 (ucT + wT + d̃ T )BT P2 ẽ+ (5.114)
+ 21 ẽT P2 (A − Ko C T )ẽ + 21 ẽT P2 B(uc + w + d̃)+

+ γ11 θ̃˙fT θ̃f + ˙T

γ2 tr[θ̃ g θ̃g ]
1

Assumption 1: For given positive definite matrices Q1 and Q2 there exist positive
definite matrices P1 and P2 , which are the solution of the following Riccati equa-
tions [524]
(A − BK T )T P1 + P1 (A − BK T ) + Q1 = 0 (5.115)
T
(A − Ko C T ) P2 + P2 (A − Ko C T )−
(5.116)
−P2 B( 2r − 1
ρ2
)BT P2 + Q2 = 0

The conditions given in Eqs. (5.115) and (5.116) are related to the requirement that the
systems described by Eqs. (5.85), (5.86) and (5.95), (5.96) have stable eigenvalues.
Substituting Eqs. (5.115) and (5.116) into V̇ yields
208 5 Differential Flatness Theory and Industrial Robotics

V̇ = 21 êT {(A − BK T )T P1 + P1 (A − BK T )}ê + ẽT CKoT P1 ê+

+ 21 ẽT {(A − Ko C T )T P2 + P2 (A − Ko C T )}ẽ+ (5.117)
1 ˙T ˙T
+ẽT P2 B(uc + w + d̃) + γ1 θ̃f θ̃f + γ2 tr[θ̃ g θ̃g ]
1

that is

V̇ = − 21 êT Q1 ê + ẽT CKoT P1 ê − 21 ẽT {Q2 − P2 B( 2r − 1

ρ2
)BT P2 }ẽ+
(5.118)
1 ˙T ˙T
+ẽT P2 B(uc + w + d̃) + γ1 θ̃f θ̃f + γ2 tr[θ̃ g θ̃g ]
1

The supervisory control uc is decomposed in two terms, ua and ub .

• The control term ua is given by

1
ua = − ẽT P2 B + Δua (5.119)
r

where assuming that the measurable elements of vector ẽ are {ẽ1 , e˜3 , . . . , e˜k }, the
term Δua is such that
⎛ ⎞
p11 ẽ1 + p13 ẽ3 + · · · + p1k ẽk
⎜p13 ẽ1 + p33 ẽ3 + · · · + p3k ẽk ⎟
− 1r ẽT P2 B + Δua = − 1r ⎜
⎝
⎟
⎠ (5.120)
··· ······
p1k ẽ1 + p3k ẽ3 + · · · + pkk ẽk

• The control term ub is given by

ub = −[(P2 B)T (P2 B)]−1 (P2 B)T CKoT P1 ê (5.121)

• ua is an H∞ control used for the compensation of the approximation error w

and the additive disturbance d̃. Its first component − 1r ẽT P2 B has been chosen so
as to compensate for the term 1r ẽT P2 BBT P2 ẽ, which appears in Eq. (5.118). By
including also the second component Δua , one has that ua is computed based
only on the feedback of the measurable variables {ẽ1 , e˜3 , · · · , e˜k }, out of the
complete vector ẽ = [ẽ1 , ẽ2 , . . . , ẽn ]. According to Eq. (5.119) ua is written as
ua = − 1r ẽT P2 B + Δua .
• ub is a control used for the compensation of the observation error (the control term
ub has been chosen so as to satisfy the condition ẽT P2 Bub = −ẽT CKoT P1 ê.
The control scheme is depicted in Fig. 5.7.
Substituting Eqs. (5.119) and (5.121) in V̇ and assuming that Eqs. (5.115) and
(5.116) hold, one gets
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 209

Fig. 5.7 The proposed H∞ control scheme

V̇ = − 21 êT Q1 ê + ẽT CKoT P1 ê − 21 ẽT Q2 ẽ + 1r ẽT P2 BBT P2 ẽ − 1 T

2ρ 2
ẽ P2 BBT P2 ẽ+
1 ˙T ˙T
+ẽT P2 Bua + ẽT P2 Bub + ẽT P2 B(w + d̃) + γ1 θ̃f θ̃f + γ2 tr[θ̃ g θ̃g ]
1

(5.122)

or equivalently,

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ+
(5.123)
1 ˙T ˙T
+ẽT P2 B(w + d̃ + Δua ) + γ1 θ̃f θ̃f + γ2 tr[θ̃ g θ̃g ]
1

It holds that θ̃˙f = θ̇f − θ˙f∗ = θ˙f and θ̃˙g = θ̇g − θ˙g∗ = θ˙g . The following weight
adaptation laws are considered:

θ̇f = −γ1 Φ(x̂)T BT P2 ẽ

(5.124)
θ̇g = −γ2 Φ(x̂)T BT P2 ẽuT

where assuming N fuzzy rules and associated kernel functions, the matrices dimen-
sions are θf ∈ RN×1 , θg ∈ RN×2 , Φ(x) ∈ R2×N , B ∈ R4×2 , P ∈ R4×4 and ẽ ∈ R4×1 .
210 5 Differential Flatness Theory and Industrial Robotics

The update of θf is a gradient type algorithm. The update of θg is also a gradi-

ent type algorithm, where uc implicitly tunes the adaptation gain γ2 [22, 31, 414].
Substituting Eq. (5.124) in V̇ gives

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + BT P2 ẽ(w + d + Δua )+
+ γ11 (−γ1 )ẽT P2 BΦ(x̂)(θf − θf∗ )+ (5.125)
+ γ12 (−γ2 )tr[uẽT P2 BΦ(x̂)(θg − θg∗ )]

or

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + BT P2 ẽ(w + d̃ + Δua )+
+ γ11 (−γ1 )ẽT P2 BΦ(x̂)(θf − θf∗ )+ (5.126)
+ γ12 (−γ2 )tr[uẽT P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))]

Taking into account that u ∈ R2×1 and ẽT PB(ĝ(x|θg ) − ĝ(x|θg∗ )) ∈ R1×2 it holds

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + BT P2 ẽ(w + d̃ + Δua )+
+ γ11 (−γ1 )ẽT P2 BΦ(x̂)(θf − θf∗ )+ (5.127)
+ γ12 (−γ2 )tr[ẽT P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))u]

Since ẽT P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))u ∈ R1×1 it holds

tr(ẽT P2 B(ĝ(x|θg ) − ĝ(x|θg∗ )u) =

(5.128)
= ẽT P2 B(ĝ(x|θg ) − ĝ(x|θg∗ ))u

Therefore, one finally obtains

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + BT P2 ẽ(w + d̃ + Δua )+
+ ∗
γ1 (−γ1 )ẽ P2 BΦ(x̂)(θf − θf )+
1 T (5.129)
+ ∗
γ2 (−γ2 )ẽ P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg ))u
1 T

Next, the following approximation error is defined

wα = [fˆ (x̂|θf∗ ) − fˆ (x̂|θf )] + [ĝ(x̂|θg∗ ) − ĝ(x̂|θg )]u (5.130)

Thus, one obtains

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ +
(5.131)
+ BT P2 ẽ(w + d̃) + ẽT P2 Bwα
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 211

Denoting the aggregate approximation error and disturbances vector as

w1 = w + d̃ + wα + Δua (5.132)

the derivative of the Lyapunov function becomes

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + ẽT P2 Bw1 (5.133)

which in turn is written as

V̇ = − 21 êT Q1 ê − 21 ẽT Q2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ +
(5.134)
+ 1 T
2 ẽ PBw1 + 1 T T
2 w1 B P2 ẽ

Lemma: The following inequality holds

1 T
2 ẽ P2 Bw1 + 21 w1T BT P2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ
(5.135)
≤ 21 ρ 2 w1T w1

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ 2 a2 + 1 2
ρ2
b − 2ab ≥ 0 ⇒

2 ρ a + 2ρ 2 b − ab ≥ 0 ⇒
1 2 2 1 2
(5.136)
ab − 2ρ1 2 b2 ≤ 21 ρ 2 a2 ⇒

2 ab + 2 ab − 2ρ 2 b ≤ 2 ρ a
1 1 1 2 1 2 2

The following substitutions are carried out: a = w1 and b = ẽT P2 B and the previous
relation becomes

2 w1 B P2 ẽ + 2 ẽ P2 Bw1 − 2ρ 2 ẽ P2 BB P2 ẽ
1 T T 1 T 1 T T
(5.137)
≤ 21 ρ 2 w1T w1

The above relation is used in V̇ , and the right part of the associated inequality is
enforced
1 1 1
V̇ ≤ − êT Q1 ê − ẽT Q2 ẽ + ρ 2 w1T w1 (5.138)
2 2 2
Thus, Eq. (5.138) can be written as

1 1
V̇ ≤ − E T QE + ρ 2 w1T w1 (5.139)
2 2
212 5 Differential Flatness Theory and Industrial Robotics

where

ê Q1 0
E= , Q= = diag[Q1 , Q2 ] (5.140)
ẽ 0 Q2

As explained in Sect. 5.3.4, a sufficiently small value for the attenuation coefficient
ρ assures that the right part of Eq. (5.139) will be upped bounded by 0.
Hence, the H∞ performance criterion is derived. For ρ sufficiently small Eq. (5.138)
will be true and the H∞ tracking criterion will be satisfied. In that case, the integration
of V̇ from 0 to T gives
T T T
0 ||E||2 dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
V̇ (t)dt ≤ − 21 0
T T
2V (T ) − 2V (0) ≤ − 0 ||E||2Q dt + ρ 2 0 ||w1 ||2 dt ⇒ (5.141)
T T
2V (T ) + 0 ||E||2Q dt ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt
∞
It is assumed that there exists a positive constant Mw > 0 such that ||w1 ||2 dt ≤
T 0
Mw . Therefore, for the integral 0 ||E||2Q dt one gets
 ∞
||E||2Q dt ≤ 2V (0) + ρ 2 Mw (5.142)
0
∞
Thus, the integral 0 ||E||2Q dt is bounded and according to Barbalat’s Lemma

limt→∞ E(t) = 0 ⇒
limt→∞ ê(t) = 0 (5.143)
limt→∞ ẽ(t) = 0

Therefore limt→∞ e(t) = 0.

5.3.7 The Role of Riccati Equation Coefficients

in Observer-Based Adaptive Fuzzy Control

The linear system of Eqs. (5.95) and (5.96) is considered again

ẽ˙ = (A − Ko C T )ẽ + Buc + B{[f (x, t) − fˆ (x̂, t)] + [g(x, t) − ĝ(x̂, t)]u + d̃}

e1 = C T ẽ

The aim of H∞ control is to eliminate the impact of the modeling errors w =

[f (x, t) − fˆ (x̂, t)] + [g(x, t) − ĝ(x̂, t)]u and the external disturbances d̃ which are
not white noise signals. This implies the minimization of the quadratic cost function
[132, 259, 316]:
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 213

1 T T
J(t) = 2 0 ẽ (t)ẽ(t) + ruc (t)uc (t)−
T
(5.144)
−ρ 2 (w + d̃)T (w + d̃)dt, r, ρ > 0

The weight r determines how much the control signal should be penalized and the
weight ρ determines how much the disturbances influence should be rewarded in the
sense of a mini-max differential game. The control input uc has been defined as the
sum of the terms described in Eqs. (5.119) and (5.121).
Parameter ρ in Eq. (5.144), is an indication of the closed-loop system robustness.
If the values of ρ > 0 are excessively decreased with respect to r, then the solution
of the Riccati equation is no longer a positive definite matrix. Consequently, there
is a lower bound ρmin of ρ for which the H∞ control problem has a solution. The
acceptable values of ρ lie in the interval [ρmin , ∞). If ρmin is found and used in
the design of the H∞ controller, then the closed-loop system will have increased
robustness. Unlike this, if a value ρ > ρmin is used, then an admissible stabilizing
H∞ controller will be derived but it will be a suboptimal one. The Hamiltonian matrix
 
A − Ko C T −( 2r − ρ12 )BBT
H= (5.145)
−Q −(A − Ko C T )T

provides a criterion for the existence of a solution of the Riccati equation (5.116). A
necessary condition for the solution of the algebraic Riccati equation to be a positive
semidefinite symmetric matrix is that H has no imaginary eigenvalues [132].
It is noted that several methods on adaptive neural/fuzzy control of nonlinear
dynamical systems assumes that the system is already described in the canonical
Brunovsky form. However, for the majority of electromechanical systems this does
not hold. For example, robotic systems, UAVs (unmanned aerial vehicles), land
vehicle and ship models, electric power generators of the PMSG and DFIG type,
electric motors of the DC or induction type, and power electronics models such as
voltage converters are some examples of dynamical systems which are not inherently
found in the canonical form [427]. Adaptive fuzzy control methods usually try to
invert the system’s dynamics, and thus to succeed convergence of its output to the
desirable setpoints, starting from a description of the system in the canonical form.
This means that the system is a-priori taken to be in the description given in Eq. (5.78)
or (5.79); however, in the majority of electromechanical systems, the dynamics is in
the form of Eq. (2.4) or (3.44). Differential flatness theory enables to transform the
system’s description ẋ = f (x, u) into that of Eq. (5.78) or (5.79) and from that point
on to develop adaptive control schemes. Consequently, differential flatness theory
extends the class of nonlinear control systems to which adaptive neural/fuzzy control
can be applied and this is a significant benefit for adaptive control theory. Additional
results about adaptive fuzzy controllers which are based on transformations into
canonical forms can be found in [396, 581, 591].
214 5 Differential Flatness Theory and Industrial Robotics

5.3.8 Simulation Tests

The performance of the proposed observer-based adaptive fuzzy MIMO controller

was tested in the benchmark problem of the 2-DOF rigid-link robotic manipulator.
The differentially flat model of the robot and its transformation to the Brunovksy
form has been analyzed in Sect. 4.2.5.
The state feedback gain was K ∈ R2×4 . The basis functions used in the estimation
x̂−cj 2
of fi (x̂, t), i = 1, 2 and gij (x̂, t), i = 1, 2, j = 1, 2 were μAj (x̂) = e( σ ) , j =
1, . . . , 3. Since there are four inputs x̂1 , x̂˙ 1 and x̂3 , x̂˙ 3 and each one of them consists
of 3 fuzzy sets, for the approximation of functions fi (x̂, t) i = 1, 3, there will be 81
fuzzy rules of the form:

Rl : IF x̂1 is Al1 AND x̂˙ 1 is Al2

(5.146)
AND x̂3 is Al3 AND x̂˙ 3 is Al4 THEN fˆil is bl

The

aggregate

output of the neuro-fuzzy approximator (rule base) is fˆi (x̂, t) =
81 ˆ l 4
i=1 μAi (x̂i )
l
l=1 fi (l)
81 4 . The centers ci , i = 1, . . . , 4 and the variances v(l) of each rule
l=1 i=1 μA (x̂i )
l
i
are as follows are summarized in Table 5.1.
The estimation of the control input gain functions ĝij (x̂, t) i = 1, 2 was derived
in a similar way. The sampling period was taken to be 0.01 s. In the beginning of
the training of the neuro-fuzzy approximators, their weights were initialized to zero.
Moreover, the elements of the robot’s state vector were also initialized to zero. The
positive definite matrices P1 ∈ R4×4 and P2 ∈ R4×4 stem from the solution of the
algebraic Riccati equations (5.115) and (5.116), for Q1 and Q2 also positive definite.
The approximations fˆ and ĝ were used in the derivation of the control law, given
by Eq. (5.80). To show the disturbance rejection capability of the proposed adaptive
fuzzy controller, at the beginning of the second half of the simulation time additive
sinusoidal disturbances of amplitude A = 0.5 and period T = 10 s were applied to
the robot’s joints.

Table 5.1 Parameters of the (l) (l) (l) (l)

Rule c1 c2 c3 c4 v(l)
fuzzy rule base
R(1) −1.0 −1.0 −1.0 −1.0 3
R(2) −1.0 −1.0 −1.0 0.0 3
R(3) −1.0 −1.0 −1.0 1.0 3
R(4) −1.0 −1.0 0.0 −1.0 3
R(5) −1.0 −1.0 0.0 0.0 3
R(6) −1.0 −1.0 0.0 1.0 3
··· ··· ··· ··· ··· ···
··· ··· ··· ··· ··· ···
R(81) 1.0 1.0 1.0 1.0 3
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 215

(a) 2 (b) 5

state variable x 2 of the robot

1.5
state variable x 1 of the robot

0
1

0.5 −5
0

−0.5 −10

−1
−15
−1.5

−2 −20
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 5.8 a Tracking of a sinusoidal position setpoint by joint 1 of the robot, b Tracking of sinusoidal
velocity setpoint by joint 1 of the robot

(a) 2 (b) 5

1.5
state variable x of the robot

state variable x of the robot

0
1
−5
0.5
3

0 −10

−0.5
−15
−1
−20
−1.5

−2 −25
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 5.9 a Tracking of a sinusoidal position setpoint by joint 2 of the robot, b Tracking of sinusoidal
velocity setpoint by joint 2 of the robot

In the simulation results that follow the position and velocity setpoints are noted
as continuous line while the position and velocity signals of the robot’s joints are
denoted as dashed lines. The position variation for the first joint of the robotic manip-
ulator when tracking a sinusoidal setpoint is depicted in Fig. 5.8a, while the velocity
variation is shown in Fig. 5.8b. For the second joint of the 2-DOF robot the tracking
of the position setpoint is depicted in Fig. 5.9a while the tracking of the velocity
setpoint is shown in Fig. 5.9b. The control inputs (motor torques) applied to the first
and second joint of the robotic manipulator are shown in Fig. 5.10a, b, respectively.
The performance of the proposed observer-based adaptive fuzzy control scheme
was also tested in the case of various other setpoints. The associated results, about the
216 5 Differential Flatness Theory and Industrial Robotics

(a) 300 (b) 300

200 200
control signal of joint 1

control signal of joint 2

100 100

0 0

−100 −100

−200 −200

−300 −300
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t (sec) t (sec)

Fig. 5.10 Tracking of a sinusoidal setpoint: a Control input to joint 1 of the robot, b control input
to joint 2 of the robot

(a) (b)
state variable x2

state variable x1

state variable x2
state variable x1

8 100 2 10
6 50 1 5
4 0 0 0
2 −50 −1 −5
0 −100 −2 −10
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) t (sec) t (sec)
state variable x3

state variable x4

state variable x3

state variable x4

8 100 2 10
6 50 1 5
4 0 0 0
2 −50 −1 −5
0 −100 −2 −10
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) t (sec) t (sec)

Fig. 5.11 Convergence of state vector elements of the robot to the desirable setpoints: a When
tracking a piecewise constant setpoint, b when tracking a sinusoidal setpoint

accuracy of tracking of the state vector elements, are shown in Figs. 5.11 and 5.12.
The control loop enabled fast convergence of the state variables to the desirable
setpoints although there was no previous knowledge about the dynamic model of the
robotic manipulator and despite the effects of external disturbances.
The RMSE (root mean square error) of the examined control loop is also calcu-
lated (assuming the same parameters of the controller) in the case of tracking of the
previous setpoints: (a) The piecewise constant setpoint of Fig. 5.11a, (b) the sinu-
soidal setpoint of Fig. 5.11b, (c) the seesaw setpoint of Fig. 5.12a, (d) the setpoint
obtained from the sum of different sinusoidal signals as shown in Fig. 5.12b. The
5.3 Observer-Based Adaptive Fuzzy Control of MIMO Robots 217

(a) (b)

state variable x2

state variable x1

state variable x2
state variable x1
2 10 4 20
1 5 2
0
0 0 0
−20
−1 −5 −2
−2 −10 −4 −40
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) t (sec) t (sec)
state variable x3

state variable x4

state variable x3

state variable x4
2 10 5 40
1 5
20
0 0 0
0
−1 −5
−2 −10 −5 −20
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) t (sec) t (sec)

Fig. 5.12 Convergence of state vector elements of the robot to the desirable setpoints: a When
tracking a seesaw setpoint, b when tracking a setpoint obtained from the sum of different sinusoidal
signals

Table 5.2 RMSE of joints’ Parameter θ1 θ2

angles
RMSEa 0.0471 0.0449
RMSEb 0.0418 0.0427
RMSEc 0.0495 0.0288
RMSEd 0.0449 0.0472

results are summarized in Table 5.2. It can be seen that the transient characteristics
of the control scheme are also quite satisfactory.
The definition of the parameters of the fuzzy rule base (81 rules) has been given
in Table 5.1. For the 2-DOF robotic model considered as application example, the
number of the rules is computed as follows: There are four input variables in the
antecedent part of the each rule, namely position and angular velocity for each one
of the robot’s joints. Each input variable is partitioned into three fuzzy sets, therefore
taking into account all possible combinations of the fuzzy sets to which the input
variables can be assigned one has 34 = 81 fuzzy rules. The placement of the centers
of the fuzzy sets at specific points of the four-dimensional state-space is performed
following the so-called grid partitioning. This means that the centers of the fuzzy sets
are chosen to occupy equally distant positions along the axes of the input variables.
This results in a uniform coverage of the state-space.
Through the presented simulation experiments, it has been shown that the pro-
posed control method succeeds excellent accuracy in the tracking of the reference
setpoints. This is a confirmation of the stability analysis that resulted in an H-infinity
tracking performance criterion and in an asymptotic stability condition. Moreover, it
has been shown that the control input exhibits smooth variations. This is an advantage
comparing, for instance, to adaptive fuzzy sliding mode control since in the latter
218 5 Differential Flatness Theory and Industrial Robotics

case there are abrupt control input variations. What is also important to compare
is the extent of the class of systems to which the various control methods can be
applied and the constraining assumptions made about the structure of the controlled
systems. From this point of view, the proposed Differential flatness theory approach
to observer-based adaptive fuzzy control of MIMO nonlinear dynamical systems
outperforms all other adaptive neural/fuzzy control methods. The method can be
applied to any type of MIMO differentially flat systems that admit a transformation
to the canonical Brunovsky form and does not require the system’s dynamic model
to have a particular structure (e.g., affine-in-the-input, triangular, or other). Besides,
the control method is based exclusively on feedback of the outputs of the system
and does not require measurements of all elements of the state vector. Performing
estimation of the state vector instead or precise measurements of the state variables
has significant advantages. It is less costly because it requires less sensors to be used
in the control loop. It reduces faulty conditions for the control loop because it is free
of sensors that frequently undergo failures when used in harsh operating conditions
of electromechanical systems (such as robots, autonomous vehicles, electric motors,
electric power generators, power electronics, aerial vehicles, vessels, or other).

5.4 State Estimation-Based Control

of Underactuated Robots

5.4.1 Overview

The problem of control of underactuated robotic manipulators that was examined

in Sect. 5.2 is now revisited. Unlike robotic manipulators with a free-end effector,
machines with closed-chains include actuators which are usually placed on an fixed
base. This enables to develop mechatronic systems with low moving inertia and fast
motion control. The particular problem studied in this chapter is the controller design
for a closed-chain mechatronic system, characterized by model uncertainty, and sub-
jected to external perturbations. Moroever, the control system is an underactuated
one, which means that the number of actuators is less than the number of degrees
of freedom in the associated dynamic model. Typical cases of underactuated sys-
tems are: (1) Mobile manipulators carrying out cooperative tasks and (2) fixed-base
robots that manipulate elastically deformable objects. Underactuated closed-chain
robotic systems, containing fewer actuators than the degrees of freedom have little
been studied. This is mainly due to high-nonlinear nature of the associated dynamic
model and the complexity in the controller’s design.
In this section, an approach to solve the problem of closed-chain underactuated
robots is developed with the use of flatness-based controller and a Kalman Filter-
based disturbances estimator (the latter being based on the Derivative-free non-
linear Kalman Filter). Dynamic analysis for the closed-chain underactuated robot
is first provided. Closed-chain robotic models can be underactuated and the effi-
cient suppression of disturbance inputs is important for attaining the performance
5.4 State Estimation-Based Control of Underactuated Robots 219

objectives of the control loop. The elements of the state vector are variables denoting
the linear and rotation displacement and rotation of the robot’s joints. The control
inputs to the model is the torque applied by an actuator to a joint that is mounted on its
basis. Disturbance inputs can be due to model uncertainties or external perturbation
forces and torques.
In Sect. 5.2 it has been shown that the underactuated model of the robotic manip-
ulator can be transformed into the linear canonical (Brunovsky) form. For the lat-
ter model, it is possible to design a state feedback controller that enables accurate
tracking of position and velocity setpoints by the robot’s joints. However, since mea-
surements are available only for certain elements of the transformed state vector, to
implement a state feedback control loop, the rest of the elements of the robot’s trans-
formed state vector have to be estimated with the use of an observer or filter. To this
end, the Derivative-free nonlinear Kalman Filter is used. By avoiding linearization
approximations, the proposed filtering method improves the accuracy of estimation
of the system’s state variables [436, 437, 448].
A particular difficulty, in the case of state estimation for such a robotic model is
the existence of the unmodeled disturbance forces. It is shown that it is possible to
redesign the Kalman Filter in the form of a disturbance observer and using the esti-
mation of the disturbance to develop an auxiliary control input that compensates for
the disturbances effects [85, 107, 108, 354]. In this way, the robot’s control system
can become robust with respect to uncertainties in the model’s parameters or uncer-
tainties about external forces and torques. It is also noted that in terms of computation
speed, the proposed Kalman Filter-based disturbance estimator is faster than distur-
bance estimators that may be based on other nonlinear filtering approaches (e.g.,
Extended Kalman Filter, Unscented Kalman Filter, or Particle Filter) thus becoming
advantageous for the real-time estimation of the unknown dynamics of the robot. The
efficiency of the proposed control and Kalman Filter-based disturbances estimation
scheme is evaluated through numerical simulation tests. It has been shown that the
accurate estimation of the disturbance forces which are exerted on the robot enables
their efficient compensation. This is succeeded by introducing an additional element
in the controller that produces a counter-disturbance input based on the estimated
value for the disturbance variable. This control scheme finally results in minimizing
the effects of the disturbances that affect the robot.

5.4.2 Derivative-Free Nonlinear Kalman Filter

for the Closed-Chain 2-DOF Robotic System

5.4.2.1 State Estimation with the Derivative-Free Nonlinear

Kalman Filter

Previous results about state estimation through transformation to linear canonical

forms can be found in [436, 437, 448]. In Sect. 5.2, it was shown that the dynamical
220 5 Differential Flatness Theory and Industrial Robotics

model of the robot can be written in the MIMO canonical form of Eq. (5.76). Thus
one has a MIMO linear model of the form

ẏf = Af yf + Bf v
(5.147)
zf = Cf yf

where yf ∈ R4×1 and matrices Af , Bf , Cf are in the form

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 1
⎜0 0 1 0⎟ ⎜0⎟ T ⎜0⎟
Af = ⎜
⎝0
⎟B =⎜ ⎟C =⎜ ⎟ (5.148)
0 0 1⎠ f ⎝0 ⎠ f ⎝0⎠
0 0 0 0 1 0

where, as explained in Sect. 5.2, variable y = z1 is associated with the linear displace-
ment and the rotational motion of joint A in the robotic model. For the aforementioned
model, and after carrying out discretization of matrices Af , Bf , and Cf with common
discretization methods one can apply linear Kalman filtering. This is Derivative-free
nonlinear Kalman filtering for the model of the robot which is performed without
the need to compute Jacobian matrices and does not introduce numerical errors due
to approximative linearization with Taylor series expansion.

5.4.2.2 Kalman Filter-Based Estimation of Robot Disturbance Forces

Disturbances affecting the robot’s model can be due to: (i) Uncertainty and changes
in model parameters, (ii) unknown external torques exerted on the robot’s joints,
(iii) unknown forces exerted on the masses of the robotic mechanism (e.g., friction).
Considering the effects of disturbances on the robotic model and after applying a
transform on the system’s state variables according to the differential flatness theory
it has been shown that the robot model of Eq. (5.147) is described by
(4)
z1 = v + φ(τ ) (5.149)

The robot’s state-space model of Eq. (5.76) will be extended to take into account also
the dynamics and the effects of the disturbance input φ(t). The extended state vector
of the robot model is defined as z ∈ R8×1 with z1 = y, z2 = ẏ, z3 = ÿ, z4 = y(3) ,
z5 = φ, z6 = φ̇, z7 = φ̈, z8 = φ (3) . The dynamics of the disturbance is assumed to
be defined by its fourth order derivative, i.e., φ (4) = fd (y, ẏ, ÿ, y(3) ). Thus one has
the extended state-space model

ż = Ã · z + B̃ · ṽ
(5.150)
q = C̃z
5.4 State Estimation-Based Control of Underactuated Robots 221

with
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 0 0 0 1
⎜0 0 1 0 0 0 0 0⎟ ⎜0 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 1 0 0 0⎟ ⎜ 0⎟ ⎜ ⎟
à = ⎜ ⎟ , B̃ = ⎜1 ⎟ , C̃ T = ⎜0⎟ (5.151)
⎜0 0 0 0 0 1 0 0⎟ ⎜0 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 1 0⎟ ⎜0 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 1⎠ ⎝0 0⎠ ⎝0⎠
0 0 0 0 0 0 0 0 0 1 0

where the measurable variable is z1 and the control input is

T
ṽ = u, φ (4) (5.152)

The disturbance estimator has the following structure

ẑ˙ = Ão · ẑ + B̃o · ṽ + K(z1 − ẑ1 )

(5.153)
q̂ = C̃o ŷ

where the estimator’s gain K ∈ R8×1 and matrices Ão , B̃o and C̃o are defined as
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 0 00 1
⎜0 0 1 0 0 0 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 1 0 0 0⎟⎟ ⎜ ⎟
1 0⎟ T ⎜ ⎟
Ão = ⎜ , B̃ = ⎜ , C̃ = ⎜0 ⎟ (5.154)
⎜0 0 0 0 0 1 0 0⎟⎟ o ⎜ ⎟ o ⎜ ⎟
⎜ ⎜0 0⎟ ⎜0 ⎟
⎜0 0 0 0 0 0 1 0⎟⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎜0 0⎟ ⎜0 ⎟
⎝0 0 0 0 0 0 0 1 ⎠ ⎝00 ⎠ ⎝ 0⎠
0 0 0 0 0 0 0 0 00 0

The disturbance estimator’s gain K ∈ R8×1 will be computed through the Kalman
Filter recursion.
Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents of matrices Ão , B̃o
and C̃o , respectively, a Derivative-free nonlinear Kalman Filter can be designed for
the aforementioned representation of the system dynamics [31, 191, 229, 405]. The
associated Kalman Filter-based disturbance estimator is given by
measurement update:

K(k) = P− (k)C̃dT [C̃d · P− (k)C̃dT + R]−1

x̂(k) = x̂ − (k) + K(k)[z(k) − C̃d x̂ − (k)] (5.155)
P(k) = P− (k) − K(k)C̃d P− (k)
222 5 Differential Flatness Theory and Industrial Robotics

Fig. 5.13 Control scheme for the closed-chain underactuated robot consisting of a flatness-based
control term and a Kalman Filter-based disturbances estimator

time update:

P− (k + 1) = Ãd (k)P(k)ÃTd (k) + Q(k)

(5.156)
x̂ − (k + 1) = Ãd (k)x̂(k) + B̃d (k)ṽ(k)

To compensate for the effects of the disturbance forces it suffices to use in the control
loop the modified control input vector v1 = u − φ̂(t).

5.4.3 Simulation Tests

The efficiency of the proposed Kalman Filter-based control scheme for the under-
actuated robotic manipulator was evaluated in the tracking of various setpoints for
state variable q1 i.e., the linear displacement of mass M1 of the mechanism and state
variable q2 , i.e., the rotation angle of joint A depicted in Fig. 5.1 (Fig. 5.13).
As it can be observed in Figs. 5.14a, 5.15 and 5.16a the proposed control scheme
enabled accurate tracking of the reference setpoints by the robot’s state variables.
Moreover, as it can be observed in Figs. 5.14b, 5.15 and 5.16b the Kalman Filter-
based disturbance estimator was capable of identifying in real-time the external
disturbances that affected the robotic system. By including an additional term in the
control loop that was based on the disturbances estimation, it became possible to
compensate for the disturbances effects.
5.5 Distributed Filtering Under External Disturbances 223

(a) 15 2.5 (b)

3.5

x2 − xd2
x1 − xd1

10
2 3
5
1.5 2.5
0
2
−5 1

φ −est
0 20 40 60 0 20 40 60 1.5
time (sec) time (sec)
1
6 0.3
x3 − xd3

x4 − xd4
0.5
4 0.2
2 0.1 0

0 0 −0.5

−2 −0.1 −1
0 20 40 60 0 20 40 60 0 20 40 60 80 100
time (sec) time (sec) time

Fig. 5.14 Tracking of setpoint 1 a Convergence of the robot’s state variables to reference setpoints,
b estimation of disturbance input

(a) 30 3 (b)
1.5
20 2.5
x1 − x1d

x2 − x2d

10 2 1
0 1.5
−10 1 0.5
0 50 100 0 50 100
φ −est

time time
10 0.6 0
0.4
x3 − x3d

x4 − x4d

5
0.2 −0.5
0
0
−5 −0.2 −1
0 50 100 0 50 100 0 20 40 60 80 100
time time time

Fig. 5.15 Tracking of setpoint 2 a Convergence of the robot’s state variables to reference setpoints,
b estimation of disturbance input

5.5 Distributed Filtering Under External Disturbances

5.5.1 Overview

The distributed filtering procedure which was presented in Chap. 4 can be affected
by inaccuracies of the model of the dynamical system which can be due to external
disturbances (e.g., additive input disturbances) or parametric uncertainties and vari-
ations. In the case of localized state estimators, simultaneous estimation of the sys-
tem’s state vector and of the disturbances vector can be succeeded using disturbance
224 5 Differential Flatness Theory and Industrial Robotics

(a) 30 3
(b)
4

x2 − x2d
20 2.5
x1 − x1d

10 2 3

0 1.5 2
−10 1
0 50 100 0 50 100 1

φ −est
time time
10 x4 − x4d 0.6 0

0.4
x3 − x3d

5 −1
0.2
0 −2
0
−5 −0.2 −3
0 50 100 0 50 100 0 20 40 60 80 100
time time time

Fig. 5.16 Tracking of setpoint 3 a Convergence of the robot’s state variables to reference setpoints,
b estimation of disturbance input

observers [86, 268, 596]. These state estimators initially conceived with a static
observer gain can be suitably modified so as to be based on dynamic adaptation of
the observer gain through the Kalman Filter recursion. Once the disturbances vector
is estimated a supervisory control term can be introduced in the control law so as
to annihilate the disturbances effects. A common technique is the Unknown Input
Observer which estimates both the states of the system and the disturbance by aug-
menting a linear design model with a linear disturbance model. Another approach
is based on the Extended State Observer. Originally proposed in the form of a non-
linear Unknown Input Observer and later simplified to the linear version with one
tuning parameter, the Extended State Observer has the state and disturbance esti-
mation power of an Unknown Input Observer while being also simpler in tuning.
Another solution to the problem of simultaneous state and disturbance estimation
comes from the Perturbation Observer. The Perturbation observer is suitable not only
for estimation of additive disturbances but also for estimation of unmodeled system
variations. In place of static observer, gain for the aforementioned observers one can
consider online adaptation of the observer’s gain through the Kalman Filter recur-
sion. Therefore, it is possible to design Kalman Filter-based disturbance observers
exhibiting the advantages of Kalman Filter estimation such as minimization of the
estimation error and smoother convergence of the estimated state variables toward
the real state variables. In the case of distributed filtering, to compensate for such
disturbances and uncertainties the local Kalman Filters that constitute the distributed
state estimation scheme need to be redesigned. To this end, one can consider the
previously mentioned results on disturbance observers. The objective of this section
is to exploit results on disturbance observers so as to make the proposed distributed
filtering approach (derivative-free distributed nonlinear Kalman Filter) and the asso-
ciated state estimation-based control loop be more robust to external disturbances
and model uncertainties of the control system.
5.5 Distributed Filtering Under External Disturbances 225

5.5.2 Dynamics and Control of the Robot

5.5.2.1 The Robotic Model

The 1-DOF robotic model to be controlled with the use of visual servoing over a
distributed cameras network consists of a rigid link which is rotated by a DC motor,
as shown in Figs. 5.17 and 5.18. The model of the DC motor is described by the set
of equations:
L İ = −ke ω − RI + V
(5.157)
J ω̇ = ke I − kd ω − Γd

with the following notations L: armature inductance, I: armature current, ke : motor

electrical constant, R: armature resistance, and V : input voltage, taken as control
input, J: motor inertia, ω: rotor rotation speed, kd : mechanical dumping constant,
Γd : disturbance, or external load torque. It is assumed that Γd = mgl · sin(θ ), i.e.,
that the DC motor rotates a rigid robotic link of length l with a mass m attached to its
end. Then, denoting the state vector as [x1 , x2 , x3 ]T = [θ, θ̇ , θ̈ ]T , a nonlinear model
of the robot’s actuation is obtained

ẋ = f (x, t) + g(x, t)u (5.158)

where f (x, t) = [f1 (x, t), f2 (x, t), f3 (x, t)]T is a vector field function with elements:

f1 (x, t) = x2
f2 (x, t) = x3 (5.159)
k 2 +k R RJ+Kd L Rmgl mgl
f3 (x, t) = − e JL d x2 − JL x3 − JL sin(x1 ) − J cos(x1 )x2

Similarly, g(x, t) = [g1 (x, t), g2 (x, t), g3 (x, t)]T is a vector field function with
ke
elements: g1 (x, t) = 0, g2 (x, t) = 0, g3 (x, t) = JL . Having chosen the joint’s angle

Fig. 5.17 Diagram of the

state estimation-based
control loop for the 1-DOF
robot
226 5 Differential Flatness Theory and Industrial Robotics

Fig. 5.18 Visual servoing based on fusion of distributed EKF state estimates

to be the system’s output, and applying a transformation in accordance to differential

flatness theory, the state-space equation of the 1-DOF robot manipulator can be
rewritten as
x (3) = f¯ (x) + ḡ(x)u (5.160)

where functions f¯ (x) and ḡ(x) are given by

k +k R
2
f¯ (x) = − e JL d x2 − RJ+Kd L
x3 − Rmgl
JL sin(x1 ) −
mgl
J cos(x1 )x2
JL
ke
(5.161)
and ḡ(x) = JL

This is a system in the trivial (canonical) form; therefore, a state estimator can
be designed according to the previous results on derivative-free nonlinear Kalman
Filtering. The controller has to make the system’s output (angle θ of the motor) follow
a given reference signal xd . For measurable state vector x and uncertain functions
f (x, t) and g(x, t) an appropriate control law for the 1-DOF robotic model is

(n)
u= g(x,t) [xd
1
− f (x, t) − K T e + uc ] (5.162)

with e = x − xd , eT = [e, ė, ë, . . . , e(n−1) ]T , K T = [kn , kn−1 , . . . , k1 ], such that the
polynomial e(n) + k1 e(n−1) + k2 e(n−2) + · · · + kn e is Hurwitz. The previously defined
control law results into e(n) = −K T e + uc + d̃, where the supervisory control term
uc aims at the compensation of modeling errors as well as of the additive disturbance
d̃. Suitable selection of the feedback gain K assures the error dynamics will result
into limt→∞ e(t) = 0. In case of state estimation-based (sensorless control), and
5.5 Distributed Filtering Under External Disturbances 227

denoting, x̂ as the estimated state vector and ê = x̂ − xd as the estimated tracking

error one has
1 (n)
u= [x − f (x̂, t) − K T ê + uc ] (5.163)
g(x̂, t) d

As explained in the previous sections, the estimated state vector x̂ can be obtained
by fusing local state estimates from the vision nodes with the use of the Extended
Information Filter, or with the use of the Derivative-free Extended Information Filter
(distributed Derivative-free nonlinear Kalman Filter) (Fig. 5.18).

5.5.3 Simulation Tests

In the proposed distributed state estimation-based control scheme, simultaneous esti-

mation of the nonmeasurable state vector elements and of the external disturbance
affecting the system was performed (Fig. 5.24). The system’s model was first trans-
formed to the observer canonical form and next a disturbance state estimator was
used to obtain the values of the system’s state state vector and of the input disturbance
through the processing of the input disturbance through the processing of a sequence
of the robot’s link position measurements. The update of the state estimator’s gain
was performed using the standard Kalman Filter recursion. The augmented state
vector also contained the external disturbance as state variable, i.e., x = [θ, θ̇ , θ̈ , d].
There was no need to compute the system’s Jacobian matrix. The structure of the
local state estimators was as follows:

x̂˙ = Ax̂ + Bv(k) + Kf (z − ẑ) (5.164)

where ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 kf1 (z1 − ẑ1 )
⎜0 0⎟ ⎜ ⎟ ⎜ ⎟
A=⎜
0 1 ⎟ B = ⎜0⎟ Kf = ⎜kf2 (z1 − ẑ1 )⎟ (5.165)
⎝0 0 0 1⎠ ⎝1⎠ ⎝kf3 (z1 − ẑ1 )⎠
0 0 0 0 0 kf4 (z1 − ẑ1 )

while the transformed control input v was given by v = f (x̂) + g(x̂)u, where the
actual control input was given by Eq. (5.163).
The simulation experiments show that for the considered class of nonlinear sys-
tems, and particularly for the multi-cameras visual servoing model, it is possible
to apply the standard Information Filter for fusing distributed state estimates from
local derivative-free Kalman Filters. This fusion is performed at a central information
processing unit noted as master station. The proposed distributed filtering scheme is
capable of compensating for external disturbances and model uncertainties. More-
over, the proposed distributed filtering scheme enables to avoid use of the Extended
Information Filter which is based on local Extended Kalman Filters, thus also avoid-
ing to introduce in the estimation procedure linearization errors due to the truncation
228 5 Differential Flatness Theory and Industrial Robotics

(a) EIF master station EIF master station (b) EIF master station
1
1 1
state variable x1

state variable x2
0.5 0.5 0.95

estimation of external disturbance

0 0
0.9
−0.5 −0.5
−1 −1 0.85
0 5 10 15 20 0 5 10 15 20
0.8
t (sec) t (sec)

EIF master station EIF master station 0.75

200
state variable x3

control input u

2 100 0.7

0 0 0.65

−2 −100
0.6
0 5 10 15 20
−200
0 5 10 15 20 0 5 10 15 20 t (sec)
t (sec) t (sec)

Fig. 5.19 Tracking of a sinusoidal setpoint after fusing position measurements from distributed
cameras with the use of the Extended Information Filter: a estimation of the robot’s state variables,
b estimation of the disturbance force

(a) EIF master station EIF master station (b) EIF master station
1
state variable x1

state variable x2

1 1
estimation of external disturbance

0.5 0.5 0.95

0 0
0.9
−0.5 −0.5
−1 −1 0.85
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) 0.8

EIF master station EIF master station 0.75

200
state variable x3

control input u

2 100 0.7

0 0 0.65

−2 −100 0.6
0 5 10 15 20
−200
0 5 10 15 20 0 5 10 15 20 t (sec)
t (sec) t (sec)

Fig. 5.20 Tracking of a seesaw setpoint after fusing position measurements from distributed cam-
eras with the use of the Extended Information Filter: a estimation of the robot’s state variables, b
estimation of the disturbance force

of higher order Taylor expansion terms. Consequently, this means that one can have
a more robust state estimation.
The performance of the control loop which is based on Extended Information Fil-
tering is shown in Figs. 5.19 and 5.20, while the performance of the control loop which
is based on derivative-free Extended Information Filtering is shown in Figs. 5.21
and 5.22. The position, velocity, and acceleration setpoints for the robot’s joint are
denoted with a red line, the real values of these state variables are denoted with a
5.5 Distributed Filtering Under External Disturbances 229

(a) DEIF master station DEIF master station (b)

DEIF master station
state variable x1

state variable x2
1 1 1
0.5 0.5

estimation of external disturbance

0.95
0 0
−0.5 −0.5 0.9
−1 −1
0.85
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) 0.8
DEIF master station DEIF master station
200 0.75
state variable x3

control input u

2 100 0.7
0 0
0.65
−2 −100

−200 0 5 10 15 20
0 5 10 15 20 0 5 10 15 20
t (sec)
t (sec) t (sec)

Fig. 5.21 Tracking of a sinusoidal setpoint after fusing position measurements from distributed
cameras with the use of the derivative-free Extended Information Filter: a estimation of the robot’s
state variables, b estimation of the disturbance force

(a) DEIF master station DEIF master station (b)

DEIF master station
state variable x1

state variable x2

1 1 1
0.5 0.5
estimation of external disturbance

0.95
0 0
−0.5 −0.5 0.9
−1 −1
0.85
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) 0.8
DEIF master station DEIF master station
200 0.75
state variable x3

control input u

2 100
0.7
0 0
0.65
−2 −100

−200 0 5 10 15 20
0 5 10 15 20 0 5 10 15 20
t (sec)
t (sec) t (sec)

Fig. 5.22 Tracking of a seesaw setpoint after fusing position measurements from distributed cam-
eras with the use of the derivative-free Extended Information Filter: a estimation of the robot’s state
variables, b estimation of the disturbance force

blue line and their estimates produced by filtering are denoted with a green line.
It can be seen that both the control loop based on the Extended Information fil-
ter and the control loop based on the derivative-free Extended Information Filter
succeeded satisfactory tracking of the setpoints associated with the elements of the
robot’s state vector. Moreover, it can be observed that, under the same parameters
of the filtering-based control loop, the derivative-free Extended Information Filter
produced an accurate estimate (dashed blue line) of the disturbance force applied to
the robot (continuous red line) in shorter time than the Extended Information Filter.
230 5 Differential Flatness Theory and Industrial Robotics

5.6 Distributed Nonlinear Filtering Under

Measurement Delays

5.6.1 Networked Control Under Communication Disturbances

This section presents an approach to distributed state estimation-based control of

nonlinear MIMO systems, capable of incorporating delayed measurements in the
estimation algorithm while being also robust to packet losses. First, the section exam-
ines the problem of distributed nonlinear filtering over a communication/sensors
network, and the use of the estimated state vector in a control loop. As a possible
filtering approach, the Extended Information Filter is proposed [427]. However, the
Extended Information Filter introduces numerical errors due to the approximative lin-
earization of the system’s nonlinear model performed by the local Extended Kalman
Filters. To overcome the aforementioned weaknesses of the Extended Information
Filter, a derivative-free approach to Extended Information Filtering (EIF) is a possi-
ble solution [415, 434, 436]. Distributed filtering is now based on a derivative-free
implementation of Kalman Filtering which is shown to be applicable to MIMO non-
linear dynamical systems [428, 430, 437]. As previously explained, in the proposed
derivative-free Kalman Filtering method, the system is first subject to a lineariza-
tion transformation that is based on the differential flatness theory [152, 495, 429,
535]. Next state estimation is performed by applying the standard Kalman Filter
recursion to the linearized model. Unlike EIF, the proposed method provides esti-
mates of the state vector of the nonlinear system without the need for derivatives
and Jacobians calculation. By avoiding linearization approximations, the proposed
filtering method improves the accuracy of estimation of the system state variables,
and results in smooth control signal variations and in minimization of the tracking
error of the associated control loop. The section extends the results of [429, 430,
436, 437] toward a networked control scheme that is subject to random delays and
measurement packets drops.
Moreover, the section proposes a method for the compensation of random delays
and packet drops which may appear during the transmission of measurements and
of state vector estimates, and which can cause the deterioration of the performance
of the distributed filtering-based control scheme [476, 477, 569]. Two problems are
treated: (i) There are time delays and packet drops in the transmission of information
between the distributed local filters and a master (central) filter, (ii) there are time
delays and packet drops in the transmission of information from distributed sensors
to each one of the local filters. In the first case, the structure and calculations of
the master filter for estimating the aggregate state vector remain unchanged. In the
second case, the effect of the random delays and packets drops has to be taken into
account in the redesign of the local Kalman Filters, which implies a modified Riccati
equation for the computation of the covariance matrix of the state vector estimation
error, as well as the use of a correction (smoothing) term in the update of the state
vector’s estimate so as to compensate for delayed measurements arriving at the local
Kalman Filters. Finally, the section shows that the aggregate state vector produced
5.6 Distributed Nonlinear Filtering Under Measurement Delays 231

by a derivative-free Extended Information Filter, suitably modified to compensate

for communication delays and packet drops, can be used for sensorless control and
robotic visual servoing (see Fig. 5.24).

5.6.2 Networked Kalman Filtering for an Autonomous System

Fusion from distributed local filters and in particular distributed state estimation
using the Extended Information Filter has been analyzed in Chap. 3. The structure
of networked Kalman Filtering is shown in Fig. 5.23. The problem of distributed
filtering becomes more complicated if random delays and packet drops affect the
transmission of information between the sensors and local processing units (filters),
or between the local filters and the master filter where the fused state estimate is
computed. First, results on the stability of the networked linear Kalman Filter will be
presented [569]. The general state-space form of a linear autonomous time-variant
dynamical system is given by

x(k) = Ax(k − 1) + w(k, k − 1) (5.166)

where x(k) ∈ Rm×1 is the system’s state vector, A ∈ Rn×n is the system’s state
transition matrix, and w(k, k − 1) is the white process noise between time instants k
and k − 1. The sensor measurements are received starting at time instant k ≥ 1 and
are described by the measurement equation

z(k) = Cx(k) + v(k) (5.167)

where C ∈ Rp×m , z(k) ∈ Rp×1 and v(k) is the white measurement noise. Mea-
surements z(k) are assumed to be transmitted over a communication channel. To
denote the arrival or loss of a measurement to the local Kalman Filter, through the
communication network, one can use variable γk ∈ {0, 1}, where 1 stands for suc-
cessful delivery of the packet, while 0 stands for loss of the packet. Thus, in the
case of packet losses, the discrete-time Kalman Filter recursion that was described
in Sect. 4.2.2 (measurement update) is modified as

Fig. 5.23 Distributed

filtering over sensors
network with communication
delays and packet drops
232 5 Differential Flatness Theory and Industrial Robotics

K(k) = γk P− (k)C T [CP− (k)C T + R]−1 (5.168)

where γk ∈ {0, 1}. This modification implies that the value of the estimated state
vector x̂(k) remains unchanged if the a packet drop occurs, i.e., when γk = 0.

5.6.3 Smoothing Estimation in Case of Delayed Measurements

An issue which is associated to the implementation of such networked control systems

is how to compensate for random delays and packet losses so as to enhance the
accuracy of estimation and consequently to improve the stability of the control loop.
The idea of incorporating delayed measurements within a Kalman Filter framework is
a possible solution for the compensation of network-induced delays and packet losses,
and is also known as update with out-of-sequence measurements [30]. The solution
proposed in [30] is optimal under the assumption that the delayed measurement was
processed within the last sampling interval (one-step-lag problem). There have been
also some attempts to extend these results to nonlinear state estimation [178, 219].
More recently there has been research effort in the redesign of distributed Kalman
Filtering algorithms for linear systems so as to eliminate the effects of delays in
measurement transmissions and packet drops, while also alleviating the one-step-lag
assumption [569].
The smoothing approach that will be presented in the following is according to
the results given in [569]. The smoothing approach considers that for a linear time
invariant nonautonomous system

x(k) = Ax(k − 1) + Bu(k − 1) + w(k − 1)

(5.169)
z(k) = Cx(k) + v(k)

it holds

x(k) = AN x(k − N) + N j=1 A
N−j Bu(k − N + j − 1)+
 (5.170)
+ N j=1 A
N−j w(k − N + j − 1)

Denoting
AN = Φ(k, k − N) (5.171)

one has

x(k) = Φ(k, k − N)x(k − N) + Nj=1 A
N−j Bu(k − N + j − 1)+
N N−j (5.172)
+ j=1 A w(k − N + j − 1)
5.6 Distributed Nonlinear Filtering Under Measurement Delays 233

Setting
N N
w1 (k, k − N) = j=1 A
N−j Bu(k − N + j − 1) + j=1 A
N−j w(k − N + j − 1),
(5.173)

one has that x(k) can be written in a compact form as

x(k) = Φ(k, k − N)x(k − N) + w1 (k, k − N) (5.174)

Using that matrix Φ(k, k − N) is invertible, one has

x(k − N) = Φ(k, k − N)−1 x(k) − Φ(k, k − N)−1 w1 (k, k − N). (5.175)

The following notation is used

Φ1 (k − N, k) = −Φ(k, k − N)−1 (5.176)

while for the retrodiction of w1 (k, k − N) it holds

wa (k − N, k|k) = −Φ(k, k − N)−1 w1 (k, k − N) (5.177)

Then, to smooth the state estimation at time instant k − N, using the measurement
of output zi (k − N) received at time instant k one has the state equation

x̂(k − N, k) = Φ1 (k − N, k)x̂(k|k) + ŵa (k − N, k|k) (5.178)

while the associated measurement equation becomes

z(k − N) = Cx(k − N) + v(k − N) (5.179)

Substituting the above equation for x(k − N) into the equation for z(k − N) provides

z(k − N) = CΦ1 (k − N, k)x(k) + Cwa (k − N, k) + v(k − N) (5.180)

and the associated estimated output at time instant k − N is

ẑ(k − N) = CΦ1 (k − N, k)x̂(k|k) + C ŵa (k − N, k) (5.181)

From Eq. (5.181) the innovation for the delayed measurement can be obtained

z̃(k − N) = z(k − N) − ẑ(k − N) (5.182)

that is
z̃(k − N) = CΦ1 (k − N)x̃(k|k) + C w̃a (k − N), (5.183)
234 5 Differential Flatness Theory and Industrial Robotics

where x̃(k|k) = x(k) − x̂(k|k) is the state estimation error and w̃a (k − N, k|k) =
wa (k − N, k) − ŵa (k − N, k) is the estimation error for wα . With this innovation,
the estimation of the state vector x(k|k) at time instant k is corrected. The correction
(smoothing) relation is [569]

x̂ ∗ (k|k) = x̂(k|k) + M z̃(k − N, k) (5.184)

Therefore, again the basic problem for the implementation of the smoothing relation
provided by Eq. (5.184) is the calculation of the term wa (k − N, k) i.e.,

wa (k − N, k) = −Φ(k, k − N)−1 w1 (k, k − N) (5.185)

The estimation of the term wa (k − N, k) is provided by [569]

ŵa (k − N, k) = −Φ1 (k − N, k)·

N−1 (5.186)
· n=0 · C̃(n)[CP(k − n|k − n − 1)C T + R(k − n)]−1 z̃(k − n)

where z̃(k − n) = z(k − n) − ẑ(k − n) is the innovation for time instant k − n, while

C̃(n) = {An Q(k − n, k − n − 1) + N j=n+2 A
j−1 Q(k − j + 1, k − j)·
j−1 (5.187)
·[ m=n+1 A[I − Ki (k − m)C T ]}C T

with Q(k − n, k − n − 1) and R(k − n) denoting the process and measurement noise
covariance matrices at k − n. Then, the smoothing filter for the processing of the
delayed measurements at the i-th local filter is given by Eq. (5.184), i.e.,

x̂i∗ (k|k) = x̂i (k|k) + Mi [zi (k − N) − ẑi (k − N|k)] (5.188)

with matrix Mi to be given by [569]

Mi = [Pi (k|k)Φ1 (k − N, k)T + Pix̃w̃ ]Ci T Wi−1 (5.189)

and matrices Wi and Pix̃w̃ to be defined as

Wi = Ci {Φ1 (k − N, k)Pi (k|k)Φ1 (k − N, k)T + Φ1 (k − N, k)Pix̃w̃ +

(5.190)
+[Φ1 (k − N, k)Pix̃w̃ ]T + Qi∗ (k − N, k)} · Ci T + Ri (k − N)

and
N−1
Pix̃w̃ = AN n=0 Pi (k − N|k − N)Di (n)T · [Ci Pi (k − n|k − n − 1)Ci T + Ri (k − n)]−1 ·
·C̃i (n)T Φ1 (k − N, k)T − AN Qi∗ (k − N, k)
(5.191)
while matrix Di (n) is defined as
5.6 Distributed Nonlinear Filtering Under Measurement Delays 235

Di (n) = Ci A if N = 1 and
N−2 (5.192)
Di (n) = Ci A j=n [I − Ki (k − j − 1)Ci ] · A, if N > 1

The covariance matrix of the estimated noise term wαi (k − N, k) is calculated as

Qi∗ (k − N, k) = Q(k − N, k) − Φ1 (k − N, k)·

N−1  −1 T
· n=0 C̃i (n) Ci Pi (k − n|k − n − 1)Ci T + Ri (k − n) · C̃i (n) Φ1 (k − N, k)T
(5.193)
where
 
N T
Q(k − N, k) = Φ1 (k − N, k) j=1 A
j−1 · Q(k − j + 1, k − j)(Aj−1 ) Φ1 (k − N, k)T
(5.194)

After smoothing, the covariance matrix of the modified state estimation error
becomes

Pi∗ (k|k) = Pi (k|k) − [Pix̃w̃ + Pi (k|k)Φ1 (k − N, k)T ]×

×Ci (k − N)T Wi−1 Ci (k − N)× (5.195)
×[Pix̃w̃ + Pi (k|k)Φ1 (k − N, k)T ]T

5.6.4 Distributed Filtering-Based Fusion of the Robot’s

State Estimates

The description of the nonlinear robotic model in the canonical form of Eq. (3.49)
with matrices A, B and C given by Eq. (3.48) enables the application of the previous
analysis for the compensation of time delays and packet drops through smoothing
in the computation of the linear Kalman Filter. The fact that matrices A, B, and C
described in Eq. (3.48) are time invariant, facilitates the computation of the smoothing
Kalman Filter according to Eqs. (5.188) and (5.189). Using the time invariant matrices
defined in Eq. (5.91) and applying common discretization methods, one gets the
discrete-time equivalents of matrices A, B, and C which are defined as Ad , Bd , and
Cd , respectively. It is noted that due to the specific form of matrix B, the term Bu(k−1)
appearing in Eq. (5.169) is a variable of small magnitude with mean value close to
zero. Thus the term w1 (k, k − 1) = Bu(k − 1) + w(k, k − 1) differs little from
w(k, k − 1).
Fusion of the local state estimates which are provided by filters running on the
vision nodes can improve the accuracy and robustness of the performed state estima-
tion, thus also improving the performance of the robot’s control loop [330, 332, 342,
506, 507]. Under the assumption of Gaussian noise, a possible approach for fusing
the state estimates from the distributed local filters is the derivative-free Extended
Information Filter (DEIF). As explained in Sect. 4.2.5, the derivative-free Extended
Information Filter provides an aggregate state estimate by weighting the state vectors
236 5 Differential Flatness Theory and Industrial Robotics

Fig. 5.24 Visual servoing based on fusion of state estimates provided by local derivative-free
nonlinear Kalman Filters

produced by local Kalman Filters with the inverse of the associated estimation error
covariance matrices.
Visual servoing over the previously described cameras network is considered for
the nonlinear dynamic model of a single-link robotic manipulator. The robot can
be programmed to execute a manufacturing task, such as welding or cutting [532].
The position of the robot’s end effector in the cartesian space (and consequently the
angle for the robotic link) is measured by the aforementioned m distributed cameras.
The proposed multi-camera-based robotic control loop can be also useful in other
vision-based industrial robotic applications where the vision is occluded or heavily
disturbed by noise sources, e.g., in the case of welding. In such applications, there is
need to fuse measurements from multiple cameras so as to obtain redundancy in the
visual information and permit the robot to complete safely and within the specified
accuracy its assigned tasks [365, 594]. The considered 2-DOF robotic model consists
of two rigid links which are rotated by DC motors, as shown in Fig. 5.24.

5.6.5 Simulation Tests

The performance of the derivative-free nonlinear Kalman Filter was tested in state
estimation-based control, under measurement transmission delays, for a 2-DOF
5.6 Distributed Nonlinear Filtering Under Measurement Delays 237

(a) DEIF master station DEIF master station (b) DEIF master station
1 2 10
state variable x1

state variable x2

control input u1
0.5 0
1
0 −10
0
−0.5 −20
−1 −1 −30
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) t (sec)

DEIF master station DEIF master station DEIF master station

1 2 5
state variable x3

state variable x4

control input u2
0.5 0
1
0 −5
0
−0.5 −10
−1 −1 −15
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) t (sec)

Fig. 5.25 a Estimation and control of the motion of the robotic manipulator under transmission
delays, when tracking a sinusoidal trajectory. b Control torques on the robot’s joints

(a) (b)
DEIF master station DEIF master station DEIF master station
1 2 10
state variable x1

state variable x2

control input u1

0.5 1 0

−10
0 0
−20
−0.5 −1
0 5 10 15 20 0 5 10 15 20 −30
0 5 10 15 20
t (sec) t (sec)
t (sec)
DEIF master station DEIF master station DEIF master station
1 2 10
state variable x3

state variable x4

control input u2

0.5 1 0

0 0 −10

−0.5 −1 −20
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
t (sec) t (sec) t (sec)

Fig. 5.26 a Estimation and control of the motion of the robotic manipulator under transmission
delays, when tracking a seesaw trajectory. b Control torques on the robot’s joints

rigid-link robotic manipulator (Fig. 5.24). The considered visual servoing control
can find applications in several robotic tasks requiring dexterous manipulations (e.g.,
packing, welding, cutting, etc.) [478]. The differentially flat model of the robot has
been analyzed in Sect. 6.3.2. It was assumed that only measurements of the angle
of the robot’s joints could be obtained through the vision system. In the simulation
tests presented in Figs. 5.25 and 5.26, the setpoints for the state variables are marked
238 5 Differential Flatness Theory and Industrial Robotics

as red lines, the estimated state variables were denoted as green lines; whereas, the
real state variables were denoted as blue lines.
The delayed measurements were processed by the Kalman Filter recursion accord-
ing to the stages explained in the previous subsections. The smoothing of the
delayed measurements that was performed by the Kalman Filter was based on
Eq. (5.184), i.e., x̂ ∗ (k|k) = x(k|k) + M z̃(k − N, k). As explained in Sect. 5.6, matrix
M is a gain matrix calculated according to Eq. (5.189). The innovation is given by
z̃(k − N) = z(k − N) − ẑ(k − N). The tracking accuracy of the distributed filtering-
based control loop is depicted in Figs. 5.25 and 5.26.
Additionally, some performance indexes were used to evaluate the distributed
filtering-based control scheme. Table 5.3, shows the variation of the RMSE for state
variables x2 = θ̇1 and x4 = θ̇2 with respect to delay levels (d1 , d2 = k · Ts i.e.,
multiples of the sampling period Ts ), as well as with respect to the probability of
delay occurrence in the transmission of the measurement packets (p ∈ [0, 1]).
Moreover, the traces of the covariance error matrices Pi , i = 1, 2 and of the
smoothed covariance error matrices Pi∗ , i = 1, 2 appearing both at the local filters,
as well as the trace of the covariance error matrix P at the master (aggregation) filter,
for various delay levels, are shown in Table 5.4.
It can be noticed that despite the raise of the delay levels in the transmission of
measurements from the sensors (cameras) to the local information processing nodes
(local derivative-free Kalman Filters), only slight variations of the tracking errors for
state variables xi , i = 1, . . . , 4 were observed. Similarly, the changes of the traces
of the estimation error covariance matrices, both at the local filters and at the master
filter, were small.

Table 5.3 RMSE for robot state variables under measurement delays
d1 d2 p RMSEx2 RMSEx4
15 − 23 17 − 21 0.5 3.284e − 3 3.283e − 3
10 − 16 12 − 18 0.4 3.221e − 3 3.219e − 3
8 − 12 15 − 19 0.6 3.235e − 3 3.231e − 3
20 − 25 25 − 30 0.5 3.275e − 3 3.272e − 3
14 − 20 9 − 15 0.5 3.239e − 3 3.236e − 3
14 − 20 9 − 15 0.7 3.256e − 3 3.252e − 3

Table 5.4 Traces of the covariance error matrices for various delay levels
d1 d2 p Tr(Pi ) Tr(Pi∗ ) Tr(P)
15 − 23 17 − 21 0.5 2.118 · 10−1 1.046 · 10−1 1.699 · 10−3
10 − 16 12 − 18 0.4 2.398 · 10−1 1.180 · 10−1 1.743 · 10−3
8 − 12 15 − 19 0.6 2.277 · 10−1 1.121 · 10−1 1.719 · 10−3
20 − 25 25 − 30 0.5 2.212 · 10−1 1.090 · 10−1 1.709 · 10−3
14 − 20 9 − 15 0.5 2.081 · 10−1 1.029 · 10−1 1.697 · 10−3
14 − 20 9 − 15 0.7 2.098 · 10−1 1.037 · 10−1 1.698 · 10−3
Chapter 6
Differential Flatness Theory in Mobile
Robotics and Autonomous Vehicles

6.1 Outline

The chapter proposes differential flatness theory-based methods of filtering and

control for MIMO nonlinear dynamical systems, such as unmanned vehicles. These
can be of different types such as Unmanned Ground Vehicles (UGVs), Unmanned
Surface Vessels (USVs), Autonomous Underwater Vessels (AUVs), and Unmanned
Aerial Vehicles (UAVs). The considered nonlinear filtering schemes which are based
on differential flatness theory can be applied to autonomous vehicle models without
the need for calculation of Jacobian matrices. Nonlinear systems such as unmanned
ground vehicles, satisfying the differential flatness property, can be written in the
Brunovsky (canonical) form via a transformation of their state variables and control
inputs. After transforming the nonlinear system to the canonical form, it is straight-
forward to apply for it the standard Kalman Filter recursion. The performance of
the proposed derivative-free nonlinear filtering scheme is tested through simulation
experiments on the problem of state estimation-based control for autonomous navi-
gation of unmanned ground vehicles.
First, the chapter proposes state estimation-based control for cooperating mobile
robots, which have the kinematic model of a unicycle. To estimate with accuracy
the position of these robotic vehicles as well as their motion characteristics, fusion
of estimates from multiple sensors is performed with the use of the Derivative-free
distributed nonlinear Kalman Filter. The proposed derivative-free nonlinear filtering
method enables distributed state estimation, by substituting the Extended Information
Filter with the standard Information Filter recursion. This filtering approach has sig-
nificant advantages because, unlike the Extended Information Filter, it is not based on
local linearization of the nonlinear dynamics and computation of Jacobian matrices.
The proposed nonlinear control is in accordance with the principles of differential
flatness theory. The performance of the considered distributed filtering-based control
is tested through simulation experiments on the problem of autonomous navigation
of automatic ground vehicles (AGVs) under a master-slave scheme.

© Springer International Publishing Switzerland 2015 239

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_6
240 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Additionally, the chapter proposes the Derivative-free distributed nonlinear

Kalman Filter, for integrity monitoring of navigation sensors in automatic ground
vehicles (AGV). Unlike the Extended Information Filter (EIF), the proposed filter
avoids approximation errors caused by the linearization of the AGV kinematic model
and does not require the computation of Jacobians. The use of a statistical fault detec-
tion and isolation algorithm for processing the residuals generated by the proposed
filtering method, can provide an indication about the condition of the navigation
sensors and about failures that may have appeared. As an application example, the
chapter considers failure diagnosis for wheel encoders or IMU devices of an AGV.
Next, the chapter explains how controller design for autonomous 4-wheeled
ground vehicles can be performed using differential flatness theory. Using a 3-DOF
nonlinear model of the vehicle’s dynamics and through the application of differential
flatness theory, an equivalent model in linear canonical (Brunovksy) form is obtained.
For the latter model, a state feedback controller is developed that enables accurate
tracking of velocity setpoints. Moreover, it is shown that with the use of Kalman
Filtering it is possible to dynamically estimate the disturbances due to unknown
forces and torques exerted on the vehicle. The processing of velocity measurements
(provided by a small number of onboard sensors) through a Kalman Filter which has
been redesigned in the form of a disturbance observer results in accurate identification
of external disturbances affecting the vehicle’s dynamic model. By including in the
vehicle’s controller an additional term that compensates for the estimated disturbance
forces, the vehicle’s motion characteristics remain unchanged. Numerical simulation
confirms the efficiency of both the proposed controller and of the disturbance forces
estimator.
Moreover, in this chapter a solution to the problem of active control and dis-
turbances compensation in vehicles’ suspensions is proposed. It is shown that the
suspension model satisfies differential flatness properties and the associated flat out-
put is a weighted sum of the system’s state vector elements. Differential flatness
of the suspension’s model enables transformation into a linear canonical form for
which it is possible to design a state feedback controller. Kalman filtering is used for
estimating the nonmeasurable elements of the suspension’s transformed state vector
through the processing of measurements provided by a small number of onboard
sensors. Moreover, by reformulating the Kalman Filter as a disturbance observer
it is possible to simultaneously estimate the external disturbances and the system’s
transformed state vector. The inclusion of an additional control term based on the
disturbances estimation enables to compensate for the disturbances’ effects and to
attenuate vibrations. The performance of the proposed Kalman Filter-based active
suspension control scheme has been tested through numerical simulation experi-
ments.
Another topic treated by the chapter is the use of the Derivative-free nonlinear
Kalman Filter for developing a robust controller which can be applied to quadropters.
The control problem for quadropters is nontrivial and becomes further complicated if
this robotic model is subjected to uncertainties and external disturbances. Using dif-
ferential flatness theory, it is shown that the model of a quadropter can be transformed
6.1 Outline 241

to linear canonical form. For the linearized equivalent of the quadropter it is shown
that a state feedback controller can be designed. Since certain elements of the state
vector of the linearized system cannot be measured directly, it is proposed to estimate
them with the use of a novel filtering method, the so-called Derivative-free non-
linear Kalman Filter. Moreover, by redesigning the Kalman Filter as a disturbance
observer, it is shown that one can estimate simultaneously external disturbances terms
that affect the quadropter or disturbance terms which are associated with paramet-
ric uncertainty. The efficiency of the proposed control scheme is confirmed through
simulation experiments.
Finally, the chapter proposes a nonlinear control approach for an underactu-
ated vessel model based on differential flatness theory and using the Derivative-
free nonlinear Kalman Filter as a state and disturbance estimator. It is proven that
the sixth-order nonlinear model of the hovercraft is a differentially flat one. It is
shown that this model cannot be subjected to static feedback linearization; how-
ever, it admits dynamic feedback linearization which means that the system’s state
vector is extended by including as additional state variables the control inputs and
their derivatives. Next, using the differential flatness properties it is also proven that
this model can be subjected to input-output linearization and can be transformed
to an equivalent canonical (Brunovsky) form. Based on this latter description, the
design of a state feedback controller is carried out enabling accurate maneuvering
and trajectory tracking. Additional problems that are solved in the design of this
feedback control scheme are the estimation of the nonmeasurable state variables in
the hovercraft’s model and the compensation of modeling uncertainties and exter-
nal perturbations affecting vessel. To this end, the application of the Derivative-free
nonlinear Kalman Filter is proposed. This nonlinear filter consists of the Kalman
Filter’s recursion on the linearized equivalent of the vessel and of an inverse non-
linear transformation based on the differential flatness features of the system which
enables to compute estimates for the state variables of the initial nonlinear model.
The redesign of the filter as a disturbance observer makes possible the estimation and
compensation of additive perturbation terms affecting the hovercraft’s model. The
efficiency of the proposed nonlinear control and state estimation scheme is shown
again through simulation experiments.

6.2 State Estimation-Based Control of Autonomous Vehicles

6.2.1 Localization and Autonomous Navigation

of Ground Vehicles

Filtering-based state estimation for unmanned ground vehicles (UGVs) is a signifi-

cant topic because it enables their accurate localization and autonomous navigation
[39]. For nonlinear systems such as UGVs, and under the assumption of Gaussian
242 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

noise, the Extended Kalman Filter (EKF) is frequently applied for estimating the
nonmeasurable state variables through the processing of input and output sequences
or for obtaining estimates of the state vector through the fusion of measurements
coming from various sensors [31, 68, 192, 229, 372]. The Extended Kalman Filter is
based on linearization of the system’s dynamics using a first-order Taylor expansion
[244, 405, 408, 411, 412, 421, 573]. Although EKF is efficient in several estimation
and fusion problems, it is characterized by cumulative errors due to the local lineariza-
tion assumption and this may affect the accuracy of the UGV’s motion estimation or
even risk the stability of the UGV state estimation-based control loop.
First, in this section, and using differential flatness theory, the nonlinear system
is subjected to a linearization transformation. This makes possible (i) to design an
efficient control law for trajectories tracking, and (ii) to apply to the nonlinear system
a filtering method that it is based on the standard Kalman Filter recursion. Unlike the
Extended Kalman Filter (EKF), the proposed filtering method provides estimates of
the state vector of the UGV without the need for derivatives and Jacobians calculation.
By avoiding linearization approximations, the proposed derivative-free nonlinear
Kalman filtering method improves the accuracy of estimation of the system state
variables, and results in smooth control signal variations and in minimization of the
tracking error of the associated control loop.
Next, the section extends the results of [424, 429] toward nonlinear dynamical
systems, such as UGVs, which are described by multi-input multi-output (MIMO)
models. Actually, the section’s results are applicable to differentially flat MIMO
nonlinear dynamical systems which after applying the differential flatness theory can
be written in the Brunovksy (canonical) form [263, 340]. Simulation experiments on
the problem of autonomous navigation of a ground vehicle are provided to test the
performance of the proposed derivative-free Kalman Filter.

6.2.2 Application of Derivative-Free Kalman Filtering

to MIMO UGV Models

6.2.2.1 Kinematic Models for Autonomous Vehicles

The proposed method of derivative-free Kalman Filtering for MIMO nonlinear sys-
tems will be analyzed through an application example. The kinematic model of a
UGV (robotic unicycle) is considered. This is given by

ẋ = vcos(θ )
ẏ = vsin(θ ) (6.1)
θ̇ = ω = Lv tan(φ)

where v(t) is the velocity of the vehicle, L is the distance between the front and
the rear wheel axis of the vehicle, θ is the angle between the transversal axis of
6.2 State Estimation-Based Control of Autonomous Vehicles 243

the vehicle and axis O X , and φ is the angle of the steering wheel with respect to
the transversal axis of the vehicle. The position of such a vehicle is described by
the coordinates (x, y) of the center of its rear axis and its orientation is given by the
angle θ between the x-axis and the axis of the direction of the vehicle. The steering
angle φ and the speed v are considered to be the inputs of the system.
The problem of control of autonomous ground vehicles (AGVs) is first considered.
The position of such a vehicle is described by the coordinates (x, y) of the center of
its rear axis and its orientation is given by the angle θ between the x-axis and the axis
of the direction of the vehicle. The steering angle φ and the speed v are considered
to be the inputs of the system. The kinematic model of autonomous vehicles can be
expressed in the general form [429]
⎛ ⎞ ⎛ ⎞
ẋ cos(θ ) 0  
⎝ ẏ ⎠ = ⎝sin(θ ) v(t)
0⎠ · (6.2)
v(t)ρ(t)
θ̇ 0 1

where (x, y) are the coordinates of the center of the vehicle’s rear wheels axis,
v(t) is the velocity of the vehicle, and θ is the angle between the transversal axis
of the vehicle and axis O X . The autonomous vehicle is a nonholonomic system.
Nonholonomic systems are characterized by nonintegrable differential expressions,
such as
n
f i j (q1 , q2 , . . . , qn , t)q̇i = 0, j = 1, 2, . . . , m (6.3)
i=i

where q̇i represents the nth generalized coordinate (state variable), m is the number
of equations defining the nonholonomic constraints, q̇i represents the generalized
speed, and f i j are nonlinear functions of qi at time t. For the kinematic model of
Eq. (6.2), the following nonholonomic constraint exists:

ẋsin(θ ) − ẏcos(θ ) = 0 (6.4)

The curvature radius for any path can be written as

1 L
R(t) = = (6.5)
ρ(t) tan(φ)

where L is the distance between the front and the back wheels, and φ (namely the
steering angle) is the angle defined by the main axis of the vehicle and the velocity
vector of the front wheel (for cart-like vehicles as shown in Fig. 6.1), or the central
front point (for car-like vehicles as shown in Fig. 6.1). The value of R(t) is usually
bounded by Rmin , the minimum curvature radius (Fig. 6.2).
244 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Fig. 6.1 The model of the

autonomous robotic vehicle
(cart-like vehicle)

Fig. 6.2 The model of the

4-wheel autonomous vehicle
(car-like vehicle)

6.2.3 Controller Design for UGVs

Flatness-based control can be used for steering the vehicle along a desirable trajectory.
In the case of the autonomous vehicle of Eq. (6.1), the flat output is the cartesian
position of the center of the wheel axis, denoted as η = (x, y), while the other model
parameters can be written as:
 
cos(θ ) η̇
v = ±||η̇|| = tan(φ) = ldet (η̇η̈)/v3 (6.6)
sin(θ ) v
6.2 State Estimation-Based Control of Autonomous Vehicles 245

These formulas show simply that θ is the tangent angle of the curve and tan(φ)
is the associated curvature. With reference to a generic driftless nonlinear system
q̇, q ∈ R n , w ∈ R m , dynamic feedback linearization consists in finding a feedback
compensator of the form
ξ̇ = α(q, ξ ) + b(q, ξ )u
(6.7)
w = c(q, ξ ) + d(q, ξ )u

with state ξ ∈ R v and input u ∈ R m , such that the closed-loop system of Eqs. (6.1)
and (6.7) is equivalent under a state transformation z = T (q, ξ ) to a linear system.
The starting point is the selection of a m-dimensional output η = h(q) to which
a desired behavior can be assigned (this is the previously defined flat output). One
then proceeds by successively differentiating the output until the input appears in a
nonsingular way. If the sum of the output differentiation orders equals the dimension
n + v of the extended state space, full input-state-output linearization is obtained.
The closed-loop system is then equivalent to a set of decoupled input-output chains
of integrators from u i to ηi . The exact linearization procedure is illustrated for the
vehicle’s model of Eq. (6.2). As flat output η = (x, y) the coordinates of the center
of the wheel axis is considered. Differentiation with respect to time then yields
[382, 422]      
ẋ cos(θ ) 0 v
η̇ = = · (6.8)
ẏ sin(θ ) 0 ω

showing that only v affects η̇, while the angular velocity ω cannot be recovered from
this first-order differential information. To proceed, one needs to add an integrator
(whose state is denoted by ξ ) on the linear velocity input
 
cos(θ )
v = ξ, ξ̇ = α⇒η̇ = ξ (6.9)
sin(θ )

where α denotes the linear acceleration of the vehicle. Differentiating further one
obtains    
cos(θ ) sin(θ )
η̈ = ξ̇ + ξ θ̇ =
 sin(θ )   )
cos(θ
(6.10)
cos(θ ) −ξ sin(θ ) α
=
sin(θ ) ξ cos(θ ) ω

and the matrix multiplying the modified input (α, ω) is nonsingular if ξ = 0. Under
this assumption one defines
     
α cos(θ ) −ξ sin(θ ) u
= · 1 (6.11)
ω sin(θ ) ξ cos(θ ) u2

and η̈ is denoted as    
η̈1 u1
η̈ = = =u (6.12)
η̈2 u2
246 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

which means that the desirable linear acceleration and the desirable angular velocity
can be expressed using the transformed control inputs u 1 and u 2 . Then, the resulting
dynamic compensator is (return to the initial control inputs v and ω)

ξ̇ = u 1 cos(θ ) + u 2 sin(θ )
v=ξ (6.13)
ω = u 2 cos(θ)−u
ξ
1 sin(θ)

Being ξ ∈R, it is n + v = 3 + 1 = 4, equal to the output differentiation order in

Eq. (6.12). In the new coordinates

z1 = x
z2 = y
(6.14)
z 3 = ẋ = ξ cos(θ )
z 4 = ẏ = ξ sin(θ )

The extended system is thus fully linearized and described by the chains of integra-
tors, in Eq. (6.12), and can be rewritten as

z̈ 1 = u 1 , z̈ 2 = u 2 (6.15)

The dynamic compensator of Eq. (6.13) has a potential singularity at ξ = v = 0, i.e.,

when the vehicle is not moving, which is a case not met while executing the trajectory
tracking. It is noted, however, that the occurrence of such a singularity is structural
for nonholonomic systems. In general, this difficulty must be obviously taken into
account when designing control laws on the equivalent linear model. A nonlinear
controller for output trajectory tracking, based on dynamic feedback linearization, is
easily derived. Assume that the autonomous vehicle must follow a smooth trajectory
1
(xd (t), yd (t)) which is persistent, i.e., for which the nominal velocity vd = (ẋd2 + ẏd2 ) 2
along the trajectory never goes to zeros (and thus singularities are avoided). On the
equivalent and decoupled system of Eq. (6.15), one can easily design an exponentially
stabilizing feedback for the desired trajectory, which has the form

u 1 = ẍd + k p1 (xd − x) + kd1 (ẋd − ẋ)

(6.16)
u 2 = ÿd + k p1 (yd − y) + kd1 ( ẏd − ẏ)

and which results in the following error dynamics for the closed-loop system

ëx + kd1 ėx + k p1 ex = 0

(6.17)
ë y + kd2 ė y + k p2 e y = 0

where ex = x − xd and e y = y − yd . The proportional-derivative gains are chosen

as k p1 > 0 and kd1 > 0 for i = 1, 2. Knowing the control inputs u 1 , u 2 , for
the linearized system one can calculate the control inputs v and ω applied to the
6.2 State Estimation-Based Control of Autonomous Vehicles 247

vehicle, using Eq. (6.7). The above result is valid, provided that the dynamic feedback
compensator does not meet the singularity. In the general case of design of flatness-
based controllers, the following theorem assures the avoidance of singularities in the
proposed control law [382]:
Theorem: Let λ11 , λ12 and λ21 , λ22 , be respectively the eigenvalues of two equations
of the error dynamics, given in Eq. (6.7). Assume that, for i = 1, 2 it is λ11 < λ12 < 0
(negative real eigenvalues), and that λi2 is sufficiently small. If
   0
ẋd (t) ε̇
mint≥0 || ||≥ x0 (6.18)
ẏd (t) ε̇ y

with ε̇x0 = ε̇x (0) = 0 and ε̇0y = ε̇ y (0) = 0, then the singularity ξ = 0 is never met.

6.2.4 Derivative-Free Kalman Filtering for UGVs

It is assumed now that the vehicle’s velocity has to be estimated through the process-
ing of the sequence of position measurements by a filtering algorithm. To this end,
the derivative-free Kalman Filter for MIMO nonlinear dynamical systems can been
used. From the application of the differential flatness theory presented in Sect. 6.4.3
for transforming the initial nonlinear vehicle’s model into a linearized equivalent that
is finally written in the Brunovsky form, one has Eq. (6.12) which means ẍ = u 1 and
ÿ = u 2 . Next, the state variables x1 = x, x2 = ẋ, x3 = y, and x4 = ẏ are defined.
Considering the state vector x∈R 4×1 , the following matrices are also defined
⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0
⎜0 0 0⎟0 ⎜1 0⎟
A=⎜
⎝0
⎟, B = ⎜ ⎟
0 1⎠0 ⎝0 0⎠
0 0 0 0 0 1 (6.19)
 
1000
C=
0010

Using the matrices of Eq. (6.26), one obtains the Brunovsky form of the MIMO
robotic unicycle
ẋ = Ax + Bv
(6.20)
y = Cx

where the new input v is given by v = [u 1 (x, t), u 2 (x, t)]T . For the robotic model
of Eq. (6.20), state estimation can be performed using the standard Kalman Filter
recursion, as described in Eqs. (4.5) and (4.6).
248 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

6.2.5 Simulation Tests

6.2.5.1 Extended Kalman Filter-Based Navigation of the Autonomous

Vehicle

The vehicle’s kinematic model of Eq. (6.1) is considered. A GPS sensor or encoders
placed at the vehicle’s wheels can be used to provide measurements of the carte-
sian coordinates of the vehicle x(t) and y(t) (displacement of the vehicle), over a
sampling period T . The values of the vehicle’s velocity components along the x and
y axes are not directly available and are estimated through the processing of the
sequence of position measurements with the use of a filtering algorithm. Computing
the vehicle’s speed through the differentiation of the position measurements would
introduce cumulative errors in the value of the vehicle’s velocity, which in turn can
affect the performance of the control loop. To avoid such errors, an estimation of
the vehicle’s velocity is obtained through the processing of the sequence of position
measurements with the use of a filtering algorithm.
Assuming a constant sampling period Δtk = T , the measurement equation is
z(k + 1) = γ (x(k)) + v(k), where z(k) is the vector containing the sequence of
measurements of the cartesian coordinates of the vehicle and v(k) is the measurement
noise vector.

z(k) = [x(k) + v1 (k), y(k) + v2 (k)], k = 1, 2, 3 . . . (6.21)

To obtain the Extended Kalman Filter (EKF), the kinematic model of the vehicle is
linearized about the estimates x̂(k) and x̂ − (k), and the control input U (k) is applied.
The EKF recursion consists of the measurement update part and of the time update
part as described in Eqs. (4.13) and (4.14), respectively. One has to use the input gain
matrix L(k) ⎛ ⎞
T cos(θ (k)) 0
L(k) = ⎝T sin(θ (k)) 0 ⎠ (6.22)
0 T

and to compute the Jacobian Jφ (x̂(k)) given by the following

⎛ ⎞
1 0 −v(k)sin(θ )T
Jφ (x̂(k)) = ⎝0 1 v(k)cos(θ )T ⎠ (6.23)
00 1

while for the elements of the process noise covariance matrix which is defined as
Q(k) = diag[σ 2 (k), σ 2 (k), σ 2 (k)] indicative values are σ 2 (k) = 10−3 .
Using the estimated state vector, function φ(x) appearing in the state equations
part and γ (x) appearing in the measurements equations part of the vehicle’s kinematic
6.2 State Estimation-Based Control of Autonomous Vehicles 249

model become φ(x̂(k)) = [x̂(k), ŷ(k)]T , and γ (x̂(k)) = [x̂(k), ŷ(k)], respectively.
The associated Jacobian matrix JγT (x̂ − (k)) is given by
 
100
Jγ (x̂ − (k)) = (6.24)
010

The vehicle is steered by the flatness-based controller analyzed in Sect. 6.4.3

u 1 = ẍd + K p1 (xd − x) + K d1 (ẋd − ẋ)

u 2 = ÿd + K p2 (yd − y) + K d2 ( ẏd − ẏ)
ξ̇ = u 1 cos(θ ) + u 2 sin(θ ) (6.25)
v = ξ, ω = u 2 cos(θ)−u ξ
1 sin(θ)

The use of EKF for estimating the vehicle’s velocity along the x-axis (denoted as
ẋ) and the vehicle’s velocity along the y-axis (denoted as ẏ) enables the successful
application of nonlinear steering control of Eq. (6.25).
Indicative results about tracking of the circular reference trajectory with the use
of the Extended Kalman Filters are shown in Figs. 6.3, 6.4, 6.5, and 6.6. In Fig. 6.3,
one can notice the accuracy of tracking of the reference trajectory, succeeded by the
proposed state estimation-based control scheme. In Fig. 6.4, the accuracy of tracking
of the x and y axis position setpoints is depicted. In Fig. 6.5, the associated x and y
axis tracking errors are shown. Finally, in Fig. 6.6, the x and y axis velocity estimation
errors are given.
Indicative results about tracking of the eight-shaped reference trajectory with use
of the Extended Kalman Filter are shown in Fig. 6.7, 6.8, 6.9, and 6.10. In Fig. 6.7

(a) 20 (b) 20
15 15

10 10

5 5
Y

0 0

−5 −5

−10 −10

−15 −15
−15 −10 −5 0 5 10 15 20 −15 −10 −5 0 5 10 15 20
X X

Fig. 6.3 a Tracking of a circular reference trajectory (green line) by the autonomous vehicle and
associated estimation of the vehicle’s position provided by the Extended Kalman Filter (continuous
yellow line). b Tracking of a circular reference trajectory (green line) by the autonomous vehicle
and real position of the vehicle (dashed red line)
250 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 15 (b) 15

10 10

vehicle Y position
vehicle X position

5 5

0 0

−5 −5

−10 −10

−15 −15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.4 Tracking of a circular reference trajectory with the use of the EKF: a tracking of the x-axis
reference setpoint, b tracking of the y-axis reference setpoint

(a) 1 (b) 10
0.8
8
0.6

0.4
tracking error x

6
tracking error y

0.2

0 4

−0.2
2
−0.4

−0.6
0
−0.8

−1 −2
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.5 Tracking of a circular reference trajectory with use of the EKF: a tracking error along the
x-axis, b tracking error along the y-axis

one can notice the accuracy of tracking of the reference trajectory, succeed by the
proposed state estimation-based control scheme. In Fig. 6.8 the accuracy of tracking
of the x and y axis position setpoints is depicted. In Fig. 6.9, the associated x and
y axis tracking errors are shown. Finally, in Fig. 6.10 the x and y axis velocity
estimation errors are given.
Indicative results about tracking of the complex-curved reference trajectory with
use of the Extended Kalman Filter are shown in Fig. 6.11, 6.12, 6.13, and 6.14.
In Fig. 6.11, one can notice the accuracy of tracking of the reference trajectory,
succeeded by the proposed state estimation-based control scheme. In Fig. 6.12, the
6.2 State Estimation-Based Control of Autonomous Vehicles 251

(a) 1.5 (b) 2.5

2
1
1.5
vehicle velocity x

vehicle velocity y
0.5
1

0 0.5

0
−0.5
−0.5
−1
−1

−1.5 −1.5
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.6 Tracking of a circular reference trajectory by the autonomous vehicle with use of the EKF:
a convergence of the estimated x-axis velocity (green line) to the associated real velocity (blue line),
b convergence of the estimated y-axis velocity (green line) to the associated real velocity (green
line)

(a) 20 (b) 20

15 15

10 10
Y

5 5
Y

0 0

−5 −5

−10 −10
−2 0 2 4 6 8 10 12 14 16 −2 0 2 4 6 8 10 12 14 16
X X

Fig. 6.7 a Tracking of an eight-shaped reference trajectory (green line) by the autonomous vehicle
and associated estimation of the vehicle’s position provided by the Extended Kalman Filter (yellow
line). b Tracking of a circular reference trajectory (green line) by the autonomous vehicle and real
position of the vehicle (dashed red line)

accuracy of tracking the x and y axis position setpoints is depicted. In Fig. 6.13, the
associated x and y axis tracking errors are shown. Finally, in Fig. 6.14, the x and y
axis velocity estimation errors are given.
252 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) (b)
14 6

12
4
vehicle X position

10

vehicle Y position
2
8

6 0

4
−2
2
−4
0

−2 −6
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.8 Tracking of an eight-shaped reference trajectory with use of the EKF: a tracking of the
x-axis reference setpoint, b tracking of the y-axis reference setpoint

1 0.2
(a) (b)
0.8
0.15
0.6
tracking error x

0.4 0.1
tracking error y

0.2
0.05
0
0
−0.2

−0.4 −0.05
−0.6
−0.1
−0.8

−1 −0.15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.9 Tracking of an eight-shaped reference trajectory with use of the EKF: a tracking error
along the x-axis, b tracking error along the y-axis

6.2.6 Derivative-Free Kalman Filter-Based Navigation

of the Autonomous Vehicle

The performance of the proposed derivative-free nonlinear Kalman Filter was tested
in the problem of state estimation-based control for autonomous navigation of the
previously described robotized vehicle (car-like robot) (Fig. 6.1). The differentially
flat model of the autonomous vehicle and its transformation to the Brunovksy form
has been analyzed in Sect. 6.2.2. It was assumed that only measurements of the
6.2 State Estimation-Based Control of Autonomous Vehicles 253

(a) 1.5 (b) 2.5

2
1
1.5
vehicle velocity x

vehicle velocity y
0.5
1

0 0.5

0
−0.5
−0.5
−1
−1

−1.5 −1.5
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.10 Tracking of an eight-shaped reference trajectory by the autonomous vehicle with use of
the EKF: a convergence of the estimated x-axis velocity (green line) to the associated real velocity
(blue line), b convergence of the estimated y-axis velocity (green line) to the associated real velocity
(green line)

(a) 20 (b) 20
15 15

10 10

5 5

0 0
Y

−5 −5

−10 −10

−15 −15

−20 −20
−15 −10 −5 0 5 10 15 −15 −10 −5 0 5 10 15
X X

Fig. 6.11 a Tracking of a complex-curved reference trajectory (green line) by the autonomous
vehicle and associated estimation of the vehicle’s position provided by the Extended Kalman Filter
(yellow line). b Tracking of a circular reference trajectory (green line) by the autonomous vehicle
(red dashed line) and real position of the vehicle (dashed red line)

cartesian coordinates of the vehicle (displacement on the x y-plane) could be obtained

through a GPS unit (localization of moderate accuracy), RTK-GPS (localization of
higher accuracy), or through laser, visual, and sonar sensors with reference to specific
landmarks (the latter measuring approaches require transformation from a local to a
global coordinates system).
254 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 10 (b) 15
8
10
6
vehicle X position

vehicle Y position
4
5
2
0 0
−2
−5
−4
−6
−10
−8
−10 −15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.12 Tracking of a complex-curved reference trajectory with use of the EKF: a tracking of
the x-axis reference setpoint, b tracking of the y-axis reference setpoint

1
(a) (b) 10
0.8
8
0.6
0.4
tracking error x

6
tracking error y

0.2
0 4
−0.2
2
−0.4
−0.6
0
−0.8
−1 −2
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.13 Tracking of a complex-curved reference trajectory with use of the EKF: a tracking error
along the x-axis, b tracking error along the y-axis

Indicative results about tracking of the circular reference trajectory with use of the
derivative-free Kalman Filter are shown in Figs. 6.15, 6.16, and 6.17. Comparing the
estimation performed by the derivative-free MIMO nonlinear Kalman Filter to the
one performed by the Extended Kalman Filter it can be noticed that the derivative-
free filtering approach results in more accurate state estimates. Finally, it is noted
that following a similar methodology the chapter’s approach can be applied also to
4-wheel autonomous vehicles.
Indicative results about tracking of the circular reference trajectory with use of the
derivative-free Kalman Filter are shown in Figs. 6.15, 6.16, and 6.17. In Fig. 6.15,
one can notice the accuracy of tracking of the reference trajectory, succeeded by
6.2 State Estimation-Based Control of Autonomous Vehicles 255

(a) 1.5 (b) 5

4
1
3
2
vehicle velocity x

vehicle velocity y
0.5
1
0 0
−1
−0.5
−2
−3
−1
−4
−1.5 −5
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.14 Tracking of a complex-curved reference trajectory by the autonomous vehicle with use of
the EKF: a convergence of the estimated x-axis velocity (green line) to the associated real velocity
(blue line), b convergence of the estimated y-axis velocity (green line) to the associated real velocity
(green line)

(a) 20 (b) 20
15 15

10 10

5 5
Y

0 0

−5 −5

−10 −10

−15 −15
−15 −10 −5 0 5 10 15 20 −15 −10 −5 0 5 10 15 20
X X

Fig. 6.15 a Tracking of a circular reference trajectory (green line) by the autonomous vehicle and
associated estimation of the vehicle’s position provided by the derivative-free Kalman Filter (yellow
line). b Tracking of a circular reference trajectory (green line) by the autonomous vehicle and real
position of the vehicle (dashed red line)

the proposed state estimation-based control scheme. In Fig. 6.16, the associated x
and y axis tracking errors are shown. Finally, in Fig. 6.17, the x and y axis velocity
estimation errors are given.
Indicative results about tracking of the eight-shaped reference trajectory with
the use of Extended Kalman Filter are shown in Figs. 6.18, 6.19, 6.20, and 6.21.
In Fig. 6.18, one can notice the accuracy of tracking of the reference trajectory,
256 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) (b) 10
0.25
0.2 8
0.15
tracking error x

tracking error y
0.1 6
0.05
0 4
−0.05
−0.1 2
−0.15
−0.2 0
−0.25
−2
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.16 Tracking of a circular reference trajectory with the use of DKF: a tracking error along
the x-axis, b tracking error along the y-axis

(a) 1.5 (b) 4

1 3
vehicle velocity x

vehicle velocity y

0.5 2

0 1

−0.5 0

−1 −1

−1.5 −2
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.17 Tracking of a circular reference trajectory by the autonomous vehicle with the use of
DKF: a convergence of the estimated x-axis velocity (green line) to the associated real velocity
(blue line), b convergence of the estimated y-axis velocity (green line) to the associated real velocity
(green line)

succeeded by the proposed state estimation-based control scheme. In Fig. 6.19, the
accuracy of tracking of the x and y axis position setpoints is depicted. In Fig. 6.20,
the associated x and y axis tracking errors are shown. Finally, in Fig. 6.21, the x and
y axis velocity estimation errors are given.
Indicative results about tracking of the complex-curved reference trajectory with
use of the Extended Kalman Filter are shown in Figs. 6.22, 6.23, 6.24, and 6.25. In
Fig. 6.22, one can notice the accuracy of tracking the reference trajectory, succeeded
by the proposed state estimation-based control scheme. In Fig. 6.23, the accuracy of
6.2 State Estimation-Based Control of Autonomous Vehicles 257

(a) 20 (b) 20

15 15

10 10

Y
5 5
Y

0 0

−5 −5

−10 −10
−2 0 2 4 6 8 10 12 14 16 −2 0 2 4 6 8 10 12 14 16
X X

Fig. 6.18 a Tracking of an eight-shaped reference trajectory (green line) by the autonomous vehicle
and associated estimation of the vehicle’s position provided by the derivative-free Kalman Filter
(yellow line). b Tracking of a circular reference trajectory (green line) by the autonomous vehicle
and real position of the vehicle (dashed red line)

(a) (b)
14 15

12
10
10
vehicle Y position
vehicle X position

5
8

6 0

4
−5
2
−10
0

−2 −15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.19 Tracking of an eight-shaped reference trajectory with use of the DKF: a tracking of the
x-axis reference setpoint, b tracking of the y-axis reference setpoint

tracking the x and y axis position setpoints is depicted. In Fig. 6.24, the associated x
and y axis tracking errors are shown. Finally, in Fig. 6.25, the x and y axis velocity
estimation errors are given.
Comparing the estimation performed by the derivative-free MIMO nonlinear
Kalman Filter with the one performed by the Extended Kalman Filter it can be
noticed that the derivative-free filtering approach results in more accurate state esti-
mates. Moreover, comparing the associated state estimation-based control loop of the
258 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) (b) 0.25

0.25
0.2
0.2
0.15 0.15
tracking error x

tracking error y
0.1
0.1
0.05
0 0.05
−0.05
0
−0.1
−0.15 −0.05
−0.2
−0.1
−0.25
−0.15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.20 Tracking of an eight-shaped reference trajectory with use of the DKF: a tracking error
along the x-axis, b tracking error along the y-axis

(a) 1.5 (b) 2.5

2
1
1.5
vehicle velocity x

vehicle velocity y

0.5
1

0 0.5

0
−0.5
−0.5
−1
−1

−1.5 −1.5
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.21 Tracking of an eight-shaped reference trajectory by the autonomous vehicle with use of
the DKF: a convergence of the estimated x-axis velocity (green line) to the associated real velocity
(blue line), b convergence of the estimated y-axis velocity (green line) to the associated real velocity
(green line)

autonomous vehicle that was based on the derivative-free MIMO nonlinear Kalman
Filter to the control that was based on the Extended Kalman Filter it was observed
that the first control scheme was significantly more robust and capable of tracking
with better accuracy the desirable trajectories. These findings show the suitability of
the considered derivative-free MIMO nonlinear Kalman Filter for localization, con-
trol, and autonomous navigation of autonomous vehicles. Finally, it is noted that the
chapter’s approach can also be applied to various types of 4-wheel robotic vehicles.
6.2 State Estimation-Based Control of Autonomous Vehicles 259

(a) 20 (b) 20
15 15

10 10

5 5

Y
0 0
Y

−5 −5

−10 −10

−15 −15

−20 −20
−15 −10 −5 0 5 10 15 −15 −10 −5 0 5 10 15
X X

Fig. 6.22 a Tracking of a complex-curved reference trajectory (green line) by the autonomous
vehicle and associated estimation of the vehicle’s position provided by the derivative-free Kalman
Filter (yellow line). b Tracking of a circular reference trajectory (green line) by the autonomous
vehicle and real position of the vehicle (dashed red line)

(a) 10 (b) 15
8
10
6
vehicle Y position
vehicle X position

4
5
2
0 0
−2
−5
−4
−6
−10
−8
−10 −15
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.23 Tracking of a complex-curved reference trajectory with use of the DKF: a tracking of
the x-axis reference setpoint, b tracking of the y-axis reference setpoint

Results about the accuracy of estimation provided by the considered nonlinear

filtering algorithms, as well as about the accuracy of tracking succeeded by the
associated state estimation-based control loop are given in Table 6.1 [437]. It can
be noticed that the Derivative-free nonlinear Kalman Filter is significantly more
accurate and robust than the Extended Kalman Filter. Its accuracy is comparable to
the one of the Unscented Kalman Filter. Moreover, its accuracy is also comparable
to the one succeeded by the Particle Filter for a moderate number of particles (e.g.,
260 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) (b) 10
0.25
0.2 8
0.15
tracking error x

0.1 6

tracking error y
0.05
0 4
−0.05
−0.1 2
−0.15
−0.2 0
−0.25
−2
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.24 Tracking of a complex-curved reference trajectory with use of the DKF: a tracking error
along the x-axis, b tracking error along the y-axis

(a) 1.5 (b) 20

1
15
vehicle velocity x

vehicle velocity y

0.5
10
0
5
−0.5

0
−1

−1.5 −5
0 10 20 30 40 50 0 10 20 30 40 50
time time

Fig. 6.25 Tracking of a complex-curved reference trajectory by the autonomous vehicle with use of
the DKF: a convergence of the estimated x-axis velocity (green line) to the associated real velocity
(blue line), b convergence of the estimated y-axis velocity (green line) to the associated real velocity
(green line)

N = 1000). The Particle Filter can succeed to diminish further the variance of
estimation; however, this requires a large number of particles (e.g., N > 1500) and
induces additional computational cost (see Table 6.2).
The aggregate runtime and the cycle time (duration of each iteration of the fil-
ter’s algorithm) for the Derivative-free nonlinear Kalman Filter, for the Extended
Kalman Filter, for the Unscented Kalman Filter, and for the Particle Filter (in case
of N = 1000 particles), using the Matlab platform on a PC with a 1.6 GHz Intel
i7 processor, is depicted in Table 6.2. It can be noticed that the Derivative-free
6.2 State Estimation-Based Control of Autonomous Vehicles 261

Table 6.1 RMSE of tracking (Gaussian noise)

RMSE x RMSE y RMSEθ
SPKF 0.0146 0.0162 0.0012
PF 0.0060 0.0031 0.0230
EKF 0.0371 0.0294 0.1835
DKF 0.0086 0.0103 0.0004

Table 6.2 RMSE of tracking (Rayleigh noise)

RMSE x RMSE y RMSEθ
SPKF 0.0218 0.0238 0.0026
PF 0.0092 0.0049 0.0717
EKF 0.0475 0.0338 0.2445
DKF 0.0091 0.0102 0.0021

Table 6.3 Run time of nonlinear estimation algorithms

SPKF PF EKF DKF
Total runtime (s) 111.14 623.92 104.39 89.69
Cycle time (s) 0.0222 0.1248 0.0209 0.0179

nonlinear Kalman Filter is faster than the other nonlinear estimation algorithms.
Actually, the Derivative-free nonlinear Kalman Filter (DKF) appears to be 20 %
faster than the Unscented Kalman Filter (UKF). The improved speed of the DKF
can be more apparent in higher dimensional, multidegree of freedom systems, where
the computation of cross-covariance matrices used by the UKF will impose more
numerical operations (Table 6.3).

6.3 State Estimation-Based Control and Synchronization

of Cooperating Vehicles

6.3.1 Overview

In the previous section, differential flatness theory was used for filtering and con-
trol in unmanned ground vehicles (UGVs). Next, the results are generalized toward
filtering and control of cooperating vehicles. There are many types of field opera-
tions that are performed by cooperating vehicles [56]. In this section, a method for
autonomous navigation of agricultural robots under a master-slave scheme is devel-
oped. The method comprises the following elements: (i) a path planner for generat-
ing automatically the trajectory that has to be followed by the cooperating robotized
262 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

vehicles, (ii) a nonlinear controller that makes the robots track with precision the
desirable trajectories, (iii) a distributed filtering scheme for estimating the motion
characteristics of the vehicles through the fusion of measurements about the vehicles’
coordinates coming from multiple position sensors (e.g., multiple GPS devices). The
autonomous navigation of the cooperating agricultural robots is finally implemented
through state estimation-based control. The nonlinear controller uses the estimated
state vector of the robots, as provided by distributed filtering, so as to generate the
control signal that defines the robots’ speed and heading angle.
A solution to decentralized information fusion over sensor networks, such as the
network that collects measurements for the system of the robotic harvesters, can be
obtained with the use of distributed Kalman Filtering [23, 163, 215, 264, 265, 326,
369, 380, 381, 485, 534, 552]. In this section, the previously analyzed derivative-free
approach to Extended Information filtering will be used. In the proposed derivative-
free Kalman Filtering method, the system is first subject to a linearization transfor-
mation and next state estimation is performed by applying the standard Kalman Filter
recursion to the linearized model. Another issue that has to be taken into account
for the autonomous functioning of the robotized vehicles is nonlinear control for
precise tracking of desirable trajectories [69]. The chapter proposes flatness-based
control for steering the cooperating UGVs along their reference paths [427, 535]. The
performance of the proposed distributed filtering-based control scheme is evaluated
through simulation experiments.
Such a multirobot system performs distributed information processing for esti-
mating the position and motion characteristics of the vehicles. At a first stage, mea-
surements from onboard sensors are combined with measurements from multiple
position sensors (e.g., GPS devices) and are initially processed by local filters to
provide local state vector estimates. At a second stage, the local state estimates for
the robotic vehicles are fused using a distributed filtering algorithm. Thus an aggre-
gate state vector of the robotic harvesters is obtained (see Fig. 6.26). Such a filtering
approach has several advantages: (i) it is fault tolerant: if a local information process-
ing unit is subject to a fault then state estimation is still possible, (ii) the information
processing scheme is scalable and can be expanded with the inclusion of more local
information processing units (local filters), (iii) the bandwidth for the exchange of
information between the local units and the aggregate filter remains limited since
there is no transmission of raw measurements but only transmission of local state
estimates and of the associated covariance matrices.
Under the assumption of a Gaussian measurement model, a solution to distrib-
uted information fusion for the robotized vehicles can be obtained with the use of
distributed Kalman Filtering [369, 520]. Distributed state estimation in the case of
non-Gaussian models has been also studied in several other research works [329,
462]. In this section, the Derivative-free distributed nonlinear Kalman Filter will be
compared against the Extended Information Filter, which is actually an approach for
fusing state estimates provided by local Extended Kalman Filters [264, 427].
6.3 State Estimation-Based Control and Synchronization of Cooperating Vehicles 263

Fig. 6.26 Sensor fusion at the local filters for obtaining local state estimates

6.3.2 Distributed Kalman Filtering for Unmanned

Ground Vehicles

Multiple cooperating mobile robots are considered. The coordinates and motion
characteristics are monitored by several sensors, as shown in Fig. 6.27. Each mobile
robot is taken to be described by the kinematic model of a unicycle vehicle. Estimation
for the individual vehicles with the use of the Extended Kalman Filter was described
in Sect. 6.2. Moreover, distributed estimation for multiple cooperating vehicles can
be based on the Extended Information Filter which was analyzed in Chap. 4.
On the other side, the concept of Derivative-free distributed nonlinear Kalman
Filter has been described in Chaps. 4 and 5. It is assumed that distributed filtering is
used to estimate with accuracy the position and motion characteristics of each agri-
cultural vehicle. Each local filtering within this distributed state estimation scheme
provides an estimate of the vehicle’s state vector by fusing measurements from mul-
tiple RTK-GPS units. If the derivative-free Kalman Filter is used in place of the
Extended Kalman Filter, then in the EIF formulation, given in Eqs. (4.53) and (4.54)
the following matrix substitutions should be performed: Jφ (k)→Ad , Jγ (k)→Cd ,
where matrices Ad and Cd are the discrete-time equivalents of matrices A and C
264 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Fig. 6.27 Precise

localization of the
cooperating vehicles through
multisensor fusion and
distributed filtering

which have been defined in Eq. (6.26). Both the stages of the Kalman Filter and
of the Extended Kalman Filter have been explained in Chap. 4. Matrices Ad and Cd
can be computed using established discretization methods. Moreover, the covariance
matrices P(k) and P − (k) are the ones obtained from the linear Kalman Filter update
equations given also in Chap. 4.
It is assumed now that the vehicle’s velocity has to be estimated through the
processing of the sequence of position measurements by a filtering algorithm. To this
end, the derivative-free Kalman Filter for MIMO nonlinear dynamical systems can
been used. From the application of differential flatness theory presented in Sect. 6.2.2
for transforming the initial nonlinear vehicle’s model into a linearized equivalent that
is finally written in the Brunovsky form, one has Eq. (6.12) which means ẍ = u 1 and
ÿ = u 2 . Next, the state variables x1 = x, x2 = ẋ, x3 = y, and x4 = ẏ are defined.
Considering the state vector x∈R 4×1 , the following matrices are also defined
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 00 10
⎜0 0 0 0⎟ ⎜1 0 ⎟ ⎜0 0 ⎟
A=⎜
⎝0
⎟, B = ⎜ ⎟, C = ⎜ ⎟ (6.26)
0 0 1⎠ ⎝0 0⎠ ⎝0 1 ⎠
0 0 0 0 01 00

Using the matrices of Eq. (6.26), one obtains the Brunovsky form of the MIMO
robot model ẋ = Ax + Bv and y = C x, where the new input v is given by
v = [u 1 (x, t), u 2 (x, t)]T . This is a robotic model, for which state estimation can
be performed using the standard Kalman Filter recursion.

6.3.3 Simulation Tests

Master-slave cooperation of two agricultural robots was considered (see Fig. 6.26).
The master tractor generates a reference path and the motion characteristics (velocity,
acceleration, orientation) that the slave tractor has to follow. It was assumed that mea-
surements of the x y coordinates of the vehicles could be obtained through multiple
GPS units (localization of moderate accuracy), or multiple local RTK-GPS stations
6.3 State Estimation-Based Control and Synchronization of Cooperating Vehicles 265

EIF Master Station

(a) 20 (b) 20
15 15

10 10

5 5
Y

Y
0 0

−5 −5

−10 −10

−15 −15
−15 −10 −5 0 5 10 15 20 −15 −10 −5 0 5 10 15 20
X X

Fig. 6.28 Extended Information Filtering and flatness-based control for cooperating robot har-
vesters: a synchronized tracking of reference path 1, b position of the synchronized vehicles every
100 sampling periods

(localization of higher accuracy). Moreover, localization of the vehicles could be

performed using measurements of their distance from a reference surface. This dis-
tance can be measured with the use of different onboard sensors, e.g., laser, sonar, or
vision sensors. The measurements from the GPS were combined with the distance
sensor measurements and were initially processed by local filters to provide local
state vector estimates. At a second stage, the local state estimates for the robotic
vehicles were fused using the Extended Information Filter. Using the outcome of the
Extended Information Filter state, estimation-based control was implemented.
Indicative results about tracking of various trajectories (e.g., reference paths fol-
lowed by the vehicles to perform harvesting) with use of the Extended Information
Filter are shown in Figs. 6.28, 6.29, 6.30, and 6.31. It can be noticed that the Extended
Information Filter provides accurate estimates of the vehicle’s state vector thus also
resulting in efficient tracking of the reference trajectories. Finally, it is noted that the
paper’s approach can be applied also to various types of 4WD agricultural vehicles.

6.4 Distributed Fault Diagnosis for Autonomous Vehicles

6.4.1 Integrity Testing in Navigation Sensors of AGVs

Integrity testing of navigation sensors in Automatic Ground Vehicles (AGVs) can be

succeeded through the processing of measurements from distributed sources [378,
427, 515]. On the one side, one can have estimation of the AGV’s cartesian coor-
dinates through the fusion of measurements coming from various onboard sensors
266 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

EIF Master Station

(a) 30 (b) 30
25 25

20 20

15 15
Y

Y
10 10

5 5

0 0

−5 −5

−10 −10
0 10 20 30 40 50 60 70 0 10 20 30 40 50 60 70
X X

Fig. 6.29 Extended Information Filtering and flatness-based control for cooperating robot har-
vesters: a synchronized tracking of reference path 2, b position of the synchronized vehicles every
100 sampling periods

EIF Master Station

(a) 30 (b) 30
25 25

20 20

15 15

10 10
Y

5 5
Y

0 0

−5 −5

−10 −10

−15 −15
0 10 20 30 40 50 60 70 −10 0 10 20 30 40 50 60 70
X X

Fig. 6.30 Extended Information Filtering and flatness-based control for cooperating robot har-
vesters: a synchronized tracking of reference path 3, b position of the synchronized vehicles every
100 sampling periods

(considered as reference sensors, e.g., GPS devices and distance measuring devices
such as laser and vision sensors). Based on these measurements, distributed filter-
ing algorithms produce estimates of the vehicle’s state vector through a multistage
fusion procedure. On the other side, one can have an estimate of the AGV’s cartesian
coordinates using the measurements provided by failure-prone sensors such as wheel
encoders or IMU. Residuals generation takes place through the comparison of the
6.4 Distributed Fault Diagnosis for Autonomous Vehicles 267

EIF Master Station

(a) 30 (b) 30

25 25

20 20

15 15
Y

Y
10 10

5 5

0 0

−5 −5
−10 0 10 20 30 40 50 60 70 −10 0 10 20 30 40 50 60 70
X X

Fig. 6.31 Extended Information Filtering and flatness-based control for cooperating robot har-
vesters: a synchronized tracking of reference path 4, b detailed motion of the synchronized vehicles

coordinates of the vehicle computed through the measurements of the monitored sen-
sors against the coordinates which are estimated by processing measurements from
the reference sensors. Finally, the statistical processing of residuals by a suitable
fault detection and isolation algorithm can provide an indication about the condition
of the navigation sensors and failures that may have appeared [31, 157, 290, 414].
This section implements distributed filtering with the use of the Derivative-free
Extended Information Filter (DEIF), for generating such residuals. As explained,
to apply the Derivative-free Extended Information Filter it is necessary first to per-
form transformation of the system’s dynamic or kinematic model into a canonical
form, and this is achieved by using a change of coordinates (diffeomorphism) that is
based on differential flatness theory. Unlike the Extended Information Filter (EIF),
the proposed filter avoids approximation errors caused by the linearization of the
AGV kinematic model and does not require the computation of Jacobians [112, 264,
369]. The Derivative-free Extended Information Filter appears to be faster than the
EIF, while also providing very accurate (in terms of variance) state estimates. The
application of the Derivative-free Extended Information Filter to AGVs confirms and
extends the initial results about the performance of derivative-free nonlinear Kalman
Filtering given in [427].

6.4.2 Sensor Fusion for AGV Navigation

To improve the accuracy of the estimation of the AGV’s state vector, fusion of the
measurements provided by onboard sensors can be performed (IMU, GPS, gyro-
compasses, laser, vision sensors). The inertial coordinates system O X Y is defined.
268 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Fig. 6.32 Reference frames

for the AGV

Furthermore, the coordinates system O  X  Y  is considered (Fig. 6.32). O  X  Y 

results from O X Y if it is rotated by an angle θ . The coordinates of the center of
gravity of the AGV with respect to O X Y are (x, y). It is assumed that the coordi-
nates of the distance measuring sensor (e.g., vision sensor or radar) with respect to
  
O  X  Y  are xi , yi . The orientation of the AGV with respect to O X Y is θi . Thus the
coordinates of the distance measuring sensor with respect to O X Y are (xi , yi ) and
its orientation is θi , and are given by:
 
xi (k) = x(k) + xi sin(θ (k)) + yi cos(θ (k))
 
yi (k) = y(k) − xi cos(θ (k)) + yi sin(θ (k)) (6.27)

θi (k) = θ (k) + θi

For the localization of the AGV, a first set of measurements comprises the AGV’s
cartesian coordinates and comes from an IMU or a GPS. A second set of measure-
ments consists of the readings of the AGV’s distance d(k) from a reference surface
P j , and can be provided from a vision sensor or laser. A reference surface P j in the
j j
AGV’s 2D motion area can be represented by Pr and Pn (see Fig. 6.32), where (i)
j j
Pr is the normal distance of the AGV from the origin O, (ii) Pn is the angle between
the normal line to the plane and the x-direction. The sensor providing this second set
of measurements is at position [xi (k), yi (k)] with respect to the inertial coordinates
system O X Y and its orientation is θi (k). Using the above notation, the distance of
j j
the GPS sensor, from the plane P j is represented by Pr , Pn (see Fig. 6.32) [427]:

j j j j
di (k) = Pr − xi (k)cos(Pn ) − yi (k)sin(Pn ) (6.28)
6.4 Distributed Fault Diagnosis for Autonomous Vehicles 269

In this section, it will be assumed that the measurements vector is given by γ (x(k)) =
[x(k), y(k), d(k)], where [x(k), y(k)] are the AGV’s cartesian coordinates, and d(k)
is the distance measurement. Moreover, it is assumed that at every time instant more
than one GPS measurements are available and that multiple estimates of the vehicle’s
coordinates can be obtained by fusing the individual GPS measurements with the
measurement of the distance from the reference surface.
In the case of multisensor fusion with the use of the EKF a constant sampling
period Δtk = T is assumed and the measurement equation for the vehicle is z(k) =
γ (x(k)) + v(k), where z(k) is the vector containing the sequence of measurements
of the cartesian coordinates of the vehicle and v(k) is the measurement noise vector.
Apart from the coordinates of the vehicle which are provided by the IMU or GPS one
can also consider a distance measurement d(k) from a landmark surface, as shown
in Fig. 6.32. Thus, the measurements vector becomes

z(k) = [x(k) + v1 (k), y(k) + v2 (k), d(k) + v3 (k)], k = 1, 2, 3 . . . (6.29)

To obtain the Extended Kalman Filter (EKF), the kinematic model of the vehicle
is linearized about the estimates x̂(k) and x̂ − (k), and the control input U (k) is
applied, as already explained in Sect. 6.2.5.1. The EKF recursion consists of the
measurement update part and of the time update part as described in Eqs. (4.13) and
(4.14), respectively. The input gain matrix L(k) for the vehicle model can be written
in the form ⎛ ⎞
T cos(θ (k)) 0
L(k) = ⎝T sin(θ (k)) 0 ⎠ (6.30)
0 T

while the Jacobian matrices Jφ (x̂(k)) and JγT (x̂ − (k)) are given by
⎛ ⎞
1 0 −v(k)sin(θ̂ (k))T
Jφ (x̂(k)) = ⎝0 1 v(k)cos(θ̂(k))T ⎠ (6.31)
00 1

⎛ ⎞
1 0 0
Jγ (x̂ − (k)) = ⎝ 0 1 0 ⎠
−cos(Pn ) −sin(Pn ) −xi cos(θ̂(k) − Pn ) − yi sin(θ̂ (k) − Pn )
(6.32)

For the elements of the noise covariance matrix Q(k) = diag[σ 2 (k), σ 2 (k), σ 2 (k)]
indicative values can be σ 2 (k) = 10−3 . Using the estimated state vector, function
γ (x) appearing in the measurements equations part of the vehicle’s kinematic model
becomes γ (x̂(k)) = [x̂(k), ŷ(k), d̂(k)].
270 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

6.4.3 Canonical Form for the AGV Model

From the application of the differential flatness theory presented in Sect. 6.2, the
initial nonlinear vehicle’s model is transformed into a linearized equivalent. This is
finally written in the Brunovsky form, and one has Eq. (6.15) which means ẍ = u 1
and ÿ = u 2 . Next, the state variables x1 = x, x2 = ẋ, x3 = y and x4 = ẏ are used.
Considering the state vector x∈R 4×1 , the following matrices are also defined
⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0
⎜0 0 0⎟0 ⎜1 0⎟
A=⎜
⎝0
⎟ B=⎜ ⎟
0 1⎠0 ⎝0 0⎠
0 0 0 0 0 1 (6.33)
 
1000
C=
0010

Using the matrices of Eq. (6.33), one obtains the Brunovsky (canonical) form of the
MIMO model of the AGV.
ẋ = Ax + Bv
(6.34)
y = Cx

where the new input v is given by v = [u 1 (x, t), u 2 (x, t)]T . For the system of
Eqs. (6.33) and (6.34), state estimation is possible by applying the standard Kalman
Filter. The system is first turned into discrete-time form using common discretization
methods and then the recursion of the linear Kalman Filter described in Eqs. (4.5)
and (4.6) is applied.

6.4.4 Derivative-Free Extended Information Filtering

for UGVs

As mentioned above, for the system of Eq. (6.33), state estimation is possible by
applying the standard Kalman Filter. The system is first turned into discrete-time
form using common discretization methods and then the recursion of the linear
Kalman Filter described in Eqs. (4.5) and (4.6) is applied.
If the derivative-free Kalman Filter is used in place of the Extended Kalman Filter
then in the EIF formulation, given in Eqs. (4.53) and (4.54), the following matrix
substitutions should be performed: Jφ (k)→Ad , Jγ (k)→Cd , where matrices Ad and
Cd are the discrete-time equivalents of matrices A and C which have been defined
Eq. (6.33) and which appear also in the measurement and time update of the standard
Kalman Filter recursion. Matrices Ad and Cd can be computed using established
discretization methods. Moreover, the covariance matrices P(k) and P − (k) are the
ones obtained from the linear Kalman Filter update equations given in Sect. 4.2.2.
6.4 Distributed Fault Diagnosis for Autonomous Vehicles 271

Finally, state estimates for the AGV which are provided by multiple local Kalman
Filters are fused into one single state estimate, either with the use of the Extended
Information Filter or with the use of the Derivative-free distributed nonlinear Kalman
Filter.

6.4.5 Simulation Tests

Distributed state estimation takes place in two levels (i) at the lower level local state
estimates are generated by local nonlinear filters, such as Extended Kalman Filters
or derivative-free nonlinear Kalman Filters. The latter, provide local state estimates
through the fusion of the cartesian coordinates of the AGV with the distance of
the AGV from a reference surface, (ii) at the higher level, fusion of the local state
estimates is performed with the use of distributed state estimation approaches, such
as the Extended Information Filter, and the previously described Derivative-Free
Extended Information Filter.
It is considered that the previously analyzed AGV model is monitored by n = 2
local filters that fuse different GPS measurements of the vehicle’s coordinates with
a measurement of the distance from the reference surface (provided by a vision
or laser sensor). The distributed filtering architecture and the residuals generation
procedure is shown in Fig. 6.33. The reference frames defining the AGV’s motion
and the AGV’s distance from the reference surface are according to Fig. 6.32 and
have been analyzed in [408, 427]. To decide on the existence of a navigation sensor
failure or intermittence, the produced residuals can be tested against a fault threshold
that is chosen according to the Generalized Likelihood Ratio criterion or the χ 2 test
[31, 290, 414].
In the simulation experiments, various AGV trajectories were considered and an
indicative one (circular path of radius equal to 10 m) is depicted in Fig. 6.34 and
6.37. Regarding computation time it was observed that the Derivative-free Extended
Information Filter was significantly faster than the Extended Information Filter, due
to no need to compute online Jacobian matrices and partial derivatives. Results on the
EIF and the Derivative-free EIF performance, in estimating the state vector the AGV
when fusing measurements from multiple onboard sensors, are given in Figs. 6.35,
6.36, 6.38, and 6.39, respectively. In case of an IMU or encoder fault (additive fault
to the sensor’s output), one can observe deviation between the distributed filtering-
based coordinates and the coordinates which are computed using the measurements
of the monitored sensor. Further processing of the residuals by a FDI algorithm with
optimally selected fault threshold enables detection of incipient navigation sensor
faults as well as fault isolation [31, 157, 290, 414].
272 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Fig. 6.33 Integrity monitoring for the AGV navigation sensors using a distributed filtering scheme

EIF DEIF
(a) 20 (b) 20
15 15

10 10

5 5
Y

0 0

−5 −5

−10 −10

−15 −15
−15 −10 −5 0 5 10 15 20 −15 −10 −5 0 5 10 15 20
X X

Fig. 6.34 Fault-free sensor case. Trajectory generated by state estimation-based control of the
AGV with the use of: a standard Extended Information Filter, b Derivative-free nonlinear Extended
Information Filter
6.5 Velocity Control of 4-Wheel Vehicles 273

EIF DEIF
(a) 0.2 (b) 0.2

0.15 0.15

0.1 0.1

0.05 0.05
residual x

residual x
0 0

−0.05 −0.05

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2
0 10 20 30 40 50 0 10 20 30 40 50
time (sec) time (sec)

Fig. 6.35 Fault-free sensor case. Residual of the x-axis position of the AGV generated with a
standard Extended Information Filter, b Derivative-free nonlinear Extended Information Filter

0.2
EIF 0.2
DEIF
(a) (b)
0.15 0.15

0.1 0.1

0.05 0.05
residual y

residual y

0 0

−0.05 −0.05

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2
0 10 20 30 40 50 0 10 20 30 40 50
time (sec) time (sec)

Fig. 6.36 Fault-free sensor case. Residual of the y-axis position of the AGV generated with a
standard Extended Information Filter, b Derivative-free nonlinear Extended Information Filter

6.5 Velocity Control of 4-Wheel Vehicles

6.5.1 Overview

Next, it will be shown that differential flatness theory-based filtering and control can
be applied to more advanced models of autonomous vehicles, such as the 4-wheel
ground vehicle. The precise modeling of the vehicles’ dynamics improves the effi-
ciency of vehicles controllers in adverse cases, for example in high velocity, when
performing abrupt maneuvers, under mass and loads changes or when moving on
274 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) EIF (b) DEIF

20 20

15 15

10 10

5 5
Y

Y
0 0

−5 −5

−10 −10

−15 −15
−15 −10 −5 0 5 10 15 20 −15 −10 −5 0 5 10 15 20
X X

Fig. 6.37 Faulty sensor case. Trajectory generated by state estimation-based control of the AGV
with the use of: a standard Extended Information Filter, b Derivative-free nonlinear Extended
Information Filter

EIF DEIF
(a) 0.2 (b) 0.2

0.15 0.15

0.1 0.1

0.05 0.05
residual x

residual x

0 0

−0.05 −0.05

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2
0 10 20 30 40 50 0 10 20 30 40 50
time (sec) time (sec)

Fig. 6.38 Faulty sensor. Residual of the x-axis position of the AGV generated with a standard
Extended Information Filter, b Derivative-free nonlinear Extended Information Filter

rough terrain. Using model-based control approaches, it is possible to design a non-

linear controller that maintains the vehicle’s motion characteristics within desirable
ranges [39, 331, 337, 349, 350, 360, 404, 406, 516, 589]. When the vehicle’s dynam-
ics is subject to modeling uncertainties or when there are unknown forces and torques
exerted on the vehicle it is important to be in position to estimate in real-time dis-
turbances and unknown dynamics so as to compensate for them through the control
input and to maintain the satisfactory performance of the vehicle’s automated steer-
ing system. To this direction, estimation for the unknown dynamics of the vehicle and
state estimation-based control schemes have been developed [209, 331, 496, 545].
6.5 Velocity Control of 4-Wheel Vehicles 275

EIF DEIF
(a) 0.2 (b) 0.2

0.15 0.15

0.1 0.1

0.05 0.05
residual y

residual y
0 0

−0.05 −0.05

−0.1 −0.1

−0.15 −0.15

−0.2 −0.2
0 10 20 30 40 50 0 10 20 30 40 50
time (sec) time (sec)

Fig. 6.39 Faulty sensor case. Residual of the y-axis position of the AGV generated with a standard
Extended Information Filter, b Derivative-free nonlinear Extended Information Filter

The objective of this section is two-fold. On the one side it analyzes the design
of a controller for autonomous navigation of automatic ground vehicles (AGVs). On
the other side, it proposes a solution to the problem of four-wheel vehicle control
under model uncertainties and external disturbances. Considering, that only under
ideal conditions the dynamic model of the vehicle is precisely known (e.g., there
may be variations in the transported mass, or in the cornering stiffness coefficients
characterizing the interaction of the tires with the ground, or in the position of the
vehicle’s center of gravity) and that in several cases there is uncertainty about the
forces and torques developed on the vehicle (e.g., traction and braking torques on the
wheels, forces due to traction of implements, or lateral forces which generate torques
affecting the yaw stability of the vehicle), the need for designing robust controllers
of the autonomous vehicles becomes obvious [44, 488, 497, 554]. By compensating
efficiently such disturbance forces and torques, safety features of the vehicle are
improved and its autonomous functioning remains reliable even under adverse road
conditions.
Dynamic analysis for the 4-wheel vehicle is provided. A 3-DOF model is intro-
duced having as elements the vehicle’s velocity along the horizontal and vertical
axis of an inertial reference frame as well as the rate of change of its orientation
angle (this is the angle defined by the vehicle’s longitudinal axis and the horizontal
axis of the frame). Lateral forces are shown to affect the vehicle’s motion and to be
dependent on the longitudinal and lateral velocity of the vehicle, on the yaw rate and
on the cornering stiffness coefficients for the front and rear tires. The control inputs
to the vehicles’ dynamic model are the traction/bracking wheel torque and the turn
angle of the steering wheel. Since the parameters of the dynamic model of the vehi-
cle cannot always be known with precision or may be time-varying (e.g., cornering
stiffness coefficients, transported mass) and since there may be unmodeled external
forces and torques exerted on the vehicle (e.g., due to road condition, disturbances
276 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

in traction forces) it is important to design a control loop with robustness to the

aforementioned sources of uncertainty and disturbances, as well as to be in position
to estimate in real-time such disturbances through the processing of measurements
from a small number of onboard sensors.
Next, it is shown how a nonlinear controller for the aforementioned vehicle’s
model can be obtained through the application of differential flatness theory [152,
465, 516, 535]. The flat output for the vehicle’s model is a vector comprising the
x-axis velocity and a second variable based on a linear relation between the y-axis
velocity and the rate of change of the orientation angle [349, 350]. By expressing all
state variables and the control input of the four-wheel vehicle model as functions of
the flat output and its derivatives, the system’s dynamic model is transformed into
the linear Brunovksy (canonical) form [317, 479]. For the latter model, it is possible
to design a state feedback controller that succeeds accurate tracking of the vehicle’s
velocity setpoints.
By exploiting the vehicle’s exactly linearized model and its transformation into a
canonical form it is possible to design a linear state estimator for approximating the
system’s state vector through the processing of measurements coming from a small
number of onboard sensors. To this end, Derivative-free nonlinear Kalman Filtering
is used. Unlike the Extended Kalman Filter, the proposed filtering method provides
estimates of the state vector of the nonlinear system without the need for derivatives
and Jacobians calculation [422, 427]. By avoiding linearization approximations, the
proposed filtering method improves the accuracy of estimation of the system’s state
variables. Moreover, it is shown that it is possible to redesign the Kalman Filter in
the form of a disturbance observer, and using the estimation of the disturbance to
develop an auxiliary control input that compensates for their effects. In this way, the
vehicle’s control and autonomous navigation system can become robust with respect
to uncertainties in the model’s parameters or uncertainties about external forces and
torques. It is also noted that in terms of computation speed, the proposed Kalman
Filter-based disturbance estimator for the vehicle is faster than disturbance estimators
that may be based on other nonlinear filtering approaches (e.g., Extended Kalman
Filter, Unscented Kalman Filter or Particle Filter) thus becoming advantageous for
the real-time estimation of the unknown vehicle dynamics [436, 437].
The efficiency of the proposed nonlinear control and Kalman Filter-based distur-
bances estimation scheme is evaluated through numerical simulation tests. It will be
shown that by accurately estimating disturbance forces and torques, the control loop
succeeds elimination of the tracking error for all state variables of the vehicle.

6.5.2 Dynamic Model of the Vehicle

6.5.2.1 Definition of Parameters in 4-Wheel Vehicle Dynamic Model

With reference to Fig. 6.40 (where the lateral forces applied on the wheels are consid-
ered to define the vehicle’s motion) one has the following parameters: β is the angle
6.5 Velocity Control of 4-Wheel Vehicles 277

Fig. 6.40 Nonlinear

4-wheeled vehicle model

between the velocity and the vehicle’s transversal angle, V is the velocity vector of
the vehicle, ψ is the yaw angle (rotation round the z axis), f x : is the aggregate force
along the x axis, f y is the aggregate force along the y axis, Tz is the aggregate torque
round the z axis, and δ is the steering angle of the front wheels [349, 360, 535].
The motion of the vehicle is described by the following set of equations:
1. Longitudinal motion

− mV (β̇ + ψ̇)sin(β) + m V̇ cos(β) = f x (6.35)

2. Lateral motion
mV (β̇ + ψ̇)cos(β) + m V̇ sin(β) = f y (6.36)

3. yaw turn
I ψ̈ = Tz (6.37)

The above-described vehicle dynamics can be also written in matrix form

⎛ ⎞⎛ ⎞ ⎛ ⎞
−sin(β) cos(β) 0 mV (β̇ + ψ̇) fx
⎝ cos(β) sin(β) 0⎠ ⎝ m V̇ ⎠ = ⎝ f y⎠ (6.38)
0 0 1 I ψ̈ T z
278 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

Finally, a matrix relation is provided about the transformation of forces on a tire

into forces and torques along the vehicle’s axes:
⎛ ⎞ ⎛ ⎞
fx −sin(δ) 0  
⎝ f y ⎠ = ⎝ cos(δ) Ff
1 ⎠ (6.39)
Fr
Tz L 1 cos(δ) −L 2

6.5.2.2 Vehicle Dynamical Model with Longitudinal and Lateral Forces

The previous model of Fig. 6.40 is rexamined considering that β̇ = 0 and that ψ
is the yaw angle formed between the vehicle’s longitudinal axis and the horizontal
axis of an inertial reference frame. Moreover, it is assumed that apart from the lateral
forces, there are traction torques transferred from the engine to the front wheels as
well as braking torques on the rear and front wheels. Due to the distance between
the wheels axes and the vehicle’s center of gravity, torques are also generated along
the vehicle’s z-axis. With reference to Fig. 6.41, the model of the vehicle’s dynamics
is formulated as follows [349, 360, 535]:

mαx = m(V̇x − ψ̇ V̇y ) = Fx1 + Fx2

mα y = m(V̇y + ψ̇ V̇x ) = Fy1 + Fy2 (6.40)
Iz ψ̈ = Tz 1 + Tz 2

where ax and a y are accelerations along the axes of the inertial reference frame and
V̇x , V̇y in a reference frame that rotates with the yaw rate ψ̇. The forces Fxi , i = 1, 2

Fig. 6.41 Vehicle model with longitudinal and lateral forces

6.5 Velocity Control of 4-Wheel Vehicles 279

on the vehicle’s longitudinal axis and Fyi , i = 1, 2 on the vehicle’s transversal axis
are computed from the horizontal and vertical forces applied on the vehicle’s wheels
as follows:
Fx1 = Fx f cos(δ) − Fy f sin(δ)
Fx2 = Fxr
Fy1 = Fy f sin(δ) + Fy f cos(δ)
(6.41)
Fy2 = Fyr
Tz 1 = L f (Fy f cos(δ) + Fx f sin(δ))
Tz 2 = −L r Fyr

About the longitudinal and the lateral forces applied to the vehicle one has:
1. Longitudinal force on the front wheel
 
1
Fx f = (Ir ω̇ f + Tm − Tb f ) (6.42)
R

2. Longitudinal force on the rear wheel

 
1
Fxr = − (Tbr + Ir ω̇r ) (6.43)
R

3. Lateral force on the front wheel (taking that the angle β between the vehicle’s lon-
Vy +ψ̇ L f
gitudinal axis and the wheel’s velocity vector is approximated by β = Vx )
 
Vy + ψ̇ L f
Fy f = C f δ− (6.44)
Vx

4. Lateral force on the rear wheel (taking that for the rear wheel the steering angle is
δ = 0 and that the angle β between the vehicle’s longitudinal axis and the wheel’s
Vy −ψ̇ L r
velocity vector is approximated by β = Vx ).

Vy − ψ̇ L r
Fyr = −Cr (6.45)
Vx

where C f and Cr are the cornering stiffness coefficients for the front and rear
tires, respectively. Nominal values of these cornering stiffness coefficients can
be estimated through identification procedures. The substitution of Eq. (6.42) to
Eq. (6.45) into Eq. (6.40) results into

Ir Vy +ψ̇ L f
m V̇x = m ψ̇ Vy − R (ω̇r + ω̇ f ) + R (Tm − Tb f − Tbr ) + C f (
1
Vx )δ − C f δ 2
V +ψ̇ L V −ψ̇ L
m V̇y = −m ψ̇ Vx − C f ( y Vx f ) − Cr ( y Vx f ) + ( R1 )(Tm − Tb f )δ + (C f − IRr ω̇ f )δ
Vy +ψ̇ L f Vy −ψ̇ L f L
Iz ψ̈ = −L f C f ( Vx ) + L r Cr ( Vx ) + Rf (Tm − Tb f )δ + L f (Tm − IRr )δ
(6.46)
280 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

The motion of the vehicle along its longitudinal axis is controlled by the traction or
braking wheel torque Tω = Tm − Tb with Tb = Tb f + Tbr and the lateral movement
via the steering angle δ. The two control inputs of the four-wheel vehicle model are

u 1 = Tω
(6.47)
u2 = δ

A first form of the vehicle’s dynamic model is

ẋ = f (x, t) + g(x, t)u + g1 u 1 u 2 + g2 u 22 (6.48)

where ⎛ ⎞
Ir
 m R ( ω̇r + ω̇ f ) 
⎜ (V +L ψ̇) (V −L ψ̇) ⎟
f (x, t) = ⎜ψ̇ Vx + m1 −C f y Vx f − Cr y Vx f ⎟ (6.49)
⎝  ⎠
(Vy +L f ψ̇) (V −L ψ̇)
Iz −L f C f + L r Cr y Vx f
1
Vx

⎛  
1 Cf Vy +L f ψ̇ ⎞
⎜ mR 
m Vx 
⎟
g(x, t) = ⎜
⎝ 0
C f R−Ir ω̇ f
mR
⎟
⎠ (6.50)
(L f C f R−L f Ir ω̇ f )
0 Iz R

⎛ ⎞
0 ⎛ Cf ⎞ ⎛ ⎞
−m Vx  
⎜ 1 ⎟ u1
g1 = ⎜ ⎟ g
⎝ mR ⎠ 2 = ⎝ 0 ⎠ x = ⎝ V y ⎠ u = (6.51)
u2
Lf 0 ψ̇
Iz R

The previously analyzed nonlinear model of the vehicle’s dynamics can be simplified
if the control inputs u 1 u 2 and u 22 are not taken into account. In the latter case, the
dynamics of the vehicle takes the form

ẋ = f (x, t) + g(x, t)u (6.52)

6.5.3 Flatness-Based Controller for the 3-DOF Vehicle Model

6.5.3.1 Flatness-Based Controller for the 4-Wheel Vehicle

To show that the four-wheel vehicle is differentially flat, the following flat outputs
are defined [349, 350]:
y1 = Vx
(6.53)
y2 = L f mVy − Iz ψ̇
6.5 Velocity Control of 4-Wheel Vehicles 281

Then it holds that all elements of the system’s state vector can be written as functions
of the flat outputs and their derivatives. Indeed, for x = [Vx , Vy , ψ̇]T it holds

Vx = y1 (6.54)
  
y2 Iz L f m y1 ẏ2 + Cr (L f + L r )y2
Vy = − (6.55)
Lfm Lfm Cr (L f + L r )(Iz − L f L r m) + (L f m y1 )2

L f m y1 ẏ2 + Cr (L f + L r )y2
ψ̇ = (6.56)
Cr (L f + L r )(Iz − L f L r m) + (L f m y1 )2

Expressing the system’s state variables as functions of the flat outputs, one has the
following state-space description for the system
   
ẏ1 u
= Δ(y1 , y2 , ẏ2 ) 1 + Φ(y1 , y2 , ẏ2 ) (6.57)
ÿ2 u2

where  
Δ11 (y1 , y2 , ẏ2 ) Δ12 (y1 , y2 , ẏ2 )
Δ(y1 , y2 , ẏ2 ) = (6.58)
Δ21 (y1 , y2 , ẏ2 ) Δ22 (y1 , y2 , ẏ2 )

with
1
Δ11 (y1 , y2 , ẏ2 ) = (6.59)
mR
Cf Vy +L f ψ̇
Δ12 (y1 , y2 , ẏ2 ) = m ( y1 ) (6.60)

Cr (L f +L r )(Vy −L r ψ̇)−L f m ψ̇ y12

Δ21 (y1 , y2 , ẏ2 ) = m R y12
(6.61)

 
L r Cr (L f +L r ) (L f C f R−L f Ir ω̇ f )
Δ22 (y1 , y2 , ẏ2 ) = −L f m y1 + y1 Iz R +
((Cr (L f +L r ))(Vy −L r ψ̇)−L f m ψ̇ y12 ) C f (Vy +L f ψ̇) Cr (L f +L r ) RC f −Ir ω̇ f (6.62)
+ y2
· m y1 − y1 mR
1

Moreover, about matrix Φ(y1 , y2 , ẏ2 ) it holds

 
Φ1 (y1 , y2 , ẏ2 )
Φ(y1 , y2 , ẏ2 ) = (6.63)
Φ2 (y1 , y2 , ẏ2 )

with elements
Ir
Φ1 (y1 , y2 , ẏ2 ) = ψ̇ Vy − m R (ω̇r + ω̇ f ) (6.64)

Cr (L f +L r )
Φ2 (y1 , y2 , ẏ2 ) = −L f m y1 f 3 (x, t) − y1 f 2 (x, t)+
C f (L f +L r )(Vy −L r ψ̇)−L f m ψ̇ y12 L r Cr (L f +L r ) (6.65)
+ y2
f 1 (x, t) + y1 f 3 (x, t)
1
282 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

According to the above, the system’s control input can be also written as a function
of the flat output and its derivatives. Thus one has
   
ẏ1 u
= Δ(y1 , y2 , ẏ2 ) 1 + Φ(y1 , y2 , ẏ2 ) (6.66)
ÿ2 u2

i.e.,    
u1 ẏ
= Δ−1 (y1 , y2 , ẏ2 )−1 ( 1 − Φ(y1 , y2 , ẏ2 )) (6.67)
u2 ẏ2

which means that provided that matrix Δ(y1 , y2 , ẏ2 ) is invertible, the control input
u = [u 1 , u 2 ]T can be written as a function of the flat output and its derivatives. The
nonsingularity of matrix Δ(y1 , y2 , ẏ2 ) depends on the determinant

(Ir ω̇ f −C f R)(L 2f y12 m 2 −Cr (L f +L r )L r L f m+Cr Iz L r )

det (Δ(y1 , y2 , ẏ2 )) = Iz R 2 y1 m 2
(6.68)

This determinant has nonzero values because it holds:

(i) (Ir ω̇ f − C f R) = 0 since for the wheels rotational acceleration one has ω̇ f <
Cf R
Ir , and also
(ii) (L 2f y12 m 2 − Cr (L f + L r )L r L f m + Cr Iz L r ) = 0 when Iz > L f m.
The differentially flat model of the vehicle can be also written in a canonical form
after defining the new control input vector
   
v1 u
= Δ(y1 , y2 , ẏ2 ) 1 + Φ(y1 , y2 , ẏ2 ) (6.69)
v2 u2

thus one obtains a MIMO system description into canonical form, i.e.,
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 000 y1 1 0  
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0 v
0⎠ 1 (6.70)
v2
ÿ2 000 ẏ2 0 1

Once the vehicle’s model is written in the differentially flat form, the controller that
ref ref ref
enables tracking of a desirable trajectory defined by y1 , y2 , ẏ2 is given by
ref ref
v1 = ẏ1 − k p1 (y1 − y1 )
ref ref ref (6.71)
v2 = ÿ2 − kd2 ( ẏ2 − ẏ2 ) − k p2 (y2 − y2 )

ref ref
and defining the error variables e1 = y1 − y1 and e2 = y2 − y2 one has the
following tracking error dynamics for the closed-loop system

ė1 + k p1 e1 = 0
(6.72)
ë2 + kd2 ė2 + k p2 e2 = 0
6.5 Velocity Control of 4-Wheel Vehicles 283

Therefore, the suitable selection of gains k p1 >0 and k p2 > 0, kd2 > 0 assures
the asymptotic elimination of the tracking errors, i.e., lim t→∞ e1 (t) = 0 and
lim t→∞ e2 (t) = 0.
The control input that is finally applied for the vehicle’s steering is given by
   
u1 −1 −1 v
= Δ (y1 , y2 , ẏ2 ) ( 1 − Φ(y1 , y2 , ẏ2 )) (6.73)
u2 v2

or equivalently
  
ref ref
u1 ẏ1 − k p1 (y1 − y1 )
= Δ−1 (y1 , y2 , ẏ2 )−1 [ r e f ref ref − Φ(y1 , y2 , ẏ2 )]
u2 ÿ2 − kd2 ( ẏ2 − ẏ2 ) − k p2 (y2 − y2 )
(6.74)

The transformation of the vehicle’s model into a canonical form, through the appli-
cation of differential flatness theory, facilitates not only the design of a feedback
controller for trajectory tracking but also the design of filters for the estimation of
the state vector of the vehicle out of a limited number of sensor measurements.

6.5.4 Estimation of Vehicle Disturbance Forces

with Kalman Filtering

6.5.4.1 State Estimation with the Derivative-Free

Nonlinear Kalman Filter

To account for model uncertainties and external disturbances, observer-based esti-

mation has been proposed in various forms (disturbance observers, extended state
observers, etc.) [85, 107, 108, 354].
It was shown that the initial nonlinear model of the vehicle can be written in the
MIMO canonical form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 000 y1 10  
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0 0⎠ v1 (6.75)
v2
ÿ2 000 ẏ2 01

Thus one has a MIMO linear model of the form

ẏ f = A f y f + B f v
(6.76)
z f = Cf yf

where y f = [y1 , y2 , ẏ2 ]T and matrices A f ,B f ,C f are in the MIMO canonical form
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
000 1 0 1 0
A f = ⎝0 0 1⎠ B f = ⎝0 0⎠ C Tf = ⎝0 1⎠ (6.77)
000 0 1 0 0
284 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

where the measurable variables y1 = Vx , y2 = L f mVy − Iz ψ̇ are associated with

the linear velocity of the vehicle Vx , Vy and with its angular velocity ψ̇. For the
aforementioned model, and after carrying out discretization of matrices A f , B f , and
C f with common discretization methods, one can perform linear Kalman filtering
using Eqs. (4.5) and (4.6). This is Derivative-free nonlinear Kalman filtering for the
model of the vehicle which, unlike EKF, is performed without the need to compute
Jacobian matrices and does not introduce numerical errors due to approximative
linearization with Taylor series expansion.

6.5.4.2 Kalman Filter-Based Estimation of Disturbances

It is assumed that disturbance forces affect the nonlinear vehicle model along its
longitudinal and transversal axis and that disturbance torques affect the nonlinear
vehicle model on its z axis. For example, disturbance forces be due to a force vector
that coincides with the vehicle’s longitudinal axis (e.g., traction disturbance) or dis-
turbance torques can be due to unmodeled lateral forces. These disturbance forces
and torques change dynamically in time and their dynamics is given by

d̃x = f dx (Vx , Vy , ψ̇)

d̃ y = f d y (Vx , Vy , ψ̇) (6.78)
d̃ψ = Tdψ (Vx , Vy , ψ̇)

Since the state variables of the vehicle’s dynamic model can be written as functions
of the flat outputs y1 and y2 and of their derivatives it also holds

(i) (i)
d̃x = f dx (y1 , y2 , ẏ2 )
(i) (i)
d̃ y = f d y (y1 , y2 , ẏ2 ) (6.79)
(i) (i)
d̃ψ = Tdψ (y1 , y2 , ẏ2 )

where i = 1, 2, . . . stands for the ith order derivative of the disturbance variable.
Considering the effect of disturbance functions on the initial nonlinear state equa-
tion of the vehicle and the linear relation between the initial state variables [Vx , Vy ]
and the state variables of the flat system description [y1 , y2 ] one has the appearance
of the disturbance terms in the canonical form model of Eq. (6.70)
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0  
1
ẏ1 000 y1 1 d̃x
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0 v ⎜ m
⎟
0⎠ 1 + ⎝ 0 ⎠ (6.80)
v2
ÿ2 000 ẏ2 0 1 L f d̃˙y − d̃˙ψ

Next, the state vector of the model of Eq. (7.66) is extended to include as additional
state variables the disturbance forces d̃x , d̃ y , and d̃ psi . Then, in the new state-space
6.5 Velocity Control of 4-Wheel Vehicles 285

description one has z 1 = y1 , z 2 = y2 , z 3 = ẏ2 , z 4 = f˜a = m1 d̃x , z 5 = f˙˜a ,

z 6 = f˙˜b = L f d̃˙y − d̃˙ψ , z 7 = f¨˜b , which takes the form of matrix equations

ż = ÷z + B̃·ṽ (6.81)

where the control input is

 T
ṽ = v1 v2 1 ¨ (3) (3)
L f d̃ y − d̃ψ
m d̃x
or
 T (6.82)
ṽ = v1 v2 f¨˜a f˜b(3)

with ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0 1 0 00 100 0 10
⎜0 0 1 0 0 0 0⎟ ⎜0 0 0 0⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 1 0⎟ ⎜0 1 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ T ⎜ ⎟
à = ⎜
⎜0 0 0 0 1 0 0⎟ B̃ = ⎜
⎟
⎜0 0 0
⎟
0⎟ C̃ = ⎜
⎜0 0⎟
⎟ (6.83)
⎜0 0 0 0 0 0 0⎟ ⎜0 0 1 0⎟⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎜0 0⎟
⎝0 0 0 0 0 01 ⎠ ⎝0 0 0 0 ⎠ ⎝0 0⎠
0 0 0 0 0 00 000 1 00

where the measurable state variables are z 1 and z 2 . Since the dynamics of the dis-
turbance terms f˜a and f˜b are taken to be unknown in the design of the associated
disturbances’ estimator one has the following dynamics:

ż o = Ão ·z + B̃o ·ṽ + K (Co z − Co ẑ) (6.84)

where K ∈R 7×2 is the state estimator’s gain and

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 0 0 1 0 0 0 1000 10
⎜0 0 1 0 0 0 0⎟ ⎜0 0 0 0⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 1 0⎟ ⎜0 1 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ T ⎜ ⎟
Ão = ⎜
⎜0 0 0 0 1 0 0⎟ B̃
⎟ o ⎜= ⎜0 0 0 0 ⎟ C̃
⎟ o = ⎜0 0⎟
⎜ ⎟ (6.85)
⎜0 0 0 0 0 0 0⎟ ⎜0 0 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 1⎠ ⎝0 0 0 0⎠ ⎝0 0⎠
0 0 0 0 0 0 0 0000 00

Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents of matrices Ão , B̃o , and
C̃o , respectively, a Derivative-free nonlinear Kalman Filter can be designed for the
aforementioned representation of the system dynamics [436, 437]. The associated
Kalman Filter-based disturbance estimator is given by
286 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

measurement update:

K (k) = P − (k)C̃dT [C̃d ·P − (k)C̃dT + R]−1

x̂(k) = x̂ − (k) + K (k)[z(k) − C̃d x̂ − (k)] (6.86)
P(k) = P − (k) − K (k)C̃d P − (k)

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

(6.87)
x̂ − (k + 1) = Ãd (k)x̂(k) + B̃d (k)ṽ(k)

To compensate for the effects of the disturbance forces it suffices to use in the control
loop the modified control input vector
⎛ ⎞
ˆ
˜  
v1 − f a
v=⎝ ⎠ or v = v1 − ẑ 4 (6.88)
ˆ v2 − ẑ 6
v2 − f˙˜b

6.5.5 Simulation Tests

To evaluate for the performance of the proposed nonlinear control scheme, as well as
for the performance of the Kalman Filter-based disturbances, estimator simulation
experiments have been carried. Different velocity setpoints have been assumed (for
velocity along the horizontal and vertical axis of the inertial reference frame, as well
as for angular velocity round the vehicle’s z axis). Moreover, different disturbance
forces and torques have been assumed to affect the vehicles’ dynamic model. Using
the representation of the vehicle’s dynamics given in Eqs. (6.48)–(6.51), two gener-
alized disturbance forces/torques have been considered: the first denoted as f˜a was
associated with state variable y1 , while the second one denoted as f˜b was associated
with the state variable y2 . It was also assumed that the change in time of the gen-
eralized forces and torques was defined by the second derivative of the associated
variable, i.e., f¨˜a and f¨˜b . The disturbances dynamics was completely unknown to
the controller and their identification was performed in real time by the disturbance
estimator. The control loop used in the vehicle’s autonomous navigation is given in
Fig. 6.42.
The measurable variables used by the control and disturbances’ estimation scheme
were the vehicle’s velocity Vx along the longitudinal axis, the vehicle’s velocity Vy
along the lateral axis and the vehicle’s yaw rate ψ̇. The first two variables can be
obtained with the use of onboard accelerometers, while the third variable can be
obtained with the use of a gyrocompass. The longitudinal axis of the vehicle is
6.5 Velocity Control of 4-Wheel Vehicles 287

∧ ∧
∼ ∼

Fig. 6.42 Control loop for the autonomous vehicle comprising a flatness-based nonlinear controller
and a Kalman Filter-based disturbances estimator

(a) 3 (b) 3

2.5 2.5

2 2

1.5 1.5
Velx

Vely

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 6.43 Vehicle control under disturbances profile 1: a convergence of x-axis velocity Vx to the
desirable setpoint, b convergence of the y-axis velocity Vy to the desirable setpoint

denoted as x-axis, while the lateral axis of the vehicle is denoted as y-axis. As it
can be seen in Figs. 6.43, 6.44, 6.45, 6.46, 6.47, 6.48, 6.49, and 6.50, the proposed
nonlinear controller succeeded accurate tracking of velocity setpoints. Moreover,
the efficient estimation of disturbance forces and torques that was succeeded by the
Kalman Filter-based disturbance estimator enabled their compensation through the
inclusion of an additional control term in the loop.
288 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 1 (b) 15 1

/dt
10

est
est
0.5

dfa/dt − dfa
fa − fa
0.5 5
0
0
0 −5 −0.5
Vel ψ

0 20 40 0 20 40
time time

/dt 2
−0.5 4 1.5

/dt

est
1

est
2

df 2 b/dt 2− df 2 b
dfb/dt − dfb
−1 0.5
0
0

−1.5 −2 −0.5
0 5 10 15 20 25 30 35 40 0 20 40 0 20 40

time time time

Fig. 6.44 Vehicle control under disturbances profile 1: a convergence of yaw rate ψ̇ to the desirable
setpoint, b estimation of the disturbance terms and of their rate of change

(a) 3 (b) 3

2.5 2.5

2 2

1.5 1.5
Velx

Vely

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 6.45 Vehicle control under disturbances profile 2: a convergence of x-axis velocity Vx to the
desirable setpoint, b convergence of the y-axis velocity Vy to the desirable setpoint

6.6 Active Vehicle Suspension Control

6.6.1 Overview

Another application area of differential flatness theory-based control and filtering for
intelligent vehicles design is active suspension control. In the recent years there has
been systematic effort towards designing vehicles of improved safety and comfort
and to this end the development of active suspension control systems has been an
6.6 Active Vehicle Suspension Control 289

(a) 1 (b)10 1

/dt
est
fa − fa est
0 0

dfa/dt − dfa
0.5
−10 −1

0 −20 −2
Vel ψ

0 10 20 30 40 0 10 20 30 40
time time

df 2 b/dt 2 − df 2 best /dt 2

−0.5 20 2

/dt
0 0

est
dfb/dt − dfb
−1 −20 −2

−40 −4

−1.5 −60 −6
0 5 10 15 20 25 30 35 40 0 10 20 30 40 0 10 20 30 40
time time
time

Fig. 6.46 Vehicle control under disturbances profile 2: a convergence of yaw rate ψ̇ to the desirable
setpoint, b estimation of the disturbance terms and of their rate of change

(a) 3 (b) 3

2.5 2.5

2 2

1.5 1.5
Velx

Vely

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 6.47 Vehicle control under disturbances profile 3: a convergence of x-axis velocity Vx to the
desirable setpoint, b convergence of the y-axis velocity Vy to the desirable setpoint

important research topic. One can note several results about active suspension control
systems exhibiting robustness to external disturbance forces and being capable of
efficiently suppressing the vibrations induced to the vehicle by these disturbances.
H∞ controllers have been developed taking into account worst case disturbances
on the suspension models [134, 165, 579]. Moreover, there have been results on
operating the suspension’s control loop under limited information provided by a
small number of onboard sensors. This can be seen in the case of developing some
type of state estimator or statistical filter to approximate the nonmeasurable ele-
ments of the suspension’s state vector and the unknown disturbance forces. Par-
ticularly, one can note the use of Linear Quadratic Gaussian (LQG) control where
Kalman Filtering is combined with an optimal controller [193, 196, 206, 269, 341].
290 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

1 4 1
(a) (b)

dfa/dt − dfa est /dt

fa − faest
2 0.5
0.5
0 0

0
−2 −0.5
Vel ψ

0 20 40 0 20 40
time time
30 1.5

2
−0.5

df 2 b/dt 2 − df 2 best /dt

dfb/dt − dfbest /dt
20 1

−1 10 0.5

0 0

−1.5 −10 −0.5

0 5 10 15 20 25 30 35 40 0 20 40 0 20 40
time time time

Fig. 6.48 Vehicle control under disturbances profile 3: a convergence of yaw rate ψ̇ to the desirable
setpoint, b estimation of the disturbance terms and of their rate of change

(a) 3 (b) 3

2.5 2.5

2 2

1.5 1.5
Velx

Vely

1 1

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 6.49 Vehicle control under disturbances profile 4: a convergence of x-axis velocity Vx to the
desirable setpoint, b convergence of the y-axis velocity Vy to the desirable setpoint

Moreover in [207] the application of a sliding-mode controller together with Kalman

Filtering has been proposed for implementing state estimation-based control of the
suspension’s model. Additionally, disturbance observers have been used for simul-
taneous estimation of the suspension’s state vector and of the unknown external
disturbances. The suitability of disturbance observers for vibration control problems
has been shown in [26], while the efficiency of disturbance estimators in vehicle con-
trol loops and especially in the suspension control problem has been demonstrated
in [40, 121, 123, 248, 249, 491]. Finally, a scheme of distributed Kalman Filtering
has been applied to disturbances and state vector estimation for the suspension’s
mechanism in [270].
6.6 Active Vehicle Suspension Control 291

(a) 1 (b) 5 1

/dt
est
0.5

fa − fa est
0

dfa/dt − dfa
0.5 0
−5
−0.5
0 −10 −1
0 20 40 0 20 40
Vel ψ

time time

/dt 2
−0.5 15 1.5

/dt

est
10 1

df 2 b/dt 2 − df 2 b
est
dfb/dt − dfb
−1 5 0.5

0 0

−1.5 −5 −0.5
0 5 10 15 20 25 30 35 40 0 20 40 0 20 40
time time time

Fig. 6.50 Vehicle control under disturbances profile 4: a convergence of yaw rate ψ̇ to the desirable
setpoint, b estimation of the disturbance terms and of their rate of change

In this section an approach to solve the problem of active control of vehicle suspen-
sions is developed with the use of flatness-based controller and a Kalman Filter-based
disturbances estimator. First, dynamic analysis for the vehicle’s suspension model
is provided. Active vehicle suspension control systems are underactuated and the
efficient suppression of disturbance inputs (e.g., due to rough road surface) is impor-
tant for attaining the performance objectives of the control loop. The elements of
the state vector are variables denoting the displacement of the sprung and unsprung
masses from their zero position and variables denoting the linear velocities of these
masses. The control inputs to the model are the force generated by the actuator
placed between the two masses (which aims at the suppression of vibrations) and the
unknown disturbance force that is generated due to contact of the tire with the road
surface. The model can take the form of a linear state-space equation. Moreover, by
assuming nonlinearities in the spring and damper terms of the suspension a nonlinear
dynamical model is obtained.
Next, it is shown how a controller for the aforementioned suspension model can
be obtained through the application of differential flatness theory [152, 465, 516,
535]. The flat output for the suspension’s model is a scalar variable which is equal
to the weighted sum of the elements of the suspension’s state vector. By expressing
all state variables and the control input of the suspension model as functions of the
flat output and its derivatives the system’s dynamic model is transformed into the
linear Brunovksy (canonical) form [317, 479]. For the latter model it is possible
to design a state feedback controller that enables accurate tracking of the vehicle’s
velocity setpoints. However, since measurements are available only for certain ele-
ments of the transformed state vector, to implement a state feedback control loop
the rest of the elements of the suspension’s transformed state vector have to be esti-
mated with the use of an observer or filter. To this end the concept of Derivative-free
292 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

nonlinear Kalman Filtering is employed. By avoiding linearization approximations,

the proposed filtering method succeeds better estimation of the system’s state vari-
ables [422, 427].
A particular difficulty, in the case of state estimation of the suspension’s model is
the existence of the unmodeled disturbance forces. It is shown that it is possible to
redesign the Kalman Filter in the form of a disturbance observer and using the esti-
mation of the disturbance to develop an auxiliary control input that compensates for
the disturbances effects [85, 107, 108, 183, 354]. In this way the suspension’s control
system can become robust with respect to uncertainties in the model’s parameters
or uncertainties about external forces. It is also noted that in terms of computation
speed the proposed Kalman Filter-based disturbance estimator is faster than distur-
bance estimators that may be based on other nonlinear filtering approaches (e.g.,
Extended Kalman Filter, Unscented Kalman Filter or Particle Filter) thus becoming
advantageous for the real-time estimation of the unknown suspension dynamics.
The efficiency of the proposed control and Kalman Filter-based disturbances esti-
mation scheme is evaluated through numerical simulation tests. It will be shown
that the accurate estimation of the disturbance forces which are exerted on the sus-
pension enables their efficient compensation. This is achieved by introducing an
additional element in the controller that produces a counterdisturbance input based
on the estimated value for the disturbance variable. This control scheme finally results
in minimizing the effects of the disturbances on the vehicle’s parts.

6.6.2 Dynamic Model of Vehicle Suspension

6.6.2.1 Dynamics of the 2-DOF Suspension

The suspension system is depicted in Fig. 6.51 and its dynamics is written as

m 1 ẍ1 + c1 (ẋ1 − ẋ2 ) + k1 (x1 − x2 ) = f

m 2 ẍ2 + c1 (ẋ2 ) − ẋ1 + k1 (x2 − x1 ) + (6.89)
+ c2 (ẋ2 − ζ̇ ) + k2 (x2 − ζ ) = − f

Variable x1 denotes the sprung mass displacement while variable x2 denotes the
unsprung mass displacement. Tyre’s deflection ζ and its time derivatives ζ (i) . i =
1, 2, . . . represent unknown external disturbance inputs due to road surface roughness
and are assumed to be bounded. The mass that needs regulation is the sprung mass
m 1 which is also considered to be larger than m 2 . The control force f is generated
by an actuator placed between the two masses (see Fig. 6.51).
A normalization is performed to the model  through the following procedure: (i)
k1
the normalized time is defined as τ = t m1 ,
(ii) the normalized input force is

u = kf1 . The system constant coefficients are redefined as ε = mm1
2
, γ 1 = c1
m1
m1
k1 ,
6.6 Active Vehicle Suspension Control 293

Fig. 6.51 An active vehicle

suspension system where
both the sprung and the
unsprung mass are connected
to a spring and a damper


γ2 = mc21 mk11 , κ = k2
k1 . Thus, the dynamics model of the vehicle’s suspension can
be rewritten as
ẍ1 + γ1 (ẋ1 − ẋ2 ) + (x1 − x2 ) = u
ε ẍ2 + γ1 (ẋ2 − ẋ1 ) + (x2 − x1 ) +
(6.90)
+ γ2 (ẋ2 − ζ̇ ) + κ(x2 − ζ ) = −u
y = x1

The model of Eq. (6.90) can be also written is state-space form after defining the
state variables z 1 = x1 , z 2 = ẋ1 , z 3 = x2 , z 4 = ẋ2 . Thus one has
⎛ ⎞ ⎛ ⎞⎛ ⎞
ż 1 0 1 0 0 z1
⎜ż 2 ⎟ ⎜−1 −γ1 1 γ ⎟ ⎜z 2 ⎟
⎜ ⎟=⎜ 1 ⎟⎜ ⎟+
⎝ż 3 ⎠ ⎝ 0 0 0 1 ⎠ ⎝z 3 ⎠
γ1 γ1 +γ2
ε − ⎛ε − ⎞ε
ż 4 1 1+κ z4
⎛ε ⎞ (6.91)
0 0 0  
⎜ 1 ⎟ ⎜0 0 ⎟ ζ
+⎜ ⎟ ⎜ ⎟
⎝ 0 ⎠ u + ⎝ 0 0 ⎠ ζ̇
κ γ2
− 1ε ε ε

where all terms associated with disturbance ζ can be represented by the new variable
Δ. Thus one obtains
d
xs = Axs + Bu + Δ (6.92)
dτ
294 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

6.6.2.2 A Nonlinear Model of Vehicle Suspension Dynamics

The dynamical model of the two-degrees of freedom vehicle suspension (see

Fig. 6.52) is given as follows [491]

m s z̈ s + Fsd + Fsk = FA
(6.93)
m u z̈ u − Fsk − Fsd + K t (z u − zr ) = −FA

where FA is the force generated by the actuator, Fsk is the force associated with the
suspension’s spring term, Fsc is the force associated with the suspension’s damper
term and Ft = kt (z u − zr ) is a spring force associated with elasticity coefficient kt
and denoting the spring-type behavior of the wheel when in contact with the road’s
surface.

Fig. 6.52 An active

suspension system
6.6 Active Vehicle Suspension Control 295

It holds that
Fsk (z s , z u ) = ks (z s − z u ) + kn s (z s − z u )3 (6.94)

Fsc (z s , z u ) = bs (ż s − ż u )+
(6.95)
+bns (ż s − ż u )2 sgn(ż s − ż u )

Denoting the state variables x1 = z s , x2 = ż s , x3 = z u , x4 = ż u one has

ẋ1 = x2
ẋ2 = − m1s [ks (x1 − x3 ) + kn s (x1 − x3 )3 +
+bs (x2 − x4 ) + bns (x2 − x4 )2 sgn(x2 − x4 )] + 1
ms u

ẋ3 = x4 (6.96)

ẋ4 = − mktu x3 + m1u [ks (x1 − x3 )+

+kn s (x1 − x3 )3 + bs (x2 − x4 )+
kt
+bns (x2 − x4 )2 sgn(x2 − x4 )] − m1u u + m u zr

1
where the term m u zr can be considered as a disturbance term. Denoting the nonlinear
functions

f 1 (x, t) = − m1s [ks (x1 − x3 ) + kn s (x1 − x3 )3 +

(6.97)
+bs (x2 − x4 ) + bns (x2 − x4 )2 sgn(x2 − x4 )]

g1 (x, t) = 1
ms (6.98)

f 2 (x, t) = − mktu x3 + m1u [ks (x1 − x3 ) + kn s (x1 − x3 )3 +

(6.99)
+bs (x2 − x4 ) + bns (x2 − x4 )2 sgn(x2 − x4 )]

g2 (x, t) = − m1u (6.100)

one has the following state-space description

ẋ1 = x2
ẋ2 = f 1 (x, t) + g1 (x, t)u
(6.101)
ẋ3 = x4
ẋ4 = f 2 (x, t) + g2 (x, t)u
296 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

6.6.3 Flatness-Based Control for a Suspension Model

6.6.3.1 A Flatness-Based Controller for the Vehicle Suspension Model

It can be proven that the suspension model is differentially flat by defining the fol-
lowing flat output [491]
⎛ ⎞
z1
  −1 ⎜z 2 ⎟
F = 0 0 0 1 Co ⎜ ⎟
⎝z 3 ⎠
(6.102)
z4
εγ2 εκ−γ22 ε ε2 γ2
i.e., F = κε x1 − κ2 1
ẋ + κ2
x2 − κ2 2
ẋ

where Co stands for the system’s controllability matrix.

Co = [B, AB, A2 B, A3 B] (6.103)

It can be shown that all state variable of the system and the control input can be written
as functions of the flat output and its derivatives. Indeed for the unperturbed system,
i.e., if from the model of Eq. (6.102) the disturbance input ζ and its derivatives ζ̇ are
ignored it holds
εγ2 (εκ−γ22 )ε ε2 γ2
F = κε x1 − κ2 1
ẋ + κ2
x2 − κ2 2
ẋ
εγ2
Ḟ = κε ẋ1 + κ x2 + ε2
κ ẋ 2 (6.104)
F̈ = −εx2
F (3) = ε ẋ2

while one also has

κ γ2
x1 = εF + ε Ḟ + F̈
κ γ2
x2 = ε Ḟ + ε F̈ + F (3)
−1 (6.105)
x3 = ε F̈
−1 (3)
x4 = ε F

while with the use of u = ẍ1 + γ1 (ẋ1 − ẋ2 ) + (x1 − x2 ) it can be also concluded that
the control input is a function of the flat output and its derivatives.
Taking also into account the effects of the disturbance input ζ and of its derivative
ζ̇ the flat output and its derivative are formulated as follows

εγ2 εκ−γ22 ε
x2 − εκγ2 2 ẋ2
2
F = κε x1 − κ2 1
ẋ + κ2
εγ2 εγ 2
Ḟ = κε x1 − + εκ ẋ2 − εγκ 2 ζ (τ ) − κ 22 ζ̇ (τ )
2
κ2 2
x
εγ 2
(6.106)
F̈ = −εx2 + εζ (τ ) + εγ2 (1 − κ1 )ζ̇ (τ ) − κ 22 ζ̈ (τ )
εγ 2
F (3) = −ε ẋ2 + εζ̇ (τ ) + εγ2 (1 − κ1 )ζ̈ (τ ) − κ 22 ζ (3) (τ )
6.6 Active Vehicle Suspension Control 297

Differentiating one more time the flat output one obtains

F (4) = −x1 − γ1 ẋ1 + (1 + κ)x2 +

+(γ1 + γ2 )ẋ2 + u − κζ (τ ) − γ2 ζ̇ (τ )+
+εζ̈ (τ ) + εγ2 (1 − κ1 )ζ (3)(τ ) − (6.107)
γ22
−ε κ2
ζ (4) (τ )

Aggregating all terms other than u into one variable

φ(τ ) = x1 − γ1 ẋ1 + (1 + κ)x2 + (γ1 + γ2 )ẋ2 −

−κζ (τ ) − γ2 ζ̇ (τ ) + εζ̈ (τ )+ (6.108)
εγ22 (4)
+εγ2 (1 − κ1 )ζ (3) (τ ) − κ2
ζ (τ )

one has the system dynamics

F (4) = u + φ(τ ) (6.109)

or equivalently, in state-space form

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
Ḟ1 0 1 0 0 F1 0
⎜ Ḟ2 ⎟ ⎜0 0 1 0⎟ ⎜ F2 ⎟ ⎜0⎟  
⎜ ⎟=⎜ ⎟ ⎜ ⎟ + ⎜ ⎟ u + φ(τ ) (6.110)
⎝ Ḟ3 ⎠ ⎝0 0 0 1⎠ ⎝ F3 ⎠ ⎝0⎠
Ḟ4 0 0 0 0 F4 1

with the state variable Fi (τ ), i = 1, . . . , 4 to stand for the (i − 1)th order derivative
F (i−1) (t) of the flat output. The estimation of the term φ(τ ) by a disturbance observer
enables to design a controller for the vehicle’s suspension model as follows

(4) (3)
u(τ ) = Fd (τ ) − k1 (F (3) (τ ) − Fd (τ ))−
−k2 ( F̈(τ ) − F̈d (τ ) − k3 ( Ḟ(τ )− (6.111)
− Ḟd (τ )) − k4 (F(τ ) − Fd (τ )) − φ̂(τ )

6.6.4 Compensating for Model Uncertainty with the Use

of the H∞ Kalman Filter

The Kalman Filter for the canonical form of the suspension model given in Eq. (6.110)
can be redesigned to cope with the case of maximum errors of some linear combi-
nation of states for worst case assumptions of process noise, measurement noise
and disturbances. This can be useful in state estimation for the vehicle suspension
model, as a method for model uncertainty compensation. Filters designed to mini-
mize a weighted norm of state errors are called H∞ or minimax filters [177, 490].
298 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

The discrete-time H∞ filter uses the same state-space model as the Kalman Filter,
which has the form

x(k + 1) = A(k)x(k) + B(k)u(k) + w(k)

(6.112)
z(k) = C(k)x(k) + v(k)

E[w(k)] = 0, E[w(k)w(k)T ] = Q(k)δi j, E[v(k)] = 0, E[v(k)v(k)T ] = R(k)δi j

and E(w(k)v(k)T ) = 0. The update of the state estimate is again given by

x̂(k) = x̂ − (k) + K (k)(z(k) − C(k)x̂ − (k)) (6.113)

that minimizes the trace of the covariance matrix of the state vector estimation error
1 1
J= E{x̃(k)T ·x̃(k)} = tr (P − (k)) (6.114)
2 2

where x̃ − (k) = x(k) − x̂ − (k) and P − (k) = E[x̃ − (k)T · x̃ − (k)]. The H∞ filtering
approach defines first a transformation

d(k) = L(k)x(k) (6.115)

where L(k)∈R n×n is a full rank matrix. The use of the transformation given in
Eq. (6.115) allows certain combinations of states to be given more weight than others.
Next, defining the estimation error variable d̃1 (i) = d(i) − d̂(i), the cost function of
the H∞ filter is initially formulated as
k−1
J (k) = + 1)T S(i)d̃(i + 1)/b
i=0 d̃(i
k−1 T
b = x̃ − (0)T P − (0)−1 x̃ − (0) + i=0 w (i + 1)Q(i + 1)−1 w(i + 1)+ (6.116)
k−1 T
+ i=0 v (i)R(i)−1 v(i)

where Si is a positive-definite symmetric weighting matrix. It can be observed that

both matrices S(k) and L(k) appear in the cost function and thus affect the solution
x̂ − (k + 1) of the optimization problem. The objective is to find state vector estimates
x̂ − (k) and x̂(k) that keep the cost function below a given value 1 θ for worst case
conditions, i.e.,
1
J (k) < (6.117)
θ
By rewriting Eq. (6.116) and substituting Eq. (6.112) a modified cost functional is
obtained
k−1
Ja (k) = − θ1 x̃ − (0) P − (0)x̃ − (0) +
T
i=0 Γ (i)

Γ (i) = (x(i + 1) − x̂ − (i + 1))T Wi (x(i + 1) − x̂ − (i + 1))−

− θ (w (i + 1)Q(i + 1)−1 w(i + 1) + (y(i) − C(i)x − (i))T R(i)−1 (y(i) − C(i)x − (i)))
1 T
(6.118)
6.6 Active Vehicle Suspension Control 299

and
W (i) = L(i)T S(i)L(i) (6.119)

This cost function does not include the dynamic model of the system given in
Eq. (6.112) and this is added by using a vector of Lagrange multipliers λ(i + 1).
This gives
J (k) = − θ1 x̃ − (0)T P − (0)x̃ − (0)+
k−1
(Γi + 2 λ(i+1)
T
+ i=0 θ )(A(i)x̂(i) + B(i)u(i)+ (6.120)
2λ(0)T 2λ(0)T
+w(i) − x(i + 1)) + θ x(0) − θ x(0)

The cost function of the filter given in Eq. (6.120) can be used as the basis for
the solution. It is aimed to find equations defining x̂ − (k + 1), or equivalently a
measurement weighting matrix (similar to the Kalman gain matrix), that minimize
the cost for worst case assumptions about x(0), w(i) and y(i). Thus, the optimization
objective is formulated as

min max
J ∗ (k) = J (k) (6.121)
xi x(0), w(i), y(i)

It is noted that the estimation algorithm uses knowledge of the output measurement
y(i) but receives no information about the initial conditions of the system x(0)
and the process noise w(i). Under this assumption, the estimation should be able
to compensate for worst case values for the unknown parameters. This is a game
theoretic problem that is solved in two steps.
In the first step of optimization, partial derivatives of J (k) with respect to x(0),
w(i) and λ(i) are set to zero so as to maximize the cost function of Eq. (6.120),
now being dependant only on the terms x̂ − (k + 1) and y(k) which are included
in Γi . In the second step of optimization, the partial derivatives of J (k) with
respect to x̂ − (k + 1) and y(k) are set to zero, to obtain a condition for the filter’s
gain matrix that minimizes this cost functional. From the optimization conditions
∂ J (k)/∂ x0 = 0T , ∂ J (k)/∂w(i) = 0T , ∂ J (k)/∂λ(i) = 0T ones obtains an expres-
sion of J (k) as function of x̂ − (k+1) and y(k). Next, from the optimization conditions
∂ J (k)/∂ x̂ − (i + 1) = 0T , and ∂ J (k)/∂ y(i) = 0T one obtains the filter’s equations.
The recursion of the H∞ Kalman Filter can be formulated again in terms of a
measurement update and a time update part
measurement update:

D(k) = [I − θ W (k)P − (k) + C T (k)R(k)−1 C(k)P − (k)]−1

K (k) = P − (k)D(k)C T (k)R(k)−1 (6.122)
x̂(k) = x̂ − (k) + K (k)[y(k) − C x̂ − (k)]
300 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

time update:
x̂ − (k + 1) = A(k)x(k) + B(k)u(k)
(6.123)
P − (k + 1) = A(k)P − (k)D(k)A T (k) + Q(k)

where it is assumed that parameter θ is sufficiently small to maintain

P − (k) − θ W (k) + C T (k)R(k)−1 C(k) (6.124)

positive definite. When θ = 0 the H∞ Kalman Filter becomes equivalent to the

standard Kalman Filter. It is noted that apart from the process noise covariance
matrix Q(k) and the measurement noise covariance matrix R(k) the H∞ Kalman
filter requires tuning of the weight matrices L and S, as well as of parameter θ .

6.6.5 Robust State Estimation with the Use

of Disturbance Observers

6.6.5.1 State Estimation with the Derivative-Free

Nonlinear Kalman Filter

Previous results about state estimation through transformation to linear canonical

forms can be found in [334, 335, 436, 437]. It was shown that the dynamical model
of the suspension can be written in the MIMO canonical form of Eq. (6.110). Thus
one has a MIMO linear model of the form

ẏ f = A f y f + B f v
(6.125)
z f = Cf yf

where y f = [F1 , F2 , F3 , F4 ]T and matrices A f ,B f ,C f are in the form

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 1
⎜0 0 1 0⎟ ⎜0⎟ T ⎜0⎟
Af = ⎜
⎝0
⎟ B =⎜ ⎟C =⎜ ⎟ (6.126)
0 0 1⎠ f ⎝0⎠ f ⎝0⎠
0 0 0 0 1 0

where the measurable variables y1 = F is associated with the displacement of the

sprung and unsprung mass in the suspension model. For the aforementioned model,
and after carrying out discretization of matrices A f , B f and C f with common dis-
cretization methods one can apply linear Kalman filtering using Eqs. (4.5) and (4.6).
This is Derivative-free nonlinear Kalman filtering for the model of the suspension
which is performed without the need to compute Jacobian matrices and does not
introduce numerical errors due to approximative linearization with Taylor series
expansion.
6.6 Active Vehicle Suspension Control 301

6.6.5.2 Kalman Filter-Based Estimation of Suspension

Disturbance Forces

Considering the effects of disturbances on the suspension’s model and after applying
a transform on the system’s state variables according to the differential flatness theory
it has been shown that the suspension model is described by

F (4) = u + φ(τ ) (6.127)

The suspension’s state-space model of Eq. (6.110) will be extended to take into
account also the dynamics and the effects of the disturbance input φ(t). The extended
state vector of the suspension model is defined as z∈R 8×1 with z 1 = F, z 2 = Ḟ, z 3 =
F̈, z 4 = F (3) , z 5 = φ, z 6 = φ̇, z 7 = φ̈, z 8 = φ (3) . The dynamics of the disturbance is
assumed to be defined by its fourth-order derivative, i.e., φ (4) = f d (F, Ḟ, F̈, F (3) ).
Thus one has the extended state-space model

ż = ÷z + B̃·ṽ
(6.128)
q = C̃ z

with ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 0 00 1
⎜0 0 1 0 0 0 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 1 0 0 0⎟ ⎜1 0 ⎟ T ⎜ ⎟
à = ⎜ ⎟ , B̃ = ⎜ ⎟ , C̃ = ⎜0⎟ (6.129)
⎜0 0 0 0 0 1 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 1 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 1⎠ ⎝0 0⎠ ⎝0 ⎠
0 0 0 0 0 0 0 0 01 0

where the measurable variable is z 1 and the control input is

 T
ṽ = u, φ (4) (6.130)

The disturbance estimator has the following structure

ẑ˙ = Ão ·ẑ + B̃o ·ṽ + K (z 1 − ẑ 1 )

(6.131)
q̂ = C̃o ŷ

where the estimator’s gain K ∈R 8×1 and matrices Ão , B̃o and C̃o are defined as
302 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 0 00 1
⎜0 0 1 0 0 0 0 0⎟ ⎜0 0 ⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0 ⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 1 0 0 0⎟⎟ ⎜ ⎟
1 0⎟ T ⎜0 ⎟
Ão = ⎜
⎜0 ⎟ , B̃o = ⎜
⎜ ⎟ , C̃ o = ⎜ ⎟
⎜0 ⎟ (6.132)
⎜ 0 0 0 0 1 0 0⎟ ⎜0 0 ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 1 0⎟⎟ ⎜ ⎟ ⎜0 ⎟
⎜ ⎜0 0 ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 1 ⎠ ⎝ 00 ⎠ ⎝0 ⎠
0 0 0 0 0 0 0 0 00 0

The disturbance estimator’s gain K ∈R 8×1 will be computed through the Kalman
Filter recursion.
Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents of matrices Ão , B̃o and
C̃o respectively, a Derivative-free nonlinear Kalman Filter can be designed for the
aforementioned representation of the system dynamics [436, 437]. The associated
Kalman Filter-based disturbance estimator is given by
measurement update:

K (k) = P − (k)C̃dT [C̃d ·P − (k)C̃dT + R]−1

x̂(k) = x̂ − (k) + K (k)[z(k) − C̃d x̂ − (k)] (6.133)
P(k) = P − (k) − K (k)C̃d P − (k)

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

(6.134)
x̂ − (k + 1) = Ãd (k)x̂(k) + B̃d (k)ṽ(k)

To compensate for the effects of the disturbance forces it suffices to use in the control
loop the modified control input vector v1 = u − φ(t).

6.6.6 Simulation Tests

To evaluate for the performance of the proposed Kalman Filter-based and distur-
bances estimation scheme for the vehicle’s suspension model simulation tests were
carried out. Different disturbance forces were assumed to be exerted on the vehicle’s
wheel due to its contact with the rough road surface. The disturbances dynamics was
completely unknown to the controller and their identification was performed in real
time by the disturbance estimator. The parameters of the suspension mechanism were
as follows: m 1 = 2000 Kg, m 2 = 40 Kg, K 1 = 1.0 · 10e4 N/m, K 2 = 1.0 · 10e4 N/m,
c1 = 1790 N · s/m and c2 = 20 N · s/m. The control loop used for the vehicle’s sus-
pension is given in Fig. 6.53.
6.6 Active Vehicle Suspension Control 303

Fig. 6.53 Control loop for the vehicle’s suspension comprising a flatness-based nonlinear controller
and a Kalman Filter-based disturbances estimator

The monitored parameters were the state variables of the suspension. The control
input was the force generated by the actuator. The measured parameters were the
position and velocity of the sprung and unsprung mass. The dynamics of the distur-
bance force was assumed to be defined by its fourth-order derivative. The extended
state vector used by the disturbance observer was of dimension x∈R 8 , where the
first four state variables were describing the suspension’s model whereas the rest
four state variables were associated with the dynamics of the disturbance force. The
real-time estimation of the external disturbance that was provided by the Kalman Fil-
ter was used by an additional control term in the control loop to generate a counter
disturbance control input. Thus, the disturbance’s effects on the vehicle’s parts were
eliminated and vibrations were efficiently suppressed. As shown in Figs. 6.54, 6.55,
6.56, 6.57, 6.58, 6.59, 6.60, 6.61, 6.62, 6.63, 6.64 and 6.65 fast stabilization of the
suspension’s sprung and unsprung masses to the desirable setpoints was succeeded
and accurate estimation of the unknown disturbances forces was performed. More-
 that for common nominal values of k1 and
over, is should be taken into account
m 1 one obtains t < τ , i.e., t = mk11 τ which finally gives the real time scale of
suspension’s active control system.
304 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

−3
x 10
(a) 4 (b) 0.01
3.5

3 0.005

2.5
0
2
x1

x2
1.5
−0.005
1

0.5 −0.01
0

−0.5 −0.015
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.54 Suspension control under disturbances profile 1: a convergence of sprung mass position
x1 to the desirable setpoint, b convergence of sprung mass velocity x2 to the desirable setpoint

(a) 0.08 (b) 1

0.8
0.06
0.6
0.04
0.4

0.02 0.2
x3

x4

0
0
−0.2
−0.02
−0.4

−0.04 −0.6
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.55 Suspension control under disturbances profile 1: a convergence of unsprung mass position
x2 to the desirable setpoint, b convergence of unsprung mass velocity x4 to the desirable setpoint

6.7 State Estimation-Based Control of Quadrotors

6.7.1 Overview

Another type of autonomous vehicles in which one can apply the differential flat-
ness theory-based methods of filtering and control is Unmanned Aerial Vehicles
(UAVs). Quadrotors are four-rotor helicopters characterized by a nonlinear 6-DOF
unstable dynamical model. To succeed autonomous navigation of the quadrotors,
6.7 State Estimation-Based Control of Quadrotors 305

(a) 0.12 (b) 0.4

0.1
0.2
0.08
0
0.06
φ −est

0.04 −0.2

v
0.02 −0.4
0
−0.6
−0.02
−0.8
−0.04

−0.06 −1
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.56 Suspension control under disturbances profile 1: a estimation of the disturbance terms,
b control input generated by the actuator

0.1
(a) (b) 0.15
0.08 0.1

0.05
0.06
0
x1

x2

0.04
−0.05
0.02
−0.1

0 −0.15

−0.02 −0.2
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.57 Suspension control under disturbances profile 2: a convergence of sprung mass position
x1 to the desirable setpoint, b convergence of sprung mass velocity x2 to the desirable setpoint

it is necessary to design efficient control algorithms that will exhibit robustness to

parametric uncertainties and to external disturbances. One can cite several results
on quadrotors’ control. An approach for quadrotors’ control that is based on the
transformation of their dynamical model in the linear canonical form and which is
associated with differential flatness theory has been given in [541]. Moreover, in
[3] a flatness-based control approach is applied to quadrotors’ motion control. A
predictive controller complemented by an H∞ term for additional robustness has
been analyzed and tested in the quadrotor’s flight control problem in [397, 398]. In
[52] motion control of the quadrotor was implemented using controllers of the LQR-
type and of the PID-type, while Kalman Filtering has been used to provide position
306 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 0.14 (b)1.2

0.12 1

0.1 0.8

0.08 0.6

0.06 0.4

x4
x3

0.04 0.2

0.02 0

0 −0.2

−0.02 −0.4

−0.04 −0.6
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.58 Suspension control under disturbances profile 2: a convergence of unsprung mass position
x2 to the desirable setpoint, b convergence of unsprung mass velocity x4 to the desirable setpoint

−3
x 10
(a) 6 (b) 0.05

4 0

−0.05
2
φ −est

−0.1
0
u

−0.15
−2
−0.2

−4 −0.25

−6 −0.3
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.59 Suspension control under disturbances profile 2: a estimation of the disturbance terms,
b control input generated by the actuator

estimates out of a visual measurements system. In [77] two control strategies are
employed as baseline controllers for the quadrotor’s model: a LQR controller which
is based on a linearized model of the quadrotor and a Sliding-Mode Controller which
is based on a nonlinear model of the quadrotor. Moreover, differential flatness the-
ory has been used for trajectory planning. In [271] and in [136], adaptive control
schemes have been proposed for the quadrotor’s model. The stability of the control
loop is confirmed through the Lyapunov approach. In [43] quadrotor’s control with
the use of a sliding-mode controller and a sliding-mode disturbance observer has
been proposed.
6.7 State Estimation-Based Control of Quadrotors 307

(a) 0.02 (b) 0.03

0.015 0.025

0.01 0.02
0.015
0.005
0.01
0
x1

x2
0.005
−0.005
0
−0.01
−0.005
−0.015 −0.01
−0.02 −0.015
−0.025 −0.02
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.60 Suspension control under disturbances profile 3: a convergence of sprung mass position
x1 to the desirable setpoint, b convergence of sprung mass velocity x2 to the desirable setpoint

(a) 0.08 (b) 1

0.8
0.06
0.6
0.04
0.4
x3

x4

0.02 0.2

0
0
−0.2
−0.02
−0.4

−0.04 −0.6
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.61 Suspension control under disturbances profile 3: a convergence of unsprung mass position
x2 to the desirable setpoint, b convergence of unsprung mass velocity x4 to the desirable setpoint

In this section, a new flatness-based control method is developed for quadrotors

that is based on differential flatness theory together with the use of a disturbance
observer that is also in accordance to differential flatness theory (the previously
analyzed Derivative-free nonlinear Kalman Filter). The differential flatness theory-
based design of the controller uses a change of coordinates (diffeomorhism) that
transforms the state-space equation of the quadrotor’s model into the linear canon-
ical (Brunovsky) form [55, 152, 340, 427, 465, 535]. For the linearized equivalent
of the quadrotor it is easier to design a state feedback controller using techniques
for linear feedback controllers’ synthesis. To supply the quadrotor’s control loop
with additional robustness, a disturbance observer is used. The disturbance observer
308 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 0.03 (b) 0.4

0.02
0.2
0.01
0 0

−0.01
φ −est

−0.2
−0.02

v
−0.4
−0.03
−0.04 −0.6
−0.05
−0.8
−0.06
−0.07 −1
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.62 Suspension control under disturbances profile 3: a estimation of the disturbance terms,
b control input generated by the actuator

(a) 0.16 (b) 0.25

0.14 0.2

0.12 0.15

0.1 0.1

0.08 0.05
x1

x2

0.06 0

0.04 −0.05

0.02 −0.1

0 −0.15

−0.02 −0.2
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.63 Suspension control under disturbances profile 4: a convergence of sprung mass position
x1 to the desirable setpoint, b convergence of sprung mass velocity x2 to the desirable setpoint

makes use of the standard Kalman Filter recursion on the linearized model of the
quadrotor. It is capable of estimating simultaneously the quadrotor’s linear and rota-
tional velocities, as well as the vector of disturbances that affect the quadrotor’s
model without the need to compute Jacobian matrices. The accurate estimation of
the disturbance inputs enables to introduce an additional control term that compen-
sates for the disturbances’ effects. The accurate tracking of reference trajectories that
is performed by the quadrotor despite the existence of external disturbances is shown
in simulation experiments.
Differential flatness theory has specific advantages when used in nonlinear con-
trol systems [55, 152, 340, 427, 465, 535]. By enabling an exact linearization of the
6.7 State Estimation-Based Control of Quadrotors 309

(a) 0.25 (b) 2

0.2 1.5

0.15
1
0.1
x3

0.5

x4
0.05
0
0

−0.05 −0.5

−0.1 −1
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.64 Suspension control under disturbances profile 4: a convergence of unsprung mass position
x2 to the desirable setpoint, b convergence of unsprung mass velocity x4 to the desirable setpoint

(a) 0.03 (b) 0.2

0.025 0.1

0.02
0
0.015
φ −est

−0.1
0.01
v

−0.2
0.005
−0.3
0

−0.005 −0.4

−0.01 −0.5
0 1 2 3 4 5 0 1 2 3 4 5
time time

Fig. 6.65 Suspension control under disturbances profile 4: a estimation of the disturbance terms,
b control input generated by the actuator

system’s dynamical model it makes possible to avoid the use of linear models of
local validity in the controller’s design. The controller performs efficiently despite
the change of operating points. After the design of such a state feedback controller,
one can consider the inclusion in the control loop of supplementary control terms
that will provide additional robustness. Thus one can design flatness-based adaptive
fuzzy controllers or flatness-based sliding-mode controllers. As mentioned above,
it is also possible to use a disturbance estimator-based auxiliary control input for
compensating for the effects of disturbances in the feedback control loop. Moreover,
the use of differential flatness theory in the design of state estimators and filters has
also several strong points. One can perform estimation of the complete state vector
310 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

of the system without the need to compute partial derivatives and Jacobian matrices.
Additionally, by avoiding numerical errors which are due to approximate lineariza-
tion of the system’s dynamic model (e.g., with the use of expansion in Taylor series),
linear estimation algorithms can be implemented. In the case of Kalman Filtering,
this means that one can perform state estimation with the use of the standard Kalman
Filter recursion, thus preserving the method’s optimality features and providing state
estimates of improved precision (e.g., comparing to Extended Kalman Filtering).

6.7.2 Kinematic Model of the Quadropter

As shown in Figs. 6.66 and 6.67, two reference frames are defined [397, 398]. The
first one B = [B1 , B2 , B3 ] is attached to the quadropter’s body, whereas the second
E = [E x , E y , E z ] is considered to be an inertial coordinates system. The Euler
angles defining rotation round the axes of the inertial reference frame E 1 , E 2 , E 3 are
defined as θ , φ, and ψ, respectively. The two reference frames are connected to each
other through a rotation matrix
⎛ ⎞
CψCθ Cψ Sθ Sφ − SψCφ Cψ SθCφ + Sψ Sφ
R = ⎝ SψCθ Sψ Sθ Sφ + CψCφ Sψ SθCφ − Cψ Sφ ⎠ (6.135)
−Sθ Cθ Sφ CθCφ

where C = cos(·) and S = sin(·).

The connection between velocities in the two reference frames is as follows:

VE = R·VB (6.136)

Fig. 6.66 Quadrotor

performing a surveillance
task and the associated
reference frames
6.7 State Estimation-Based Control of Quadrotors 311

Fig. 6.67 Reference axes

for the quadropter

where VE = [u E , v E , w E ] and VB = [u B , v B , w B ] are the linear velocity vectors

expressed in the two reference frames. About the angular velocities in the two refer-
ence frames the following relation holds

η̇ = W −1 ω (6.137)

that is ⎛ ⎞ ⎛ ⎞⎛ ⎞
φ̇ 1 sin(φ)tan(θ ) cos(φ)tan(θ ) p
⎝ θ̇ ⎠ = ⎝0 cos(φ) −sin(φ) ⎠ ⎝ q ⎠ (6.138)
ψ̇ 0 sin(φ)sec(θ ) cos(φ)sec(θ ) r

where η = [φ, θ, ψ]T is the angular velocities vector in the inertial reference frame
and ω = [ p, q, r ]T is the angular velocities vector in the body-fixed reference frame.

6.7.3 Euler-Lagrange Equations for the Quadropter

The Euler-Lagrange equation for the quadropter is formulated as follows:

 
d ∂L ∂L fξ
dt ( ∂ q̇i )− ∂qi =
τη
(6.139)

where the Lagrangian is defined as L(q, q̇) = E Ctrans + E Cr ot − E p , where E Ctrans

is the kinetic energy of the quadrotor due to translational motion, E Cr ot is the kinetic
energy of the quadrotor due to rotational motion, and E p is the total potential energy
of the quadrotor due to lift. The generalized state vector is q = [ξ T , η T ]T ∈R 6 , τη ∈R 3
312 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

is the torques vector that causes rotation round the axes of the body-fixed reference
frame, and f ξ = R fˆ + αT is the translational force applied to the quadrotor due
to the main control input U1 along the z-axis direction, while αT = [A x , A y , A z ]T
are the aerodynamic forces vector, defined along the axes of the inertial reference
frame. Since the Lagrangian does not contain cross-coupling between the ξ̇ and the η̇
terms, the Lagrange-Euler equations can be divided into translational and rotational
dynamics.
The translational dynamics of the quadropter is given by

m ξ̈ + mge3 = f ξ (6.140)

where e3 = [0, 0, 1]T is the unit vector along the z axis of the inertial reference
frame. Equation (6.140) can be written using the following three equations

Ax
ẍ = m (cos(ψ)sin(θ )cos(φ) + sin(ψ)sin(φ))U1
1
+ m
Ay
ÿ = m (sin(ψ)sin(θ )cos(φ) − cos(ψ)sin(φ))U1
1
+ m
(6.141)
z̈ = −g + m1 (cos(θ )cos(φ))U1 + Amz

where m is the quadropter’s mass and g is the gravitational acceleration.

The rotational dynamics of the quadropter is given by

M(η)η̈ + C(η, η̇)η̇ = τη (6.142)

where the inertia matrix M(η) is defined as

⎛ ⎞
Ix x 0 −I x x Sθ
M(η) = ⎝ 0 I yy C 2 φ + Izz S 2 φ (I yy − Izz )Cφ SφCθ ⎠
−I x x Sθ (I yy − Izz )Cφ SφCθ I x x S θ + I yy S φC θ + Izz C φC θ
2 2 2 2 2

(6.143)

and the Coriolis matrix is

⎛ ⎞
c11 c12 c13
C(η, η̇) = ⎝c21 c22 c23 ⎠ (6.144)
c31 c32 c33

where the elements of the matrix are

c11 = 0
c12 = (I yy − Izz )(θ̇Cφ Sφ + ψ̇ S 2 φCθ ) + (Izz − I yy )ψ̇C 2 φCθ
c13 = (Izz − I yy )ψ̇Cφ SφC 2 θ
c21 = (Izz − I yy )(θ̇Cφ Sφ + ψ̇ S 2 φCθ ) + (I yy − Izz )ψ̇C 2 φCθ + I x x ψ̇Cθ
c22 = (Izz − I yy )φ̇Cφ Sφ
6.7 State Estimation-Based Control of Quadrotors 313

c23 = −I x x ψ̇ SθCθ + I yy ψ̇ S 2 φCθ Sθ + Izz ψ̇C 2 φ SθCθ

c31 = (I yy − Izz )ψ̇C 2 θ SφCφ − I x x θ̇Cθ
c32 = (Izz − I yy )(θ̇Cφ Sφ Sθ + φ̇ S 2 φCθ ) + (I yy − Izz )φ̇C 2 φCθ + I x x ψ̇ SθCθ −
−I yy ψ̇ S 2 φ SθCθ − Izz ψ̇C 2 φ SθCθ
c33 = (I yy − Izz )φ̇Cφ SφC 2 θ − I yy θ̇ S 2 φCθ Sθ −
−Izz θ̇C 2 φCθ Sθ + I x x θ̇Cθ Sθ (6.145)

Thus, the mathematical model that describes the quadrotor’s rotational motion is
given by

η̈ = M(η)−1 (τη − C(η, η̇)η̇) (6.146)

Denoting w = M(η)−1 (τη − C(η, η̇)η̇), one has the following notation for the
rotational dynamics
⎛ ⎞ ⎛ ⎞
φ̈ wa
⎝ θ̈ ⎠ = ⎝wb ⎠ (6.147)
ψ̈ wc

Considering small variations of the heading angle of the quadrotor round ψ = π2 ,

denoting w1 = U1 /m and taking also that the aerodynamic coefficients A x , A y , A z

m, a simplified quadropter is formulated as follows [541]

ẍ = w1 sin(φ) ÿ = w1 cos(φ)sin(θ ) z̈ = w1 cos(φ)cos(θ ) − g

(6.148)
φ̈ = wa θ̈ = wb ψ̈ = wc

6.7.4 Design of Flatness-Based Controller

for the Quadrotor’s Model

It will be shown, that the quadrotor’s model given in Eq. (6.148) is a differentially flat
one, i.e., that all its state variables and the associated control inputs can be written
as functions of a new variable called flat output and of its derivatives.
The following state variables are introduced x1 = x, x2 = ẋ, x3 = y, x4 = ẏ,
x5 = z, x6 = ż, x7 = φ, x8 = φ̇, x9 = θ , x10 = θ̇ , x11 = ψ, x12 = ψ̇. Thus, one
has the following state-space description for the quadrotor’s dynamic model

ẋ1 = x2 ẋ2 = w1 sin(x7 ) ẋ3 = x4 ẋ4 = w1 cos(x7 )sin(x9 )

ẋ5 = x6 ẋ6 = w1 cos(x7 )cos(x9 ) ẋ7 = x8 ẋ8 = wa (6.149)
ẋ9 = x10 ẋ10 = wb ẋ11 = x12 ẋ12 = wc

The flat output of the system is taken to be the vector y f = [x1 , x3 , x5 , x7 , x9 , x11 ]T .
It holds that
314 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

x1 = [1 0 0 0 0 0]y f x2 = [1 0 0 0 0 0] ẏ f x3 = [0 1 0 0 0 0]y f x4 = [0 1 0 0 0 0] ẏ f
x5 = [0 0 1 0 0 0]y f x6 = [0 0 1 0 0 0] ẏ f x7 = [0 0 0 1 0 0]y f x8 = [0 0 0 1 0 0] ẏ f
x9 = [0 0 0 0 1 0]y f x10 = [0 0 0 0 1 0] ẏ f x11 = [0 0 0 0 0 1]y f x12 = [0 0 0 0 0 1] ẏ f
(6.150)

According to Eq. (6.150) all state variables of the quadropter can be written as func-
tions of the flat output and its derivatives. Using this and Eq. (6.149), one also has
that the control inputs of the quadropter’s model, w1 , wa , wb , and wc can be written
as functions of the flat output and its derivatives. Therefore, it is confirmed that the
system is a differentially flat one.

Defining now the new control inputs

v1 = w1 sin(x7 ) v2 = w1 cos(x7 )sin(x9 ) v3 = w1 cos(x7 )cos(x9 )

(6.151)
v4 = wa v5 = wb v6 = wc

one has the following state-space description for the system

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ f1 010000000000 y f1 000000
⎜ ÿ f ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟ ⎜1 0 0 0 0 0⎟
⎜ 1⎟ ⎜ ⎟ ⎜ 1⎟ ⎜ ⎟
⎜ ẏ f ⎟ ⎜0 0 0 1 0 0 0 0 0 0 0 0⎟ ⎜ y f ⎟ ⎜0 0 0 0 0 0⎟
⎜ 2⎟ ⎜ ⎟ ⎜ 2⎟ ⎜ ⎟⎛ ⎞
⎜ ÿ f ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟ ⎜0 1 0 0 0 0⎟ v1
⎜ 2⎟ ⎜ ⎟ ⎜ 2⎟ ⎜ ⎟
⎜ ẏ f ⎟ ⎜0 0 0 0 0 1 0 0 0 0 0 0⎟ ⎜ y f ⎟ ⎜0 0 0 0 0 0⎟ ⎜v2 ⎟
⎜ 3⎟ ⎜ ⎟ ⎜ 3⎟ ⎜ ⎟⎜ ⎟
⎜ ÿ f ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟ ⎜0 0 1 0 0 0⎟ ⎜v3 ⎟
⎜ 3⎟ = ⎜ ⎟ ⎜ 3⎟ + ⎜ ⎟⎜ ⎟ (6.152)
⎜ ẏ f ⎟ ⎜0 0 0 0 0 0 0 1 0 0 0 0⎟ ⎜ y f ⎟ ⎜0 0 0 0 0 0⎟ ⎜v4 ⎟
⎜ 4⎟ ⎜ ⎟ ⎜ 4⎟ ⎜ ⎟⎜ ⎟
⎜ ÿ f ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟ ⎜0 0 0 1 0 0⎟ ⎝v5 ⎠
⎜ 4⎟ ⎜ ⎟ ⎜ 4⎟ ⎜ ⎟
⎜ ẏ f ⎟ ⎜0 0 0 0 0 0 0 0 0 1 0 0⎟ ⎜ y f ⎟ ⎜0 0 0 0 0 0⎟ v6
⎜ 5⎟ ⎜ ⎟ ⎜ 5⎟ ⎜ ⎟
⎜ ÿ f ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟ ⎜0 0 0 0 1 0⎟
⎜ 5⎟ ⎜ ⎟ ⎜ 5⎟ ⎜ ⎟
⎝ ẏ f6 ⎠ ⎝0 0 0 0 0 0 0 0 0 0 0 1⎠ ⎝ y f6 ⎠ ⎝0 0 0 0 0 0⎠
ÿ f6 000000000000 ẏ f6 000001

and the measurement equation for this system becomes

⎛ ⎞
y f1
⎜ ẏ f ⎟
⎜ 1⎟
⎜ ⎟
⎛ ⎞ ⎛ ⎞ ⎜ y f2 ⎟
z1 100000000000 ⎜ ⎟
⎜ ẏ f2 ⎟
⎜ z 2 ⎟ ⎜0 0 1 0 0 0 0 0 0 0 0 0 ⎟ ⎜ y f ⎟
⎜ ⎟ ⎜ ⎟ ⎜ 3⎟
⎜z 3 ⎟ ⎜0 0 0 0 1 0 0 0 0 0 0 0⎟ ⎜ ẏ f ⎟
⎜ ⎟=⎜ ⎟ ⎜ 3⎟ (6.153)
⎜ z 4 ⎟ ⎜0 0 0 0 0 0 1 0 0 0 0 0 ⎟ ⎜ y f ⎟
⎜ ⎟ ⎜ ⎟ ⎜ 4⎟
⎝z 5 ⎠ ⎝0 0 0 0 0 0 0 0 1 0 0 0⎠ ⎜ ẏ f4 ⎟
⎜ ⎟
z6 000000000010 ⎜ ⎜ y f5 ⎟
⎟
⎜ ẏ f ⎟
⎜ 5⎟
⎝ y f6 ⎠
ẏ f6
6.7 State Estimation-Based Control of Quadrotors 315

Thus, using differential flatness theory, the quadrotor’s model has been written in a
MIMO linear canonical (Brunovsky) form, which is both controllable and observable.
After being written in the linear canonical form, the quadrotor’s state-space equa-
tion comprises six subsystems of the form

ÿ fi = vi , i = 1, . . . , 6 (6.154)

For each one of these subsystems, a controller can be defined as follows:

vi = ÿ dfi − kdi ( ẏ fi − ẏ dfi ) − k pi (y fi − y dfi ), i = 1, . . . , 6 (6.155)

The control scheme is implemented in the form of two cascading loops. The inner
control loop controls rotation angles, while the outer control loop sets the desired
values of the rotation angles, so as to control position in the xyz-reference system.
The computation of the reference setpoints for the rotation angles φd (t), θd (t), and
ψd (t) and for the cartesian coordinates xd (t),yd (t), and z d (t) takes into account the
constraints imposed by the system dynamics.

6.7.5 Estimation of the Quadrotor’s Disturbance Forces and

Torques with Kalman Filtering

It was shown that the initial nonlinear model of the quadrotor can be written in
the MIMO canonical form of Eqs. (6.152) and (6.153). Next, it is assumed that the
quadrotor’s model is affected by additive input disturbances, thus one has

ẍ1 = (w1 + d1 )sin(x7 )

ẍ3 = (w1 + d1 )cos(x7 )sin(x9 )
ẍ5 = (w1 + d1 )cos(x7 )cos(x9 )
(6.156)
ẍ7 = wa + da
ẍ9 = wb + db
ẍ11 = wc + dc

or using the new state variables y fi i = 1, . . . , 12 of the differential flatness theory-

based model and the transformed inputs vi , i = 1, . . . , 6 one has

ÿ f1 = v1 + d1 sin(y f7 )
ÿ f3 = v2 + d1 cos(y f7 )sin(y f9 )
ÿ f5 = v3 + d1 cos(y f7 )cos(y f9 )
(6.157)
ÿ f7 = v4 + da
ÿ f9 = v5 + db
ÿ f11 = v6 + dc
316 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

while by redefining the disturbance terms as d̃1 = d1 sin(y f7 ), d̃2 = d1 cos(y f7 )

sin(y f9 ), d̃3 = d1 cos(y f7 )cos(y f9 ), d̃4 = da , d̃5 = db and d̃6 = dc , the dynamics of
the disturbed system can be written as

ÿ f1 = v1 + d̃1
ÿ f3 = v2 + d̃2
ÿ f5 = v3 + d̃3
(6.158)
ÿ f7 = v4 + d̃4
ÿ f9 = v5 + d̃5
ÿ f11 = v6 + d̃6

The system’s dynamics can be also written as ẏ f1 = y f2 , ẏ f2 = v1 + d̃1 , ẏ f3 = y f4 ,

ẏ f4 = v2 + d̃2 , ẏ f5 = y f6 , ẏ f6 = v3 + d̃3 , ẏ f7 = y f8 , ẏ f8 = v4 + d̃4 , ẏ f9 = y f10 ,
ẏ f10 = v5 + d̃5 , ẏ f11 = y f6 , ẏ f6 = v6 + d̃6 .
Without loss of generality, it is assumed that the dynamics of the disturbances
terms are described by their second-order derivative, i.e., d̃¨i = f di , i = 1, . . . , 6.
Next, the extended state vector of the system is defined so as to include disturbance
terms as well. Thus one has the following state variables

z f1 = y f1 z f2 = y f2 z f3 = y f3 z f4 = y f4 z f5 = y f5 z f6 = y f6
z f7 = y f7 z f8 = y f8 z f9 = y f9 z f10 = y f10 z f11 = y f11 z f12 = y f12
z f13 = d̃1 z f14 = d̃˙1 z f15 = d̃¨1 z f16 = d̃2 z f17 = d̃˙2 z f18 = d̃¨2 (6.159)
z f19 = d̃3 z f20 = d̃˙3 z f21 = d̃¨3 z f22 = d̃4 z f23 = d̃˙4 z f24 = d̃¨4
z = d̃ z = d̃˙ z = d̃¨
f 25 5 f 26 6 f 27 5 z = d̃ z = d̃˙ z = d̃¨
f 28 6 f 29 6 f 30 6

Thus, the disturbed system can be described by a state-space equation of the form

ż f = A f z f + B f v
(6.160)
z meas
f = Cfzf

where matrices A f ∈R 30×30 , B f ∈R 30×6 , and C f ∈R 6×30 , are described in Eq. (6.162).
For the model of Eqs. (6.160) and (6.162), and after carrying out discretization of
matrices A f , B f , and C f with common discretization methods, one can implement
the standard Kalman Filter algorithm using Eqs. (4.5) and (4.6). This is Derivative-
free nonlinear Kalman filtering for the model of the quadropter which, unlike EKF,
is performed without the need to compute Jacobian matrices and does not introduce
numerical errors due to approximative linearization with Taylor series expansion.
The dynamics of the disturbance terms d̃i , i = 1, . . . , 6 are taken to be unknown
in the design of the associated disturbances’ estimator. Defining as Ãd , B̃d , and C̃d ,
the discrete-time equivalents of matrices à f , B̃ f , and C̃ f , respectively, one has the
following dynamics:

ẑ˙ f = Ã f ·ẑ f + B̃ f ·ṽ + K (z meas

f − C̃ f ẑ f ) (6.161)
6.7 State Estimation-Based Control of Quadrotors 317

where K ∈R 30×6 is the state estimator’s gain. The associated Kalman Filter-based
disturbance estimator is given by [422, 453]
⎛ ⎞
01×1 1 01×28
⎜01×12 1 01×17 ⎟
⎜ ⎟
⎜ 01×3 1 01×26 ⎟
⎜ ⎟
⎜01×15 1 01×14 ⎟
⎜ ⎟
⎜ 01×5 1 01×24 ⎟
⎜ ⎟
⎜01×18 1 01×11 ⎟
⎜ ⎟
⎜ 01×7 1 01×22 ⎟
⎜ ⎟
⎜01×21 1 01×8 ⎟
⎜ ⎟
⎜ 01×9 1 01×20 ⎟
⎜ ⎟ ⎛ ⎞
⎜01×24 1 01×5 ⎟ 01×6
⎜ ⎟
⎜01×11 1 01×18 ⎟ ⎜ 1 01×5 ⎟
⎜ ⎟ ⎜ ⎟
⎜01×27 1 01×2 ⎟ ⎜ 01×1 1 01×4 ⎟
⎜ ⎟ ⎜ ⎟ ⎛ ⎞
⎜01×13 1 01×16 ⎟ ⎜ 01×6 ⎟ 1 01×29
⎜ ⎟ ⎜ ⎟
⎜01×14 1 01×15 ⎟ ⎜ 01×2 1 01×3 ⎟ ⎜ 01×2 1 01×27 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜01×30 ⎟ ⎜ ⎟ ⎜ 01×25 ⎟
Af = ⎜ ⎟ B f = ⎜ 01×6 ⎟ C f = ⎜ 01×4 1 ⎟
⎜01×16 1 01×13 ⎟ ⎜ 01×3 1 01×2 ⎟ ⎜ 01×6 1 01×23 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜01×17 1 ⎟
01×12 ⎟ ⎜ 01×6 ⎟ ⎝ 01×8 1 01×21 ⎠
⎜ ⎜ ⎟
⎜01×30 ⎟ ⎜ 01×4 1 ⎟
01×1 ⎟ 01×10 1 01×19
⎜ ⎟ ⎜
⎜01×19 1 ⎟
01×10 ⎟ ⎜ 01×6 ⎟
⎜ ⎜ ⎟
⎜01×20 1 01×9 ⎟⎟ ⎝ 01×5 1 ⎠
⎜
⎜01×30 ⎟ 018×6
⎜ ⎟
⎜01×22 1 01×7 ⎟
⎜ ⎟
⎜01×23 1 01×6 ⎟
⎜ ⎟
⎜01×30 ⎟
⎜ ⎟
⎜01×25 1 01×4 ⎟
⎜ ⎟
⎜01×26 1 01×3 ⎟
⎜ ⎟
⎜01×30 ⎟
⎜ ⎟
⎜01×28 1 01×1 ⎟
⎜ ⎟
⎝01×29 1 ⎠
01×30
(6.162)
measurement update:

K (k) = P − (k)C̃dT [C̃d ·P − (k)C̃dT + R]−1

ẑ f (k) = ẑ −f (k) + K (k)[z meas
f (k) − C̃d ẑ −f (k)] (6.163)
−
P(k) = P (k) − K (k)C̃d P (k) −

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

(6.164)
ẑ −f (k + 1) = Ãd (k)ẑ f (k) + B̃d (k)ṽ(k)
318 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

To compensate for the effects of the disturbance forces it suffices to use in the control
loop the modified control input vector
⎛ ⎞
v1 − d̃ˆ1 ⎛ ⎞
⎜ ˆ ⎟ v1 − ẑ 13
⎜v2 − d̃2 ⎟ ⎜v2 − ẑ 16 ⎟
⎜ ⎟ ⎜ ⎟
⎜ ˆ ⎟ ⎜v3 − ẑ 19 ⎟
⎜v − d̃ ⎟
v = ⎜ 3 ˆ3 ⎟ or v = ⎜
⎜v4 − ẑ 22 ⎟
⎟ (6.165)
⎜v4 − d̃4 ⎟ ⎜ ⎟
⎜ ⎟ ⎝v5 − ẑ 25 ⎠
⎜ ˆ ⎟
⎝v5 − d̃5 ⎠ v6 − ẑ 28
v6 − d̃ˆ6

6.7.6 Simulation Tests

Initial simulation experiments were concerned with flight control of the quadropter
in the disturbance-free case. The considered reference trajectories are shown in
Fig. 6.68. The implementation of the flatness-based control enabled accurate tracking
of the reference trajectories. Convergence has been succeeded for the linear position
and velocity variables to the associated setpoints as it can be seen in Fig. 6.69a, b
and in Fig. 6.70a. Moreover, there has been convergence of the angular position and
velocity variables to the associated setpoints as it can be seen in Fig. 6.70a and in
Fig. 6.71a, b.
Additional simulation experiments were concerned with control of the quadropter
in flight under disturbance forces and torques. The estimation of the disturbance
forces and torques is shown in Fig. 6.72. The implementation of the flatness-based

(b) 0.8
(a) 0.6
0.8
0.4
0.6
0.2
0.4
z

0.2
−0.2

0
1.5 −0.4
1
0.5 1.5
0 1 −0.6
0.5
−0.5 0
y −1 −0.5
−1 x −0.8
−1.5 −1.5
−1.5 −1 −0.5 0 0.5 1 1.5
x

Fig. 6.68 Control of the quadrotor in the disturbance-free case: a trajectory of the quadrotor in the
cartesian space, b projection of the quadrotor’s trajectory in the x y plane
6.7 State Estimation-Based Control of Quadrotors 319

(a) 2 (b) 2

1 1

3
1

0 0
f

yf
y

−1 −1

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
1 1

0.5 0.5
f2

f4
0 0
y

y
−0.5 −0.5

−1 −1
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.69 Control of the quadrotor in the disturbance-free case: a position and velocity along the
x axis, b position and velocity along the y axis

1.5 1.5

1
1
7

0.5
f5

f
y
y

0.5
0

0 −0.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.1 0.6
0.08 0.4
0.06
8
6

0.2
f
yf

0.04
0.02 0

0 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.70 Control of the quadrotor in the disturbance-free case: a position and velocity along the
z axis, b rotation angle φ and associated angular speed

control enabled accurate tracking of the reference trajectories. There has been con-
vergence of the linear position and velocity variables to the associated setpoints as it
can be seen in Fig. 6.73a, b and in Fig. 6.74a. Moreover, there has been convergence
of the angular position and velocity variables to the associated setpoints as it can be
seen in Fig. 6.74b and in Fig. 6.75a, b.
320 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 2 (b) 6

1 4

11
9
f

yf
y

0 2

−1 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.6 4

0.4 3
10

12
0.2 2
f

f
y

y
0 1

−0.2 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.71 Control of the quadrotor in the disturbance-free case: a rotation angle θ and associated
angular speed, b rotation angle ψ and associated angular speed

(a) 0.2 (b) 0.2

d1

d4

0 0

−0.2 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.5 0.5
d2

d5

0 0

−0.5 −0.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.5 0.1
d3

d6

0 0

−0.5 −0.1
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.72 Use of the Derivative-free nonlinear Kalman Filter in estimation of disturbances: a
associated with linear motion, b associated with the rotational motion of the vehicle (blue line real
value, green line estimated value)

6.8 State Estimation-Based Control of the Underactuated

Hovercraft

6.8.1 Overview

Another significant application field for differential flatness theory-based methods

for filtering and control is localization and autonomous navigation of unmanned
6.8 State Estimation-Based Control of the Underactuated Hovercraft 321

(a) 2 (b) 2

1 1
1

3
0 0
f

f
y

y
−1 −1

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
1 4

0.5
2

4
2

yf
yf

0
−0.5

−1 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.73 Control of the quadrotor in the presence of external disturbances a position and velocity
along the x axis, b position and velocity along the y axis (blue line real value, green line estimated
value, red line setpoint)

(a) 1.5 (b) 0.2

0.1
1
5

0
yf

f
y

0.5
−0.1

0 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.4 0.5

0.2
0
6

0
f

f
y

−0.5
−0.2

−0.4 −1
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.74 Control of the quadrotor in the presence of external disturbances: a position and velocity
along the z axis, b rotation angle φ and associated angular speed (blue line real value, green line
estimated value, red line setpoint)

surface vessels (USVs) and autonomous underwater vessels (AUVs). In particular,

the problem of autonomous navigation of surface vessels such as hovercrafts has
received particular attention, since it can find use in both security purposes and
passenger transportation [114, 401, 451, 492, 509]. The problem of control and
trajectory tracking for unmanned surface vessels of the hovercraft type is nontrivial
because the associated kinematic model is a high-order nonlinear one [17, 128, 184,
499, 531]. Another problem that has to be dealt with is that the hovercraft’s model
322 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

(a) 2 (b) 10
1 8
6

11
9

0
f

yf
y

4
−1 2
−2 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
5 20

0 10
10

12
yf

yf
−5 0

−10 −10
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 6.75 Control of the quadrotor in the presence of disturbances: a rotation angle θ and associated
angular speed, b rotation angle ψ and associated angular speed (blue line real value, green line
estimated value, red line setpoint)

is underactuated [2, 34, 41, 191, 384, 486, 502]. Indicative results on control of
underactuated dynamical systems can be found in [399, 400, 445]. Moreover, as
explained in Chap. 2, the hovercraft’s model cannot be subjected to static feedback
linearization, but admits only dynamic feedback linearization. This means that to
succeed linearization, the state-space description of the system has to be augmented
by considering as additional state variables the control inputs and their derivatives.
Thus, finally the control input that is applied to the vessel contains integral terms of
the tracking error. The present chapter proposes a solution to the control problem
of hovercrafts control with the use of differential flatness theory and of a nonlinear
filtering method, the so-called Derivative-free nonlinear Kalman Filter.
First it is shown that the hovercraft’s model is differentially flat. This means that
all its state variables and the control inputs can be written as functions of one single
algebraic variable which is the flat output and of the flat output’s derivatives [55, 152,
263, 286, 340, 427, 464, 465, 495]. By exploiting the differential flatness properties it
is shown that the system can be transformed into the linear canonical form, through
dynamic feedback linearization. To succeed this, dynamic extension is performed
which means that the state-vector’s dimension is increased by considering as addi-
tional state variables certain control inputs and their derivatives. For the linearized
equivalent of the system, the design of a state feedback controller is possible, through
the use of pole placement techniques. Next, to estimate the nonmeasurable state vari-
ables of the vessel and to identify additive disturbance terms that affect the system,
the Derivative-free nonlinear Kalman Filter is redesigned as a disturbance observer
[31, 408, 414, 435–437, 447]. This estimation algorithm consists of the standard
6.8 State Estimation-Based Control of the Underactuated Hovercraft 323

Kalman Filter recursion applied on the linearized equivalent of the hovercraft and of
an inverse transformation that is based on differential flatness theory which permits
to compute estimates of the state variables of the initial nonlinear system.

6.8.2 Lie Algebra-Based Control of the Underactuated

Hovercraft

In Chap. 2, it was shown that the extended model of the vessel is a differentially flat
one. Using its differential flatness properties, the vessel’s model was transformed
into the linear canonical (Brunovsky) form. It will be shown that one can arrive at
an equivalent description of the hovercraft’s dynamics using Lie algebra [461]. The
sixth-order state-space equation of the underactuated hovercraft model (Figs. 2.3 and
6.76) is given by
ẋ = ucos(ψ) − vsin(ψ)
ẏ = usin(ψ) + vcos(ψ)
ψ̇ = r
(6.166)
u̇ = v·r + τu
v̇ = −u·r − βv
ṙ = τr

where x and y are the cartesian coordinates of the vessel, ψ is the orientation angle,
u is the surge velocity, v is the sway velocity, and r is the yaw rate. The hovercraft’s
model is also written in the matrix form:

Fig. 6.76 Underactuated hovercraft performing maneuvers and the associated reference frames
324 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ ucos(ψ) − vsin(ψ) 0 0
⎜ ẏ ⎟ ⎜usin(ψ) + vcos(ψ)⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ 
⎜ψ̇ ⎟ ⎜ ⎟ ⎜0 0⎟
⎜ ⎟=⎜ r ⎟+⎜ ⎟ τu (6.167)
⎜ u̇ ⎟ ⎜ ⎟ ⎜1 0⎟
⎜ ⎟ ⎜ vr ⎟ ⎜ ⎟ τr
⎝ v̇ ⎠ ⎝ −ur + βv ⎠ ⎝0 0⎠
ṙ 0 0 1

or equivalently, one has the description

x̃˙ = f (x̃) + g ṽ (6.168)

The system’s state vector is denoted as x̃ = [x, y, ψ, u, v, r ]T , f (x̃)∈R 6×1 , and

g(x̃) = [ga , gb ]∈R 6×2 , while the control input is the vector ṽ = [τu , τr ]T .
The system’s state vector is extended by including as additional state variables the
control input τu and its first derivative τ̇u . These are denoted as z 1 = τu and z 2 = τ̇u .
The extended state-space description of the system becomes
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ ucos(ψ) − vsin(ψ) 0 0
⎜ ẏ ⎟ ⎜usin(ψ) + vcos(ψ)⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜ ψ̇ ⎟ ⎜ r ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ 
⎜ u̇ ⎟ ⎜ vr + z 1 ⎟ ⎜0 0⎟
⎜ ⎟=⎜ ⎟+⎜ ⎟ τ̈u (6.169)
⎜ v̇ ⎟ ⎜ −ur + βv ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟ τr
⎜ ṙ ⎟ ⎜ 0 ⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝ż 1 ⎠ ⎝ z2 ⎠ ⎝0 0⎠
ż 2 0 1 0

or equivalently, one has the description

ż = f (z) + g(z)ṽ (6.170)

The extended system’s state vector is denoted as z = [x, y, ψ, u, v, r, z 1 , z 2 ]T . More-

over, one has f (z)∈R 8×1 , and g(z) = [ga , gb ]∈R 8×2 , while the control input is the
vector ṽ = [τ̈u , τr ]T . Next, the following system outputs are defined

z 1.1 = x z 2,1 = y (6.171)

Moreover, the new state variables are defined

z 1,2 = L f z 1,1 z 2,2 = L f z 2,1

z 1,3 = L 2f z 1,1 z 2,3 = L 2f z 2,1 (6.172)
z 1,4 = L 3f z 1,1 z 2,4 = L 3f z 2,1
6.8 State Estimation-Based Control of the Underactuated Hovercraft 325

The system will be brought to a linearized input-output form using

ż 1,4 = L 4f z 1,1 + L ga L 3f z 1,1 τ̈u + L gb L 3f z 1,1 τr

(6.173)
ż 2,4 = L 4f z 2,1 + L ga L 3f z 2,1 τ̈u + L gb L 3f z 2,1 τr

It holds that z 1,1 = x. Thus one has

z 1,2 = L f z 1,1 ⇒
∂z 1,1 ∂z 1,1 ∂z 1,1 ∂z ∂z ∂z ∂z 1,1 ∂z 1,1
z 1,2 = ∂ x f1 + ∂ y f2 + ∂ψ f 3 + ∂u1,1 f 4 + ∂v1,1 f 5 + ∂r1,1 f 6 + ∂z 1 f 7 + ∂z 2 f 8 ⇒
z 1,2 = 1· f 1 ⇒z 1,2 = ucos(ψ) − vsin(ψ)
(6.174)
Similarly, one obtains

z 1,3 = L 2f z 1,1 ⇒
∂z 1,2 ∂z 1,2 ∂z 1,2 ∂z ∂z ∂z ∂z ∂z
z 1,3 = ∂ x f1 + +
∂ y f2 + ∂u1,2 f 4 + ∂v1,2 f 5 + ∂r1,2 f 6 + ∂z1,2
∂ψ f 3 1
f 7 + ∂z1,2
2
f8 ⇒
z 1,3 = (−usin(ψ) − vcos(ψ)) f 3 + cos(ψ) f 4 − sin(ψ) f 5 ⇒
z 1,3 = (−usin(ψ) − vcos(ψ))r + cos(ψ)(vr + z 1 ) − sin(ψ)(−ur + βv)⇒
z 1,3 = τu cos(ψ) + βvsin(ψ)
(6.175)
Equivalently, it holds that

z 1,4 = L 3f z 1,1 ⇒z 1,4 = L f z 1,3 ⇒

∂z 1,3 ∂z 1,3 ∂z 1,3 ∂z 1,3 ∂z 1,3 ∂z 1,3 ∂z 1,3 ∂z 1,3
z 1,4 = ∂ x f1 + ∂ y f2 + ∂ψ f 3 + ∂u f 4 + ∂v f 5 + ∂r f 6 + ∂z 1 f 7 + ∂z 2 f 8
(6.176)

while after intermediate operations one obtains

z 1,4 = (−τu sin(ψ) + βvcos(ψ)) f 3 + βsin(ψ) f 5 + cos(ψ) f 7 ⇒

z 1,4 = (−τu sin(ψ) + βcos(ψ))r + βsin(ψ)(−ur + βv) + cos(ψ)z 2 ⇒ (6.177)
z 1,4 = (τ̇u sin(ψ) + βvcos(ψ))r + βsin(ψ)(−ur + βv) + cos(ψ)z 2

or, using the extended state vector variables notation one has

z 1,4 = z 2 cos(ψ) − z 1 sin(ψ)r − βur sin(ψ) − β 2 vsin(ψ) + βvcos(ψ)r (6.178)

It also holds that

ż 1,4 = L 4f z 1,1 + L ga L 3f z 1,1 τ̈u + L gb L 3f z 1,1 τr ⇒

(6.179)
ż 1,4 = L f z 1,4 + L ga z 1,4 τ̈u + L gb z 1,4 τr

where
∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4
L f z 1,4 = ∂ x f1 + ∂ y f2 + ∂ψ f 3 + ∂u f 4 + ∂v f 5 + ∂r f 6 + ∂z 1 f 7 + ∂z 2 f 8
(6.180)
326 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

which gives

L f z 1,4 = (−z 2 sin(ψ) − z 1 cos(ψ)r − βur cos(ψ) − β 2 vcos(ψ) − βvsin(ψ)r )r +

+(−βrsin(ψ))(vr + z 1 ) + (−β 2 sin(ψ) + βcos(ψ)r )(−ur + βv)+
+(−z 1 sin(ψ) − βvsin(ψ) + βvcos(ψ))0 + (−sin(ψ)r )z 2
(6.181)
while after some intermediate computations one obtains

L f z 1,4 = −2z 2 sin(ψ)r − z 1 cos(ψ)r 2 −

−βvr 2 sin(ψ) − βz 1 rsin(ψ)−
−βur 2 cos(ψ) + β 2 ur sin(ψ)− (6.182)
−β vsin(ψ) − β 2 vr cos(ψ) − βur 2 cos2 (ψ) + β 2 vr cos(ψ)
3

−βvr 2 sin(ψ)

In a similar manner, one computes

∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4

L ga z 1,4 = ∂ x ga1 + ∂ y ga2 + ∂ψ ga3 + ∂u ga4 +
∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4
+ ∂v ga5 + ∂r ga6 + ∂z 1 ga7 + ∂z 2 ga8
(6.183)
∂z
L ga z 1,4 = ∂z1,42
⇒L ga z 1,4 = cos(ψ)

and also
∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4
L gb z 1,4 = ∂ x gb1 + ∂ y gb2 + ∂ψ gb3 + ∂u gb4 +
∂z 1,4 ∂z 1,4 ∂z 1,4 ∂z 1,4
+ ∂v gb5 + ∂r gb6 + ∂z 1 gb7 + ∂z 2 gb8
(6.184)
∂z 1,4
L gb z 1,4 = ∂r ⇒L gb z 1,4 = −z 1 sin(ψ) − βusin(ψ) + βvcos(ψ)

In an equivalent way, and using that z 2,1 = y2 = y one can compute

∂z 2,1 ∂z 2,1 ∂z 2,1 ∂z 2,1

z 2,2 = L f z 2,1 ⇒z 2,2 = ∂ x f 1 + ∂ y f 2 + ∂ψ f 3 + ∂u f 4 +
∂z ∂z ∂z ∂z
+ ∂v2,1 f 5 + ∂r2,1 f 6 + ∂z2,1
1
f 7 + ∂z2,1
2
f8⇒ (6.185)
z 2,2 = 1· f 2 ⇒z 2,2 = usin(ψ) + vcos(ψ)

Equivalently, one has

z 2,3 = L 2f z 2,1 ⇒z 2,3 = L f z 2,2 ⇒

∂z 2,2 ∂z 2,2 ∂z ∂z ∂z ∂z ∂z ∂z
z 2,3 = ∂ x f1 + + ∂ψ2,2 f 3 + ∂u2,2 f 4 + ∂v2,2 f 5 + ∂r2,2 f 6 + ∂z2,2
∂ y f2 1
f 7 + ∂z2,2
8
f8 ⇒
z 2,3 = (ucos(ψ) − vsin(ψ))r + sin(ψ)(vr + z 1 ) + cos(ψ)(−ur + βv)⇒
z 2,3 = z 1 sin(ψ) + βvcos(ψ)
(6.186)
6.8 State Estimation-Based Control of the Underactuated Hovercraft 327

In an equivalent manner, one obtains

z 2,4 = L 3f z 2,1 ⇒z 3,3 = L f z 2,3 ⇒

∂z 2,3 ∂z 2,3 ∂z ∂z ∂z ∂z ∂z ∂z
z 2,4 = ∂ x f1 + + ∂ψ2,3 f 3 + ∂u2,3 f 4 + ∂v2,3 f 5 + ∂r2,3 f 6 + ∂z2,3
∂ y f2 1
f 7 + ∂z2,3
2
f8 ⇒
z 2,4 = (cos(ψ) − βvsin(ψ)) f 3 + βcos(ψ) f 5 + sin(ψ) f 7 ⇒
z 2,4 = (z 1 cos(ψ) − βvsin(ψ))r + (βcos(ψ)(−ur + βv) + sin(ψ)z 2 ⇒
z 2,4 = z 1 cos(ψ)r − βvr sin(ψ) − βur cos(ψ) + β 2 vcos(ψ) + z 2 sin(ψ)
(6.187)
Moreover, it holds that

ż 2,4 = L 4f z 2,1 + L ga L 3f z 2,1 τ̈u + L gb L 3f z 2,1 τr (6.188)

where

L 4f z 2,1 = L f z 2,4 ⇒
∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4
L 4f z 2,1 = ∂ x f1 + ∂ y f2 + ∂ψ f 3 + ∂u f 4 + ∂v f 5 + ∂r f 6 + ∂z 1 f 7 + ∂z 2 f 8
(6.189)
which gives

L 4f z 2,1 = [−z 1 sin(ψ)r − βvr cos(ψ) + βur sin(ψ) − β 2 vsin(ψ) + z 2 cos(ψ)]r +

[−βr cos(ψ)](vr + z 1 ) + [−βrsin(ψ) + β 2 cos(ψ)](−ur + βv)+
[z 1 cos(ψ) − βvsin(ψ) − βucos(ψ)]0 + [cos(ψ)r ]z 2 + [sin(ψ)]0
(6.190)
and after additional computations one arrives at the form

L 4f z 2,1 = −z 1 r 2 sin(ψ) − βvr 2 cos(ψ) + βur 2 sin(ψ) − β 2 vr sin(ψ) + z 2 r cos(ψ)−

−βvr 2 cos(ψ) − βr z 1 cos(ψ) + βur 2 sin(ψ) − β 2 r vsin(ψ)−
−β 2 ur cos(ψ) + β 2 vcos(ψ) + z 2 r cos(ψ)
(6.191)
Proceeding as before, one computes

L ga L 3f z 2,1 = L ga z 2,4 ⇒
∂z ∂z 2,4 ∂z 2,4 ∂z 2,4
L ga L 3f z 2,1 = ∂2,4 x ga1 + ∂ y ga2 + ∂ψ ga3 + ∂u ga4 +
∂z ∂z ∂z ∂z (6.192)
+ ∂v2,4 ga5 + ∂v2,4 ga6 + ∂z2,4 1
ga7 + ∂z2,4 2
ga8 ⇒
∂z
L ga L 3f z 2,1 = ∂z2,4 2
⇒L L 3z
ga f 2,1 = sin(ψ)

Equivalently, one computes

L gb L 3f z 2,1 = L gb z 2,4 ⇒
∂z 2,4 ∂z 2,4 ∂z 2,4 ∂z 2,4
L gb L 3f z 2,1 = ∂ x gb1 + ∂ y gb2 + ∂ψ gb3 + ∂u gb4 +
∂z ∂z ∂z ∂z
+ ∂v2,4 gb5 + ∂r2,4 gb6 + ∂z2,4 1
gb7 + ∂z2,4 2
gb8 ⇒ (6.193)
∂z
L gb L f z 2,1 = ∂v ⇒
3 2,4

L gb L 3f z 2,1 = z 1 cos(ψ) = βvsin(ψ) − βucos(ψ)

328 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

The aggregate dynamics of the input-output linearized system is

x (4) = L 4f z 1,1 + L ga L 3f z 1,1 τ̈u + L gb L 3f z 1,1 τr

(6.194)
y (4) = L 4f z 2,1 + L ga L 3f z 2,1 τ̈u + L gb L 3f z 2,1 τr

By defining the new control inputs

v1 = L 4f z 1,1 + L ga L 3f z 1,1 τ̈u + L gb L 3f z 1,1 τr

(6.195)
v2 = L 4f z 2,1 + L ga L 3f z 2,1 τ̈u + L gb L 3f z 2,1 τr

one arrives at the following description for the input-output linearized system

x (4) = v1
(6.196)
y (4) = v2

which can be also written in the state-space form

z̃˙ = Az̃ + B ṽ
(6.197)
z̃ m = C z̃

or using that z̃ = [z 1,1 , z 1,2 , z 1,3 , z 1,4 , z 2,1 , z 2,2 , z 2,3 , z 2,4 ]T , equivalently one has
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1,1 0 1 0 0 0 0 0 0 z 1,1 0 0
⎜ż 1,2 ⎟ ⎜0 0 1 0 0 0 0 0⎟ ⎜z 1,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ż 1,3 ⎟ ⎜0 0 0 1 0 0 0 0⎟ ⎜ ⎟ ⎜ 0⎟
⎜ ⎟ ⎜ ⎟ ⎜z 1,3 ⎟ ⎜0 ⎟ 
⎜ż 1,4 ⎟ ⎜0 0 0 0 0 0 0 0⎟ ⎜z 1,4 ⎟ ⎜1 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟+⎜ ⎟ v1
⎜ż 2,1 ⎟ = ⎜0 0 0 0 0 1 0 0⎟ ⎜ ⎟ ⎜ 0⎟
(6.198)
⎜ ⎟ ⎜ ⎟ ⎜z 2,1 ⎟ ⎜0 ⎟ v2
⎜ż 2,2 ⎟ ⎜0 0 0 0 0 0 1 0⎟ ⎜z 2,2 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ż 2,3 ⎠ ⎝0 0 0 0 0 0 0 1⎠ ⎝z 2,3 ⎠ ⎝0 0⎠
ż 2,4 0 0 0 0 0 0 0 0 z 2,4 0 1

while the associated measurement equation is

⎞ ⎛
z 1,1
⎜z 1,2 ⎟
⎜ ⎟
⎜ ⎟
 m   ⎜z 1,3 ⎟
z1 ⎜
1 0 0 0 0 0 0 0 ⎜z 1,4 ⎟ ⎟
= (6.199)
m
z2 00001000 ⎜ ⎜z 2,1 ⎟
⎟
⎜z 2,2 ⎟
⎜ ⎟
⎝z 2,3 ⎠
z 2,4
6.8 State Estimation-Based Control of the Underactuated Hovercraft 329

A suitable feedback control law for the linearized system is

(4) (3)
v1 = xd − k11 (x (3) − xd ) − k21 (ẍ − ẍd ) − k31 (ẋ − ẋd ) − k41 (x − xd )
(6.200)
v2 = yd(4) − k12 (y (3) − yd(3) ) − k22 ( ÿ − ÿd ) − k32 ( ẏ − ẏd ) − k42 (y − yd )

One can also compute the control input that is finally applied to the hovercraft model.
It holds that
ṽ = f˜ + M̃ ũ (6.201)

where matrices and vectors ṽ, f˜, M̃, and ũ are defined as
     
v1 ˜ L 4f z 1,1 L g,a L 3f z 1,1 L gb L 3f z 1,1 τ̈u
ṽ = f = M̃ = ũ = (6.202)
v2 L 4f z 2,1 3 3
L g,a L f z 2,1 L g,b L f z 2,1 τr

The above equation can be solved with respect to ũ, which finally gives

ũ = f˜ + M̃ −1 (ṽ − f˜) (6.203)

The vector ũ is the control input that is finally applied to the system, which finally
means that the control signal contains integrals of the tracking error.
Since in the case of transformation of the nonlinear vessel dynamics to an equiv-
alent linear form, with the use of differential flatness theory, one arrives at the state-
space model of Eqs. (6.198) and (6.199). The design of the flatness-based feedback
controller for the hovercraft is also given by Eqs. (6.198)–(6.203).

6.8.3 Flatness-Based Control of the Underactuated Vessel

In Chap. 2 it has been proven that the model of the underactuated vessel given in
Eq. (6.166) is a differentially flat one. By exploiting differential flatness properties,
the hovercraft’s dynamic model was transformed into the canonical form and a state
feedback controller was designed for it.
For the linearized equivalent of the system it is also possible to perform state
estimation using the Derivative-free nonlinear Kalman Filter. Before computing the
Kalman Filter stages, the matrices A, B, and C previously defined in Eq. (6.197) are
substituted by their discrete-time equivalents Ad , Bd , and Cd . This is done through
common discretization methods. The recursion of the filter’s algorithm consists of
two stages:
measurement update:

K (k) = P − CdT [P − CdT P + R]−1

ẑ(k) = ẑ − (k) − K (k)[Cd z(k) − Cd ẑ − (k)] (6.204)
P(k) = P − (k) − K (k)Cd P − (k)
330 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

time update:
P − (k + 1) = AdT P(k)Ad + Q(k)
(6.205)
ẑ − (k + 1) = Ad ẑ(k) + Bd u(k)

Moreover, by using the nonlinear transformations which are provided by differential

flatness theory according to Eqs. (2.88), (2.91), (2.93), and (2.98) one can obtain
estimates of the state variables of the initial nonlinear hovercraft model.

6.8.4 Disturbances’ Compensation with the Use

of the Derivative-Free Nonlinear Kalman Filter

Next, it is assumed that the input-output linearized equivalent of the system, is sub-
jected to disturbance terms which express the effects of both modeling uncertainty
and of external perturbations. Thus one has

x (4) = v1 + d̃1
(6.206)
y (4) = v2 + d̃2

It is considered that the disturbance signals are equivalently represented by their

time derivatives (up to order n) and by the associated initial conditions (however
since disturbances are to be estimated with the use of the Kalman Filter, finally the
dependence on knowledge of initial conditions becomes obsolete). It holds that
(n) (n)
d̃1 = f d1 d̃2 = f d2 (6.207)

The system’s state vector is extended by including as additional state variables the
disturbances’ derivatives. Thus, taking that the derivative’s order is n = 2 one has

z d,1 = d̃1 z d,2 = d̃˙1 z d,3 = d̃2 z d,4 = d̃˙2 (6.208)

and the extended state-space description of the system becomes

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1,1 010000000000 z 1,1 0000
⎜ ż 1,2 ⎟ ⎜0 0 1 0 0 0 0 0 0 0 0 0⎟ ⎜ z 1,2 ⎟ ⎜0 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ż 1,3 ⎟ ⎜0 0 0 1 0 0 0 0 0 0 0 0⎟ ⎜ z 1,3 ⎟ ⎜0 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ż 1,4 ⎟ ⎜0 0 0 0 0 0 0 0 1 0 0 0⎟ ⎜ z 1,4 ⎟ ⎜1 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎛ ⎞
⎜ ż 2,1 ⎟ ⎜0 0 0 0 0 1 0 0 0 0 0 0⎟ ⎜ z 2,1 ⎟ ⎜0 0 0 0⎟ τ̈u
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ż 2,2 ⎟ ⎜0 0 0 0 0 0 1 0 0 0 0 0⎟ ⎜ z 2,2 ⎟ ⎜0 0 0 0⎟ ⎜ τr ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟+⎜ ⎟⎜ ⎟ (6.209)
⎜ ż 2,3 ⎟ ⎜0 0 0 0 0 0 0 1 0 0 0 0⎟ ⎜ z 2,3 ⎟ ⎜0 0 0 0⎟ ⎝ f d ⎠
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ 1
⎜ ż 2,4 ⎟ ⎜0 0 0 0 0 0 0 0 0 0 1 0⎟ ⎜ z 2,4 ⎟ ⎜0 1 0 0⎟ f d
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ 2
⎜ż d,1 ⎟ ⎜0 0 0 0 0 0 0 0 0 1 0 0⎟ ⎜z d,1 ⎟ ⎜0 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ż d,2 ⎟ ⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜z d,2 ⎟ ⎜0 0 1 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ż d,3 ⎠ ⎝0 0 0 0 0 0 0 0 0 0 0 1⎠ ⎝z d,3 ⎠ ⎝0 0 0 0⎠
ż d,4 000000000000 z d,4 0001
6.8 State Estimation-Based Control of the Underactuated Hovercraft 331

while the associated measurement equation is

   
z 1,1 1000000000
= z (6.210)
z 2,1 0001000000 e

where z e = [z 1,1 , z 1,2 , z 1,3 , z 1,4 , z 2,1 , z 2,2 , z 2,3 , z 2,4 , z d,1 , z d,2 , z d,3 , z d,4 ]T is the
extended state vector. Thus, the extended state-space description of the hovercraft
model takes the form
ż e = Ae z e + Be ve
(6.211)
z emeas = Ce z e

For the extended state-space description of the system, one can design a state esti-
mator of the form
ẑ˙ e = Ao z e + Bo ve + K (z emeas − Co ẑ e )
(6.212)
ẑ emeas = Co ẑ e

where for the matrices Ao and Co it holds Ao = Ae and Co = Ce while for matrix
Bo one has ⎛ ⎞
000100000000
⎜0 0 0 0 0 0 0 1 0 0 0 0 ⎟
BoT = ⎜
⎝0 0 0 0 0 0 0 0 0 0 0 0⎠
⎟ (6.213)
000000000000

Again the Kalman Filter recursion provides joint estimation of the nonmeasurable
state vector elements, of the disturbances’ inputs and of their derivatives. Prior to
computing the Kalman Filter stages, the previously defined matrices A,B and C are
substituted by their discrete-time equivalents Aed ,Bed and Ced . This is done through
common discretization methods. The recursion of the filter’s algorithm consists of
two stages. Thus, one has
measurement update:

K (k) = Pe− CeTd [Pe− Ced T Pe + Re ]−1

ẑ e (k) = ẑ e− (k) − K (k)[Ced z e (k) − Ced ẑe− (k)] (6.214)
Pe (k) = Pe− (k) − K (k)Ced Pe− (k)

time update:

Pe− (k + 1) = Aed T Pe (k)Aed + Q e (k)

(6.215)
ẑ e− (k + 1) = Aed zˆe (k) + Bed ve (k)

To compensate for the disturbances effects, the modified control input that is applied
to the system is

(4) (3)
v1 = xd − k11 (x (3) − xd ) − k21 (ẍ − ẍd ) − k31 (ẋ − ẋd ) − k41 (x − xd ) − ẑ d,1
(4) (3)
v2 = yd − k12 (y (3) − yd ) − k22 (ẍ − ÿd ) − k32 ( ẏ − ẏd ) − k42 (y − yd ) − ẑ d,2
(6.216)
332 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

6.8.5 Simulation Tests

The performance of the proposed nonlinear control scheme was evaluated in the
case of several reference setpoints. The associated results are presented in Figs. 6.77,
6.78, 6.79, 6.80, and 6.81. It can be observed that in all cases the nonlinear feed-
back controller succeeded fast and accurate tracking of the reference setpoints. The
Derivative-free nonlinear Kalman Filter enabled estimation of the nonmeasurable
variables of the system’s state-vector which were needed for the implementation of
the feedback control scheme. Moreover, by using the inverse transformation that was
provided by differential flatness theory it was possible to obtain estimates of the state
variables of the initial nonlinear system.
The convergence of the state variables of the hovercraft (position x,y to the desir-
able setpoints is shown in Figs. 6.77a, 6.78a, 6.79a, 6.80a, and 6.81a. The estimation
of the disturbance terms that were applied to the hovercraft model are depicted in
Figs. 6.77b, 6.78b, 6.79b, 6.80b and 6.81b, respectively. It can be noticed again that
the proposed feedback nonlinear control scheme succeeded fast and accurate tracking
to these setpoints.
For the underactuated hovercraft, one can succeed exactly the same motion and
orientation control as in the case of the fully actuated vessel. Therefore, it is possible
for the hovercraft to track complicated reference paths with excellent accuracy while
keeping also the desirable velocity. This has been demonstrated through a series of
examples, in the simulation tests section of the manuscript (Figs. 6.77, 6.78, 6.79,
6.80, and 6.81). It is noteworthy that the dynamic feedback linearization procedure
which has been implemented on the hovercraft’s model, results in the canonical

(a) (b)
11 0.8 0.15

10 0.6 0.1
d/dt d1

9
d1

0.4 0.05
8
0.2 0
7
0 −0.05
6 0 10 20 30 40 0 10 20 30 40
time time
y

5
0.6 0.3
4
0.4 0.2
d/dt d2

3
d2

0.2 0.1
2

1 0 0

0 −0.2 −0.1
0 1 2 3 4 5 6 7 8 9 10 11 0 10 20 30 40 0 10 20 30 40
x time time

Fig. 6.77 Reference path 1. a Trajectory tracking for states x, y of the underactuated hovercraft.
b Estimation of disturbance inputs affecting the hovercraft, with the use of the Derivative-free
nonlinear Kalman Filter
6.8 State Estimation-Based Control of the Underactuated Hovercraft 333

(a)6 (b) 2 0.6

1.5 0.4
4

d/dt d1
d1
1 0.2

2 0.5 0

0 −0.2
0 10 20 30 40 0 10 20 30 40
y

0 time time
4 1

−2
2 0.5

d/dt d2
d2
−4
0 0

−6 −2 −0.5
−6 −4 −2 0 2 4 6 0 10 20 30 40 0 10 20 30 40
x time time

Fig. 6.78 Reference path 2. a Trajectory tracking for states x, y of the underactuated hovercraft.
b Estimation of disturbance inputs affecting the hovercraft, with the use of the Derivative-free
nonlinear Kalman Filter

(a)6 (b) 1.5 0.3

1 d/dt d1 0.2
4
d1

0.5 0.1

2 0 0

−0.5 −0.1
0 10 20 30 40 0 10 20 30 40
0
y

time time
2 0.6
−2 0.4
d/dt d2

1
d2

0.2
−4
0
0

−6 −1 −0.2
−6 −4 −2 0 2 4 6 0 10 20 30 40 0 10 20 30 40
x time time

Fig. 6.79 Reference path 3. a Trajectory tracking for states x, y of the underactuated hovercraft.
b Estimation of disturbance inputs affecting the hovercraft, with the use of the Derivative-free
nonlinear Kalman Filter

form description of Eqs. (2.123) and (2.124) which is confirmed to be controllable.

Practically, this means that under the proposed control scheme, the vessel can reach
any point in its motion plane and can track any reference path.
The possibility for the appearance of singularities in the computation of the control
signal of the hovercraft is also present in all static or dynamic feedback linearization
algorithms which arrive at a transformed control input of the form v = f (x, t) +
g(x, t)u, that is u = g(x, t)−1 [v − f (x, t)]. There are two cases: (i) due to the
334 6 Differential Flatness Theory in Mobile Robotics and Autonomous Vehicles

3 0.6
(a) (b)
10
0.4

d/dt d1
2

d1
0.2
5 1
0

0 −0.2
0 10 20 30 40 0 10 20 30 40
0
y

time time
10 2
8
−5

d/dt d2
1
6

d2
4
0
−10 2
0 −1
−10 −5 0 5 10 0 10 20 30 40 0 10 20 30 40
x time time

Fig. 6.80 Reference path 4. a Trajectory tracking for states x, y of the underactuated hovercraft.
b Estimation of disturbance inputs affecting the hovercraft, with the use of the Derivative-free
nonlinear Kalman Filter

(a) 15 (b) 4 1

3
0.5
10
d/dt d1
d1

2
0
1
5

0 −0.5
0 10 20 30 40 0 10 20 30 40
0
y

time time
10 2

−5 8
1
d/dt d2

6
d2

−10 4
0
2

−15 0 −1
−15 −10 −5 0 5 10 15 0 10 20 30 40 0 10 20 30 40
x time time

Fig. 6.81 Reference path 5. a Trajectory tracking for states x, y of the underactuated hovercraft.
b Estimation of disturbance inputs affecting the hovercraft, with the use of the Derivative-free
nonlinear Kalman Filter

inherent model of g(x, t) its inverse never becomes 0. In such a case, the singularity
problem is avoided, (ii) for certain areas of the state vector space x∈R n the zeroing
of g(x, t)−1 becomes possible. For the latter case, the avoidance of singularities can
be succeeded by a state variable transformation into a new state-space representation
which does not include any points of singularity.
Finally, it is noted that the presented simulation experiments have been performed
under the assumption that the hovercraft was subjected to external disturbances such
6.8 State Estimation-Based Control of the Underactuated Hovercraft 335

as wind or current. The proposed control scheme is robust to modeling uncertainties

and external perturbations. This is a prerequisite in the design of control systems
for underactuated surface vessels [129, 176, 312, 359, 482, 602]. First, it has been
proven that the feedback control applied on the input-output linearized model of
the hovercraft succeeded the placement of all poles of the control loop in the left
complex semiplane. Next, it can be confirmed that the extended state-space model of
the hovercraft, which contains disturbances as additional state variables, has multiple
poles at the origin (multiple poles at zero). Thus, stabilization can be succeeded using
pole placement methods. With the use of the Derivative-free nonlinear Kalman Filter,
it became possible to identify the perturbation and modeling uncertainty terms in real-
time and subsequently to compensate for them through the inclusion of an additional
term in the control signal. This amendment in the feedback control scheme provided
the control loop with additional robustness features. Finally, it is worth mentioning
that the proposed control scheme had an excellent performance although it was
not possible to measure directly all elements of the state vector (only the cartesian
coordinates of the vessel could be measured) and several state variables had to be
estimated with the use of filtering.
Chapter 7
Differential Flatness Theory and Electric
Power Generation

7.1 Outline

This chapter examines the use of control and filtering methods based on differential
flatness theory, in the problem of electric power generation. In the design of nonlinear
controllers for power generators it is important to estimate the nonmeasurable state
variables for computing the feedback control signal. First, a Derivative-free non-
linear Kalman Filtering approach is introduced aiming at implementing sensorless
control of the Permanent Magnet Synchronous Generator (PMSG). In the proposed
Derivative-free Kalman Filtering method the system is first subject to a linearization
transformation that is based on the differential flatness theory and next state estima-
tion is performed by applying the standard Kalman Filter recursion to the linearized
model. Unlike the Lie algebra-based estimator design method, the proposed approach
provides estimates of the state vector of the permanent magnet synchronous generator
without the need for derivatives and Jacobians calculation. By avoiding linearization
approximations, the proposed filtering method improves the accuracy of estimation
of the system state variables, and results in smooth control signal variations and in
minimization of the tracking error of the associated control loop.
Next, the chapter studies differential flatness properties and an input–output lin-
earization procedure for doubly-fed induction generators (DFIGs). The complete
sixth-order model of the doubly-fed induction generator is derived with the use of
stator and rotor electrical equations. By defining flat outputs which are associated with
the rotor’s turn angle and the magnetic flux of the stator an equivalent DFIG descrip-
tion in the Brunovksy (canonical) form is obtained. For the linearized canonical model
of the generator a feedback controller is designed. Moreover, a comparison of the
differential flatness theory-based control method against Lie algebra-based control is
provided. At a second stage, a novel Kalman Filtering method (Derivative-free non-
linear Kalman Filtering) is introduced. The proposed Kalman Filter is redesigned as
disturbance observer for estimating additive input disturbances to the DFIG model.
These estimated disturbance terms are finally used by a feedback controller that

© Springer International Publishing Switzerland 2015 337

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_7
338 7 Differential Flatness Theory and Electric Power Generation

enables the generator’s active and reactive power to track desirable setpoints. The
efficiency of the proposed state estimation-based control scheme is tested through
simulation experiments.
Moreover, a sensorless control scheme for doubly-fed induction generators
(DFIG) is developed using an implementation of flatness-based control in cascad-
ing loops and a state estimation method based on Extended Kalman Filtering. The
Extended Kalman Filter is proposed for estimating the nonmeasurable elements of
the DFIG state vector (such as the rotation speed and the magnetic flux) through the
processing of measurements of the rotor’s angle and of the rotor currents. The effi-
ciency of the considered state estimation-based nonlinear control scheme is evaluated
through simulation experiments.
Finally, a control method for distributed interconnected power generation units is
developed. The power system comprises Permanent Magnet Synchronous Genera-
tors (PMSGs) which are connected to each other through transformers and tie-lines.
A Derivative-free nonlinear Kalman Filtering approach is introduced aiming at imple-
menting sensorless control of the distributed power generators. As in the case of
stand-alone functioning of PMSG, in the proposed Derivative-free Kalman Filtering
method the distributed generators’ model is first subjected to a linearization transfor-
mation based on differential flatness theory and next state estimation is performed
by applying the standard Kalman Filter recursion to the linearized model. As in the
stand-alone PMSG case, the proposed approach provides estimates of the state vector
of the permanent magnet synchronous generator without the need for derivatives and
Jacobians calculation. Moreover, by bringing the Derivative-free nonlinear Kalman
Filter in the form of a disturbance observer it is shown that it is possible to estimate
at the same time the nonmeasurable elements of each generator’s state vector, the
unknown input power (torque), and the disturbance terms induced by interarea oscil-
lations. The efficient real-time estimation of the aggregate disturbance that affects
each local generator makes possible to introduce a counter-disturbance control term,
thus maintaining the power system on its nominal operating conditions.

7.2 State Estimation-Based Control of PMSGs

7.2.1 The PMSG Control Problem

State estimation of nonlinear electric power generators with the use of filters is a
significant topic in the area of mechatronics since it can provide improved meth-
ods for sensorless control and fault diagnosis of such electromechanical systems.
In the design of nonlinear controllers for power generators it is important to mea-
sure all state variables needed to generate a feedback control law. In case of state
variables for which direct measurement is technically difficult or costly, estima-
tion can be performed with nonlinear filtering methods [15, 119, 220, 304, 327].
7.2 State Estimation-Based Control of PMSGs 339

This section proposes Derivative-free nonlinear Kalman Filtering for sensorless con-
trol of Permanent Magnet Synchronous Generators.
For nonlinear electric power generators, and under Gaussian noise, the Extended
Kalman Filter (EKF) is frequently applied for estimating the nonmeasurable state
variables through processing of input and output sequences [31, 192, 229, 372]. The
Extended Kalman Filter is based on linearization of the dynamical system using
Taylor series expansion [405, 408, 412, 422, 573]. Although EKF is efficient in
several estimation problems, it is characterized by cumulative errors due to local
linearization assumption and this may affect the accuracy of the state estimation or
even risk the stability of the observer-based control loop.
It is also known that one can attempt transformation of the permanent magnet
synchronous generator model into the canonical (Brunovsky) form through appli-
cation of the Lie-algebra theory [327, 328]. By using such differential geometric
methods it is possible to arrive at a description of the system in the linear canonical
form if the relative degree of the system is equal to the order of the system. After
transformation to canonical form, state estimation with the use of a linear system
is possible. However, this linearization procedure requires the computation of Lie
derivatives (partial derivatives on the vector fields describing the system dynamics),
which can be a cumbersome computation procedure.
Aiming at finding more efficient state estimation methods for the model of the
permanent magnet synchronous generator (with reference to the Lie algebra-based
state estimator design), in this chapter the Derivative-free nonlinear Kalman Filter
is used. In the proposed Derivative-free Kalman Filtering method the system is first
subject to a linearization transformation based on the differential flatness theory and
the next state estimation is performed by applying the standard Kalman Filter recur-
sion to the linearized model. Unlike the Lie algebra-based estimator design method,
the proposed approach provides estimates of the state vector of the permanent magnet
synchronous generator without the need for derivatives and Jacobians calculation.
By avoiding linearization approximations, the proposed filtering method improves
the accuracy of estimation of the system state variables, and results in smooth control
signal variations and in minimization of the tracking error of the PMSG control loop
[424, 434, 437].
This section is concerned with proving differential flatness of the permanent mag-
net synchronous generator model and its resulting description in the Brunovksy
(canonical) form [340, 444]. At a second stage and using the linearized model,
Kalman Filtering is proposed for estimating the nondirectly measurable elements of
the state vector of the power generator. To enable efficient operation of the power gen-
erator under variable mechanical input power (torque) the Kalman Filter is extended
toward a disturbances estimator. This enables to estimate simultaneously both the
state vector of the generator and the unknown mechanical input power (torque).
Finally, state estimation-based control is applied to assure that the rotation speed of
the generator tracks a predefined reference setpoint. The aggregate control input is
340 7 Differential Flatness Theory and Electric Power Generation

generated by including in the state-feedback control law a term that is based on the
estimation of the unknown input power and of its derivatives and which compensates
for the effects of variation of this input power.

7.2.2 Dynamic Model of the Permanent Magnet Synchronous

Generator

It is considered that the third order model of the Synchronous Generator (which also
includes as a special case the PMSG) is connected to the power grid as shown in
Fig. 7.1, thus forming the model of a Single Machine Infinite Bus (SMIB) system.
The PMSG mechanical dynamics can be represented as follows:

δ̇ = ω
ω0 (7.1)
ω̇ = − 2J
D
(ω − ω0 ) + 2J (Pm − Pe )

where δ is the turn angle of the generator’s rotor, ω is the rotation speed of the
rotor with respect to synchronous reference, ω0 is the synchronous speed of the
generator, J is the moment of inertia of the rotor, Pm is the mechanical input
torque to the generator which is associated with the mechanical input power, D is the
damping constant of the generator, and Te is the electrical torque associated with the
generated active power. Moreover, the following variables are defined: Δδ = δ − δ0
and Δω = ω − ω0 with ω0 denoting the synchronous speed.
The generator’s electric dynamics is described by [15, 119, 220, 304, 327]:

 1
Ė q = (E f − E q ) (7.2)
Tdo

where E q is the quadrature-axis transient voltage of the generator (which stands for
a non-measurable magnetic flux variable), E q is the quadrature axis voltage of the
generator, Tdo is the direct axis open-circuit transient time constant of the generator,
and E f is the equivalent voltage in the excitation coil.

Fig. 7.1 PMSG connected

to the power grid according
to the Single Machine
Infinite Bus (SMIB) model
7.2 State Estimation-Based Control of PMSGs 341

The algebraic equations of the synchronous generator are given as

xdΣ   Vs
Eq =  E q − (xd − xd )  cos(Δδ)
xd xd
Σ Σ
Vs
Iq =  sin(Δδ)
xd
Σ

Eq Vs
Id =  −  cos(Δδ)
xd xd
Σ Σ
Vs E q
 (7.3)
Pe =  sin(Δδ)
xd
Σ

Vs E q V2
Qe = cos(Δδ) − xds

xd Σ
 Σ
  
Vt = (E q − X d Id ) + (X d Iq )2
2

 
where xdΣ = xd + x T + x L , xdΣ = xd + x T + x L , xd is the direct-axis synchronous
reactance, x T is the reactance of the transformer, x L is the reactance of the transmis-
sion line, Id and Iq are direct and quadrature axis currents of the generator, Vs is the
infinite bus voltage, Q e is the generator reactive power delivered to the infinite bus,
and Vt is the terminal voltage of the generator.
Substituting the electrical equations of the PMSG given in Eq. (7.3) into the equa-
tions of electrical and mechanical dynamics of the rotor given in Eqs. (7.1) and (7.2),
respectively, the complete model of the Single Machine Infinite Bus model is obtained

δ̇ = ω − ω0

Pm Vs E q
ω̇ = − 2J
D
(ω − ω0 ) + ω0 2J − ω0 2J
1
 sin(Δδ)
xd
Σ
(7.4)

  1 xd −xd
Ė q = − 1 Eq + Tdo x  Vs cos(Δδ) + 1
Tdo Ef
Td dΣ


 x
where Td = xddΣ Tdo is the time constant of the field winding.
Σ
The previously analyzed single-machine infinite-bus model of the PMSG is
described by a nonlinear state-space model of the form

ẋ = f (x) + g(x)u (7.5)

where the state vector x is defined as

 
 T
x = Δδ Δω E q (7.6)
342 7 Differential Flatness Theory and Electric Power Generation

The vector fields f (x) and g(x) are defined as

⎛ ⎞
ω − ω0

⎜ D ⎟ Vs E q
⎜ Pm
sin(Δδ)⎟
f (x) = ⎜− 2J (ω − ω0 ) + ω0 2H − ω0 2J
1

⎟ xdσ (7.7)
⎝  ⎠
1  1 xd −xd
−  E q + Tdo  Vs cos(Δδ)
Td xdσ

 T
1
g(x) = 0 0 (7.8)
Tdo

with control input u = E f the field voltage (equivalent voltage in the excitation coil)
and measurable output the rotation speed of the rotor

y = h(x) = δ − δ0 (7.9)

7.2.3 Lie Algebra-Based Design of State Estimators

for the PMSG

7.2.3.1 Relative Degree for Nonlinear Systems

The nonlinear model of the PMSG given in Eq. (7.4) is in an affine in-the-input
form, i.e.,

ẋ(t) = f (x) + g(x)u

(7.10)
y(t) = h(x)

where x ∈ R n is the state vector, u ∈ R is the control vector, y ∈ R is the output

vector, f (x) and g(x) vector fields that belong both in R n and h(x) are the scalar
functions of x. It will be shown that a state estimator for the nonlinear model of the
PMSG can be derived using Lie algebra, which introduces a change of coordinates
(diffeomorphism) that enables to write a nonlinear system in an equivalent linear
form. Next, the following definitions from Lie algebra theory are recalled [238]:
Lie Derivative (already given in Chap. 1): For a given differentiable scalar function
h(x) of x = [x1 , x2 , x3 , . . . xn ]T and a vector field f (x) = [ f 1 , f 2 , f 3 , . . . , f n ]T , the
Lie derivative of function h(x) along the vector field f (x), is a new scalar function
defined by L f h(x) is obtained as follows:

∂h(x) n ∂h(x)
L f h(x) = f (x) = f i (x) (7.11)
∂x i=1 ∂ x i
7.2 State Estimation-Based Control of PMSGs 343

Relative degree (already given in Chap. 1): If the Lie derivative of the function
L rf−1 h(x) along vector field g(x) is not equal to zero in a neighborhood Ω, i.e.,

L g L rf−1 h(x)=0 (7.12)

then it is said that the system has relative degree r in Ω.

The relative degree of the system is a parameter to take into account the design
of controllers or observers for nonlinear dynamical systems. If the relative degree of
a system is equal to its order n, then the system is exactly linearizable. If r < n then
the system is partially linearizable.

7.2.3.2 Nonlinear Observer Design for Exactly Linearizable Systems

Under the condition that the relative degree of the system is r = n one has [238]:

L g L 1−1 2−1 n−2

f h(x) = L g L f h(x) = · · · = L g L f h(x) = 0
n−1 (7.13)
L g L f h(x) = 0

Next, a change in coordinates is performed as follows:

z 1 = y = h(x) = L 1−1
f h(x) (7.14)

Moreover, it holds that

∂h(x)
ż 1 = ẋ (7.15)
∂x
Substituting the system’s dynamics into Eq. (7.15) one gets

ż 1 = ∂h(x) ∂h(x)
∂ x f (x) + ∂ x g(x) · u (7.16)
= L 2−1 1−1
f h(x) + L g L f h(x)u

According to the condition about the system’s relative degree it holds that L g L 1−1
f
h(x) = 0 and from Eq. (7.16) one gets

ż 1 = L f h(x) = z 2 (7.17)

In a similar way,
ż 2 = L 2f h(x) = z 3
···
(7.18)
···
ż n−1 = L n−1f h(x) = z n
344 7 Differential Flatness Theory and Electric Power Generation

Again, using the property of the system’s relative degree one has L g L n−1
f h(x) =0
one obtains
ż n = L nf h(x) + L g L n−1
f h(x)u = α(x) + b(x)u = v (7.19)

where α(x) = L nf h(x), b(x) = L g L n−1

f h(x) and v = α(x) + b(x)u. After this
change of coordinates the system can be written as

ż = Az + Bu (7.20)

where ⎛ ⎞
h(x)
⎜ L f h(x) ⎟
⎜ ⎟
z = φ(x) = ⎜
⎜ ··· ⎟
⎟ (7.21)
⎝ ··· ⎠
L n−1
f h(x)

while matrices A, B appearing in the previous state-space equation are given as

⎛ ⎞
0 1 0 ··· 0
⎜0 0 1 · · · 0 ⎟
⎜ ⎟
⎜ ⎟
A = ⎜ ... ... ... · · · ... ⎟
⎜ ⎟ (7.22)
⎝0 0 0 · · · 1 ⎠
0 0 0 ··· 0
 T
B = 0 0 ··· 0 1

The state observer for the transformed system of Eq. (7.20) is

ẑ˙ = Aẑ + Bv + K (y − h(x̂)) or

˙ẑ = Aẑ + B[α(x̂) + b(x̂)u] + K (y − h(x̂)) or (7.23)
ẑ˙ = Aẑ + B[α(φ −1 (ẑ)) + b(φ −1 (ẑ))u] + K (y − h(φ −1 (ẑ)))

It is also possible to express the state observer using a nonlinear model, i.e.,

x̂˙ = f (x̂) + g(x̂)u + L(y − h(x̂)) (7.24)

It will be shown that the nonlinear observer’s gain L is now given as

L = (Jφ (x̂))−1 K (7.25)

where matrix Jφ (x̂) is the Jacobian of the new coordinates z 1 , . . . , z n which is

obtained after the nonlinear change of coordinates and K is the observer’s gain
computed for the linearized equivalent of the system of Eq. (7.23). The observer gain
7.2 State Estimation-Based Control of PMSGs 345

K for the linearized system can be obtained through the Kalman Filter recursion.
One has that ⎛ ⎞
dh(x̂)
⎜ d L f h(x̂) ⎟
d ẑ = ⎜
⎝ ···
⎟ = Jφ (x̂)d x̂
⎠ (7.26)
d L n−1
f h( x̂)

or equivalently
ẑ˙ = Jφ (x̂)x̂˙ (7.27)

It holds that the Jacobian matrix of φ(x̂) with respect to x̂ can be written as

∂φ(x̂)
Jφ (x̂) = (7.28)
∂ x̂
Using the state observer dynamics described in Eq. (7.24) one has

∂φ(x̂) ˙ ∂φ(x̂) ∂φ(x̂) ∂φ(x̂)

x̂ = f (x̂) + g(x̂)u + L(x̂)(y − h(x̂)) (7.29)
∂ x̂ ∂ x̂ ∂ x̂ ∂ x̂

Considering that for the first row of the Jacobian matrix it holds that φ(x̂) = h(x̂),
one has
∂h(x̂) ˙ ∂h(x̂) ∂h(x̂)
x̂ = f (x̂) + g(x̂)u + K (y − h(x̂)) (7.30)
∂ x̂ ∂ x̂ ∂ x̂
or equivalently
ẑ˙ 1 = L f h(x̂) + L g h(x̂)u + K 1 (y − h(x̂)) (7.31)

Moreover, it holds that L g h(x̂) = 0 and L f h(x̂) = ẑ 2 and thus one obtains

ẑ˙ 1 = ẑ 2 + K 1 (y − h(x̂)) (7.32)

In a similar manner, one has

ẑ˙ 2 = ẑ 3 + K 2 (y − h(x̂))
··· (7.33)
ẑ˙ n−1 = ẑ n + K n−1 (y − h(x̂))

Additionally, using L g L n−1

f  =0 one can finally write

ẑ˙ n = α(φ −1 (ẑ)) + b(φ −1 (ẑ))u + K n (y − h(x̂)) (7.34)

where α(φ −1 (ẑ)) = L n−1 n−1 −1 −1 n−1

f h( x̂) = L f h(φ (ẑ)), and b(φ (ẑ)) = L g L f h( x̂) =
L g L n−1 −1
f h(φ (ẑ)). Using the previous notation one obtains the formulation of the
nonlinear estimator’s gain L as a function of the observation gain K for the linearized
346 7 Differential Flatness Theory and Electric Power Generation

equivalent of the system described in Eq. (7.23). Thus one finally arrives at Eq. (7.25),
that is,
(Jφ (x̂))L = K ⇒L = (Jφ (x̂))−1 K (7.35)

7.2.3.3 Linearization of PMSG Dynamics Using Lie-Algebra

The nonlinear dynamical model of the PMSG was described in Eqs. (7.5)–(7.9). The
linearization procedure with the use of Lie algebra gives:

z 1 = L 0f h(x) = h(x) = Δδ (7.36)

⎛ ⎞
  f1
∂h ∂h ∂h ⎝
z 2 = L 1f h(x) = ∂ x1 ∂ x2 ∂ x3 · f 2⎠ ⇒
(7.37)
f3
z 2 = x2 = Δω
⎛ ⎞
  f1
∂z 2 ∂z 2 ∂z 2
z 3 = L 2f h(x) = ∂ x1 ∂ x2 ∂ x3 · ⎝ f2 ⎠ ⇒
f3 (7.38)

Pm ω0 Vs E q
z 3 = f 2 = − 2J
D
(ω − ω0 ) + ω0 2J − 2J x  sin(Δδ)
dΣ

Moreover, according to the previous analysis it holds that

ż 3 = L 3f h(x) + L g L 2f h(x)u (7.39)

where
⎛ ⎞
  f1
∂z 3 ∂z 3 ∂z 3 ⎝ f2 ⎠ ⇒
L 3f h(x) = ∂ x1 ∂ x2 ∂ x3 ·
f3
ω0 Vs
L 3f h(x) = (− 2J  x3 cos(x1 )ẋ1 )−
xdΣ

D −D Pm ω0 Vs 1 xd −xd
− 2J ( 2J x2 + ω0 2J − 2J x  x 3 sin(x 1 ))( T  x 3 + Tdo x 
1
cos(x1 )) ⇒
dΣ d dΣ
D Pm Vs ω0 Vs 1
L 3f h(x) = ( 2J ) x˙1 − ω0 2J 2J + (2J )2 ω0  x3 sin(x1 ) + 2J 
D 2 D
 x 3 sin(x 1 )−
x dΣ x T dΣ d

ω0 Vs 1 xd −xd ω0 Vs
− 2J   Vs cos(x1 )sin(x1 ) − 2H x  x 3 cos(x 1 )x 2
xdΣ Tdo xdΣ dΣ
(7.40)
7.2 State Estimation-Based Control of PMSGs 347

Finally, in an equivalent manner one obtains

⎛ ⎞
  g1
ω0 1 Vs
∂z 3 ∂z 3 ∂z 3
L g L 2f h(x) = ∂ x1 ∂ x2 ∂ x3 · ⎝g2 ⎠ ⇒ L g L 2f h(x) = −  sin(x1 )
g3 2J Tdo xdΣ
(7.41)

Defining the control input for the linearized system v = L 3f h(x) + L g L 2f h(x)u, the
state-space model can be written in the following linear canonical (Brunovsky) form:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (7.42)
ż 3 000 z3 1

7.2.4 Differential Flatness of the PMSG

It will be shown that the dynamic model of the PMSG is a differentially flat one,
i.e., it holds that all state variables and its control inputs can be written as functions
of the flat output and its derivatives. Moreover, it will be shown that by expressing
the elements of the state vector as functions of the flat output and its derivatives one
obtains a transformation of the PMSG model into the linear canonical (Brunovsky)
form.

Defining the state vector of the PMSG as x = [Δδ, Δω, E q ]T one has that

ẋ1 = x2
Pm ω0 Vs
ẋ2 = − 2J
D
x2 + ω0 2J − 2J  x 3 sin(x 1 )

xdΣ (7.43)
1 xd −xd
ẋ3 = − 1
 x3 + Tdo x  Vs cos(x1 ) + 1
Tdo u
Td dΣ

The flat output is chosen to be y = x1 . Therefore, it holds that

x1 = y (7.44)

x2 = ẏ (7.45)

while from the second row of Eq. (7.43) one has

D Pm ω0 Vs x3
ÿ = − ẏ + ω0 −  sin(y) (7.46)
2J 2J 2J xdΣ
348 7 Differential Flatness Theory and Electric Power Generation

Thus, for x1 = ±nπ (where n = 0, 1, 2, . . .) one obtains

ω0 P2Jm − ÿ− 2J
D
ẏ
x3 = ω0 Vs , or
2J  sin(y) (7.47)
xdΣ
x3 = f a (y, ẏ, ÿ)

From the third row of Eq. (7.43) one has


1 xd −xd
u = Tdo [ẋ3 + 1
 x3 T  Vs cos(x1 )], or
Td do x
dΣ (7.48)
u = f b (y, ẏ, ÿ)

Therefore, all state variables and the control input of the PMSG can be written as
functions of the flat output and its derivatives, and the PMSG model is a differentially
flat one.
Next, the following change of variables is performed: y1 = y, y2 = ẏ, y3 = ÿ. It
also holds that

ẏ3 = y (3) = ẍ2 ⇒

  (7.49)
ω0 Vs ω0 Vs
y (3) = − 2J
D
ẋ2 − 2H xdΣ ẋ 3 sin(x 1 ) − 2J xdΣ x 3 cos(x 1 ) ẋ 1

By substituting ẋ2 and ẋ3 from the second and third rows of Eq. (7.43), respectively,
and after intermediate operations one obtains

y (3) = f c (y, ẏ, ÿ) + gc (y, ẏ, ÿ)u (7.50)

where
D D Pm D Vs ω0 Vs 1
f c (y, ẏ, ÿ) = ( ) ẏ − ω0 + ω0  x3 sin( ẏ) +   x 3 sin(y)−
2J 2 2J 2J (2J ) 2
xdΣ 2J xdΣ Td

ω0 Vs 1 xd − xd ω0 Vs
−   Vs cos(y)sin(y) −  x3 cos(y) ẏ (7.51)
2J xdΣ Tdo xdΣ 2J xdΣ

ω0 1 Vs
gc (y, ẏ, ÿ) = − 2J Tdo x  sin(y) (7.52)
dΣ

Thus, the system can be written in the following linear canonical (Brunovsky) form:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (7.53)
ẏ3 000 y3 1

while the new control input for the linearized system is v = f c (y, ẏ, ÿ) +
gc (y, ẏ, ÿ)u. It can be noticed that the linearized equivalent of the system, which is
obtained after applying differential flatness theory, is the same as that obtained after
applying the Lie algebra-based method.
7.2 State Estimation-Based Control of PMSGs 349

The controller design for the linearized system described in Eq. (7.53) is carried
out using common pole placement methods. Since the overall system dynamics is
described by the differential equation

y (3) = v (7.54)

a suitable feedback control law that makes the flat output track a desirable setpoint
yd (t) is given by

(3)
v = yd (t) − k1 ( ÿ(t) − ÿd (t)) − k2 ( ẏ(t) − ẏd (t)) − k3 (y(t) − yd (t)) (7.55)

The control input that is finally applied to the PMSG is given as

u(t) = gc−1 (t)[v(t) − f c (t)] (7.56)

The control law of Eq. (7.55) results in the closed-loop dynamics

e(3) (t) + k1 ë(t) + k2 ė(t) + k3 e(t) = 0 (7.57)

By selecting the feedback control gains k1 , k2 and k3 such that the associated charac-
teristic polynomial p(s) = s 3 + k1 s 2 + k2 s + k3 is a Hurwitz one, i.e., it has poles in
the left complex semi-plane, one has that the tracking error converges asymptotically
to zero
lim t→∞ e(t) = 0⇒lim t→∞ y(t) = yd (t) (7.58)

Since the flat output converges to the desirable setpoint and since all state variables
x1 , x2 and x3 are expressed as functions of the flat output and its derivatives one has
that the state variables also converge to the desirable setpoints and therefore efficient
control of the PMSG is achieved.
Moreover, for the linearized equivalent of the system one can perform state esti-
mation with the use of the standard Kalman Filter recursion and can compute also

the state vector elements which are not directly measurable (i.e., ω and E q ).

7.2.5 Estimation of PMSG Disturbance Input with Kalman

Filtering

7.2.5.1 State Estimation with the Derivative-Free Nonlinear Kalman

Filter

It was shown that using differential flatness theory the initial nonlinear model of the
PMSG can be written in canonical form
350 7 Differential Flatness Theory and Electric Power Generation
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (7.59)
ÿ3 000 ẏ3 1

Thus, one has a linear model of the form

ẏ f = A f y f + B f v
(7.60)
z f = Cf yf

where y f = [y1 , y2 , y3 ]T and matrices A f , B f , C f are in the canonical form

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
010 0 1
A f = ⎝0 0 1⎠ B f = ⎝0⎠ C Tf = ⎝0⎠ (7.61)
000 1 0

where the measurable variable y1 = δ is associated with the turn angle of the rotor.
For the previous model, and after carrying out discretization of matrices A f , B f , and
C f with common discretization methods one can perform linear Kalman filtering
using Eqs. (4.5) and (4.6). This is Derivative-free nonlinear Kalman filtering for the
model of the generator which, unlike EKF, is performed without the need to compute
Jacobian matrices and does not introduce numerical errors due to approximative
linearization with Taylor series expansion [429, 434, 437].

7.2.5.2 Kalman Filter-Based Estimation of Disturbances

Up to now the mechanical input torque of the generator, which has been denoted
as Pm , was considered to be constant or piecewise constant while it has been also
considered that it is possible to measure it. Now it is assumed that the mechanical
input torque varies in time, and in that case the aggregate disturbance input exerted
on the generator’s model is

D ω0
Tm = −ω0 Pm + Ṗm (7.62)
(2J )2 2J

It is also assumed that the dynamics of the disturbance term Tm is defined by its
nth order derivative Tm(n) . Considering now that after expressing the system’s state
variables as functions of the flat outputs and their derivatives the PMSG’s dynamics
is given as
y (3) = f c (y, ẏ, ÿ) + gc (y, ẏ, ÿ)u + 2H1
Ṗm or
(3) ω0
y = v − ω0 D
Pm + Ṗm or (7.63)
(2J )2 2J
(3)
y = v − Tm
7.2 State Estimation-Based Control of PMSGs 351

where v = f c (y, ẏ, ÿ) + gc (y, ẏ, ÿ)v and

Pm ω0 Vs 1
f c (y, ẏ, ÿ) = ( 2JD2 ) ẏ − ω0 2J
D
2J + D
ω Vs
(2J )2 0 x 
x3 sin( ẏ) + 2J x  T  x 3 sin(y)−
dΣ dΣ d

ω0 Vs 1 xd −xd ω0 Vs
− 2J   Vs cos(y)sin(y) − 2H x  x 3 cos(y) ẏ
xdΣ Tdo xdΣ dΣ
(7.64)
D ω0 Vs
gc (y, ẏ, ÿ) = −  sin(y) (7.65)
2M Tdo xdΣ
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0  
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v − ω0 D 2 Pm + ω0
(2J ) 2J Ṗm (7.66)
ÿ3 000 y3 1

Next, the state vector of the model of Eq. (7.66) is extended to include as additional
state variables the disturbance input Tm . Then, in the new state-space description one
ω0
has z 1 = y1 , z 2 = y2 , z 3 = y3 , z 4 = Tm = −ω0 (2JD)2 Pm + 2J Ṗm , z 5 = Ṫm , and
z 6 = T̈m . Without loss of generality, the disturbance input dynamics is assumed to
(3)
be described by its third order derivative ż 6 = Tm . Using the previous definition of
state variables one has the matrix equations

ż = Ã · z + B̃ · ṽ (7.67)

where the control input is

 T
ṽ = v Tm(3) (7.68)

with ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 00 1
⎜0 0 1 0 0 0⎟ ⎜0 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0⎟ ⎜ ⎟ ⎜ ⎟
à = ⎜ ⎟ B̃ = ⎜1 0⎟ C̃ T = ⎜0⎟ (7.69)
⎜0 0 0 0 1 0⎟ ⎜0 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 1⎠ ⎝0 0 ⎠ ⎝0⎠
0 0 0 0 0 0 01 0

where the measurable state variable is z 1 . Since the dynamics of the disturbance input
are taken to be unknown in the design of the associated disturbances’ estimator one
has the following dynamics:

ż o = Ão · z + B̃o · ṽ + K (Co z − Co ẑ) (7.70)

352 7 Differential Flatness Theory and Electric Power Generation

where K ∈ R 6×1 is the state estimator’s gain and

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 00 00 1
⎜0 0 1 0 0 0⎟ ⎜0 0⎟ ⎜0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0⎟ ⎜ ⎟ ⎜ ⎟
Ão = ⎜ ⎟ B̃o = ⎜1 0⎟ C̃ T = ⎜0⎟ (7.71)
⎜0 0 0 0 ⎟
1 0⎟ ⎜ ⎟ o ⎜0 ⎟
⎜ ⎜0 0⎟ ⎜ ⎟
⎝0 0 0 0 0 1⎠ ⎝0 0⎠ ⎝0 ⎠
0 0 0 0 00 00 0

It can be confirmed that the disturbance observer model of the PMSG defined in
Eq. (7.67) is observable. Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents
of matrices Ão , B̃o and C̃o respectively, a Derivative-free nonlinear Kalman Filter
can be designed for the aforementioned representation of the system dynamics. The
associated Kalman Filter-based disturbance estimator is given as [438]
measurement update:

K (k) = P − (k)C̃dT [C̃d · P − (k)C̃dT + R]−1

x̂(k) = x̂ − (k) + K (k)[z(k) − C̃d x̂ − (k)] (7.72)
P(k) = P − (k) − K (k)C̃d P − (k)

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

(7.73)
x̂ − (k + 1) = Ãd (k)x̂(k) + B̃d (k)ṽ(k)

To compensate for the disturbance terms it suffices to use in the control loop the
modified control input, which actually removes the effects of the external disturbance
variable Tm .
v∗ = v − T̂m or v∗ = v − ẑ 4 (7.74)

7.2.6 Simulation Experiments

To evaluate the performance of the proposed nonlinear control scheme, which uses
Kalman Filtering to estimate the nonmeasurable state vector elements of the PMSG
and the external disturbances, simulation experiments have been carried out. Differ-
ent rotation speed setpoints have been assumed. Moreover, different input torques
(mechanical input power profiles) have been considered to affect the PMSG dynamic
model (Single Machine Infinite Bus model). The control loop used for the PMSG is
shown in Fig. 7.2.
First, the case of measurable piecewise constant mechanical input power (torque)
was examined. It was assumed that the input torque could be measured and could be
directly used in the controller’s design. The Derivative-free nonlinear Kalman Filter
7.2 State Estimation-Based Control of PMSGs 353

Fig. 7.2 Control loop for the PSMG comprising a flatness-based nonlinear controller and a Kalman
Filter-based disturbances estimator

(a) 0.11 (b) 1

0.1
0.8
0.09
0.08 0.6

0.07
0.4
q
Δω

0.06
E

0.2
0.05
0.04 0
0.03
−0.2
0.02
0.01 −0.4
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.3 Sensorless control of the PMSG under measurable mechanical input torque in case of
speed reference setpoint 1: a convergence of the real and estimated values of the angular speed

difference Δω, b convergence of the real and estimated values of the quadrature axis voltage E q


enabled the estimation of specific elements of the state vector, such as ω and E q .
The associated results of the tracking performance of the control loop are shown in
Figs. 7.3 and 7.4. The desirable setpoint is denoted by a red line, the real value of the
state variable is denoted by a blue line, while the estimated value of the state variable
is printed in green. The units of the PMSG state variables have been expressed in
p.u. (per unit system). One can observe that the proposed sensorless control scheme
for the PMSG succeeds fast and accurate convergence to the desirable setpoints.
354 7 Differential Flatness Theory and Electric Power Generation

(a) 0.12 (b) 0.4

0.1 0.2

0
0.08
−0.2

E /q
Δω

0.06
−0.4
0.04
−0.6

0.02 −0.8

0 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.4 Sensorless control of the PMSG under measurable mechanical input torque in case of
speed reference setpoint 2: a convergence of the real and estimated values of the angular speed

difference Δω, b convergence of the real and estimated values of the quadrature axis voltage E q

Next, the case of PMSG operation under unknown input power (torque) was exam-
ined. The input power was considered to be a disturbance input to the PMSG model
and it was assumed that its change in time was defined by the third derivative of the
associated variable, i.e., Tm(3) where Tm = −ω0 (2JD)2 Pm + 2J ω0
Ṗm . The disturbance
dynamics was completely unknown to the controller and its identification was per-
formed in real time by the disturbance estimator. It is shown that the Derivative-free
nonlinear Kalman Filter redesigned as a disturbance observer is capable of estimat-

ing simultaneously the nonmeasurable state variables of the generator (ω and E q ), as
well as the unknown disturbance input Tm . A first set of results is concerned with the
tracking performance of the control loop in case of piecewise constant disturbance
input, as shown in Figs. 7.5 and 7.6. The estimation of the piecewise constant distur-
bance input is shown in Fig. 7.7. Additionally, results of the tracking performance of
the control loop in case of time-varying disturbance input are shown in Figs. 7.8 and
7.9, while the estimation of the time-varying disturbance input is shown in Fig. 7.10.
The units of the PMSG state variables have been expressed again in p.u. (per unit
system).
The simulation experiments confirm that the proposed state estimation-based con-
trol scheme not only enables implementation of PMSG control through measurement
of a small number of variables (e.g., of only the rotor’s turn angle) but also improves
the robustness of the PMSG control loop in case of varying speed setpoints and
varying mechanical input torque.
In conclusion, there are two particular issues in the design of the PMSG controller
(i) there are certain elements in the generator’s state vector which are not directly
measurable, (ii) there may be variations in the mechanical input power (torque) and
it may also be impractical or costly to measure this input power. To address (i) a
new nonlinear Kalman Filtering approach is introduced. The method is the so-called
7.2 State Estimation-Based Control of PMSGs 355

(a) 0.15 (b) 0.04

0.02
0.1 0
−0.02
0.05 −0.04

q
Δω

E/
−0.06
0 −0.08
−0.1
−0.05 −0.12
−0.14
−0.1 −0.16
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.5 Sensorless control of the PMSG under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 1: a convergence of the real and estimated values of
the angular speed difference Δω, b convergence of the real and estimated values of the quadrature

axis voltage E q

(a) 0.25 (b) 0.05

0
0.2
−0.05
0.15
−0.1
0.1 −0.15
q
Δω

/
E

0.05 −0.2
−0.25
0
−0.3
−0.05
−0.35
−0.1 −0.4
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.6 Sensorless control of the PMSG under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 2: a convergence of the real and estimated values of
the angular speed difference Δω, b convergence of the real and estimated values of the quadrature

axis voltage E q

Derivative-free nonlinear Kalman Filtering and consists of implementation of the

standard Kalman Filter recursion on the linearized equivalent of the PMSG model,
which is obtained with the use of the differential flatness-based transformation. The
fast computation features of the Derivative-free nonlinear Kalman Filter enable the
estimation of nonmeasurable state vector elements in real-time, while the accuracy
of the provided state estimation is very satisfactory.
356 7 Differential Flatness Theory and Electric Power Generation

(a) 0.35 (b) 0.45

0.4
Estimation of input torque Tm

Estimation of input torque Tm

0.3
0.35
0.25
0.3
0.2 0.25

0.15 0.2

0.15
0.1
0.1
0.05
0.05

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.7 Estimation of the nonmeasurable (piecewise constant) input torque in sensorless control
of the PMSG through processing of rotor angle measurements: a in case of speed reference setpoint
1, b in case of speed reference setpoint 2

(a) 0.2 (b) 0.05

0.15
0
0.1

0.05 −0.05
q
Δω

E/

−0.05 −0.1

−0.1
−0.15
−0.15

−0.2 −0.2
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.8 Sensorless control of the PMSG under nonmeasurable (time varying) mechanical input
torque in case of speed reference setpoint 1: a convergence of the real and estimated values of the
angular speed difference Δω, b convergence of the real and estimated values of the quadrature axis

voltage E q

To address (ii) the Derivative-free nonlinear Kalman Filter is redesigned in the

form of a disturbance estimator. The state vector of the disturbance estimator contains
both the state vector elements of the linearized equivalent of the PMSG and new
state vector elements which stand for the unknown mechanical input (torque) and
its derivatives. With the use of the aforementioned disturbance observer one can
obtain estimates of both the nonmeasurable elements of the state vector (e.g., rotation
speed and quadrature-axis transient voltage of the rotor), as well as estimates of the
unknown disturbance input (which is due to external mechanical torque) and of its
time derivatives.
7.2 State Estimation-Based Control of PMSGs 357

(a) 0.16 (b) 0.05

0.14 0

0.12 −0.05

0.1 −0.1

q
Δω

E/
0.08 −0.15

0.06 −0.2

0.04 −0.25

0.02 −0.3

0 −0.35
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.9 Sensorless control of the PMSG under nonmeasurable (time varying) mechanical input
torque in case of speed reference setpoint 2: a convergence of the real and estimated values of the
angular speed difference Δω, b convergence of the real and estimated values of the quadrature axis

voltage E q

(a) 0.5 (b) 0.2

Estimation of input torque Tm

Estimation of input torque Tm

0.4 0.15

0.1
0.3
0.05
0.2
0
0.1
−0.05
0
−0.1
−0.1 −0.15

−0.2 −0.2
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.10 Estimation of the nonmeasurable (time varying) input torque in sensorless control of the
PMSG: a in case of speed reference setpoint 1, b in case of speed reference setpoint 2

To maintain the operating characteristics of the generator (e.g., rotation speed)

unchanged despite variations in the input power, the initial state feedback controller
of the PMSG has been complemented by additional elements based on the distur-
bance input estimates which actually compensate for the unknown input effects. The
considered disturbance estimator also makes possible to compensate for the effects
of parametric changes and modeling uncertainties in the PMSG dynamics. The effi-
ciency of the proposed state estimation-based (sensorless control) of the PMSG is
evaluated through simulation experiments.
358 7 Differential Flatness Theory and Electric Power Generation

7.3 State Estimation-Based Control of DFIGs

7.3.1 Overview

Next, it will be shown that differential flatness theory-based methods for filtering and
control can be also applied to Doubly-fed induction generators. These types of gener-
ators (DFIGs) have been widely used in variable-speed fixed frequency hydro-power
generation systems, wind-power generation systems, and turbine engine power gen-
eration systems [140, 304, 355, 575, 586]. Doubly-fed induction generators have
proved to be more efficient than squirrel-cage induction generator systems (SCIG)
and synchronous generator systems in terms of cost and losses of the associated
power electronics converters. DFIG systems can operate either in grid-connected
mode or in stand-alone mode [54, 155, 156, 171, 228, 416, 577]. Moreover, several
field-oriented control schemes have been proposed for both operation modes. Results
on the reliable connection of DFIGs to the electricity grid are presented in [385, 566,
587]. Additionally, to control electric power generators and the power electronics
that enable their connection to the grid, feedback linearization approaches have been
developed [28, 245]. In parallel, several results have been published on sensorless
control of DFIG [1, 63, 386, 442, 562, 582]. Taking into account that the installa-
tion and maintenance of sensors for measuring several parameters of the generator’s
state vector can be technically difficult or costly, the need for developing sensorless
control schemes for DFIG becomes apparent. In this section a novel sensorless con-
trol scheme is developed using flatness-based control theory and a state estimation
method based on Kalman Filtering.
Using the electric equations of the stator and rotor a dynamic model for the
doubly fed induction generator is derived. The doubly-fed induction generator is
analogous to the induction motor. In an induction motor the stator voltage plays
the role of an input variable, while the rotor voltage is a constant. In case of the
doubly-fed induction machine it is quite similar, but it is the other way round with a
dual analogy to hold between the stator and rotor parameters of the generator and the
motor. This means that the rotor voltage now acts as an input, while the stator voltage
is a constant parameter. The stator’s and rotor’s voltages, currents, and magnetic flux
are represented as vectors in a rotating orthogonal axis frame. The complete sixth
order model of the DFIG captures efficiently transients at both the stator and rotor
side.
This chapter is concerned with proving differential flatness of the model of the
doubly-fed induction generator and its resulting description in the Brunovksy (canon-
ical) form [340]. By defining flat outputs which are associated with the rotor’s angle
and with the magnetic flux of the stator, an equivalent DFIG description in the
Brunovksy (linear canonical) form is obtained. It is shown that for the linearized
DFIG’s model it is possible to design a feedback controller. At a second stage, a
novel Kalman Filtering method, the Derivative-free nonlinear Kalman Filter, is pro-
posed for estimating the state vector elements of the linearized system which are not
directly measurable. With the redesign of the proposed Kalman Filter as a disturbance
7.3 State Estimation-Based Control of DFIGs 359

observer, it becomes possible to estimate also disturbance terms affecting the DFIG
model and to use these terms in the feedback controller. By avoiding linearization
approximations, the proposed filtering method, improves the accuracy of estimation,
and results in smooth control signal variations and in minimization of the tracking
error of the associated control loop [434, 436, 437].

7.3.2 The Complete Sixth-Order Model of the Induction

Generator

The doubly-fed induction generator (DFIG) is not only the most widely used tech-
nology in wind turbines due to its good performance, but it is also essential in many
other fields such as hydro-power generation, pumped storage plants, and flywheel
energy storage systems. The DFIG model is derived from the voltage equations of
the stator and rotor. It is assumed that the stator and rotor windings are symmetrical
and symmetrically fed. Usually, the saturation of the inductances, iron losses, and
bearing friction is neglected. Moreover, the winding resistance is considered to be
constant.
This type of wound-rotor machine is connected to the grid by both the rotor
and stator sides. The DFIG stator can be directly connected to the electric power
grid while the rotor is interfaced through back-to-back converters (see Fig. 7.11).
By decoupling the power system’s electrical frequency and the rotor’s mechanical
frequency the converter allows a variable speed operation of the wind turbine.

Fig. 7.11 Configuration of a doubly-fed induction generator unit in the power grid
360 7 Differential Flatness Theory and Electric Power Generation

The doubly-fed induction generator is analogous to the induction motor. In an

induction motor the stator voltage plays the role of an input variable, while the rotor
voltage is a constant (it is usually zero). In case of the doubly-fed induction machine
it is very similar, but the other way round, with a dual analogy to hold between the
stator and rotor parameters of the generator and the motor. This means that the rotor
voltage now acts as an input, while the stator voltage depends on the voltage at the
bus to which the DFIG is connected and is a constant parameter [434, 442, 458].
A model of the doubly-fed induction generator is as follows:
Dynamic equations:
J ω̇ = Tm − K f ω − Te (7.75)

where J is the moment of inertia of the rotor, Tm is the externally applied mechanical
torque that makes the turbine rotate, Te is the electrical torque associated to the
generated active power, and finally the term k f ω expresses friction, with K f being
the friction coefficient. The wind generated mechanical torque is given by

1
Tm = ρπ R 3 Cq (λ, β)v2 (7.76)
2
where v is the wind’s speed [54]. Cq is a torque coefficient which depends on the
blade pitch angle β and the tip-speed ratio which is provided by λ = ωrv R , with ωr
being the rotor’s angular velocity, R is the rotor radius, and ρ is the air density.
Electrical equations: Using the Park transform the DFIG is described in the d − q
reference frame by the following set of equations:

dψsd
vsd = Rs i sd + − ωdq ψsq (7.77)
dt
dψsq
vsq = Rs i sq + + ωdq ψsd (7.78)
dt
dψrd
vrd = Rr ird + − ωr ψrq (7.79)
dt
dψrq
vrq = Rr irq + + ωr ψrd (7.80)
dt
where ωdq is the synchronous frequency, ωr is the rotation frequency of the rotor,
ψsd is the stator flux component along the d-axis, ψsq is the stator flux component
along the q-axis and equivalently ψrd is the rotor flux component along the d-axis,
while ψrq is the stator flux component along the q-axis (see Fig. 7.12).
Moreover, vsd and i sd are the stator’s voltage and current in the d reference, vsq
and i sq are the stator’s voltage and current in the q reference frame and equivalently
vrd and ird are the rotor’s voltage and current in the d reference frame, while vrq and
irq are the rotor’s voltage and current in the q reference frame.
7.3 State Estimation-Based Control of DFIGs 361

Fig. 7.12 The a − b stator reference frame and the d − q rotor reference frame of the induction
generator

As the d and q axes are magnetically decoupled the flux components are described
by the following equations:
ψsd = L s i sd + Mird (7.81)

ψsq = L s i sq + Mirq (7.82)

ψrd = L r ird + Mi sd (7.83)

ψrq = L r irq + Mi sq (7.84)

Moreover, the electromagnetic torque that is developed is given as

Te = η(i sq ψsd − i sd ψsq ) (7.85)

where η is a variable associated to the number of poles and to the mutual induc-
tance M.
In a compact form the doubly-fed induction generator can be described by the
following set of equations in the d − q reference frame that rotates at an arbitrary
speed denoted as ωdq [155]

dψsq 1 M
=− ψsq − ωdq ψsd + irq + vsq (7.86)
dt τs τs
362 7 Differential Flatness Theory and Electric Power Generation

dψsd 1 M
= ωdq ψsq − ψsd + ird + vsd (7.87)
dt τs τs

dirq β 1
= ψsq + βωr ψsd − γ2 irq − (ωdq − ωr )ird − βvsq + vr (7.88)
dt τs σ Lr q

dird β 1
= −βωr ψsq + ψsd + (ωdq − ωr )irq − γ2 ird − βvsd + vr (7.89)
dt τs σ Lr d

where ψsq , ψsd , irq , ird are the stator flux and the rotor currents, vsq , vsd , vrq , vrd are
the stator and rotor voltages, L s and L r are the stator and rotor inductances, ωr is
the rotor’s angular velocity, M is the magnetizing inductance. Moreover, denoting
as Rs and Rr the stator and rotor resistances, the following parameters are defined:
2 Ls
σ = 1 − LMr L s β = 1−σMσ τs = Rs (7.90)
τr = LRrr γ2 = ( 1−σ
σ τs +
1
σ τr )

The angle of the vectors that describe the magnetic flux ψs α and ψs b is first defined
for the stator, i.e.,
ψs
ρ = tan −1 ( b ) (7.91)
ψsa

The angle between the inertial reference frame and the rotating reference frame is
taken to be equal to ρ. 
ψsa ψsb
Moreover, it holds that cos(ρ) = ||ψ|| , sin(ρ) = ||ψ|| , and ||ψ|| = ψs2α + ψs2b .
Therefore, in the rotating frame d − q of the generator, and under the condition of
field orientation, there will be only one nonzero component of the magnetic flux ψsd ,
while the component of the flux along the q axis equals 0.
The dynamic model of the doubly-fed induction generator can be also written in
state-space equation form by defining the following state variables: x1 = θ , x2 = ωr ,
x3 = ψsd , x4 = ψsq , x5 = ird and x6 = irq . It holds that

ẋ1 = x2 (7.92)

Km Tm η
ẋ2 = − x2 − + (irq x3 − ird x4 ) (7.93)
J J J
1 M
ẋ3 = − x3 + ωdq x4 + x5 + vsd (7.94)
τs τs

1 M
ẋ4 = −ωdq x3 − x4 + x6 + vsq (7.95)
τs τs
7.3 State Estimation-Based Control of DFIGs 363

ẋ5 = −βx2 x4 + τβs x3 +

(7.96)
+(ωdq − x2 )x6 − γ2 x5 + σ 1L r vrd − βvsd

ẋ6 = τβs x4 + βx2 x3 −

(7.97)
−(ωdq − x2 )x5 − γ2 x6 + σ 1L r vrq − βvsq

In the above set of equations J is the moment of inertia of the rotor, Tm is the
externally applied mechanical torque that makes the turbine rotate, K m is the friction
coefficient, and η is a variable that is associated to the number of poles and to the
mutual inductance M. Variable η in turn determines the electrical torque Te which is
associated to rotor currents and stator magnetic flux. Equations (7.157)–(7.162) can
be written also in the form

ẋ = f (x) + ga (x)vrd + gb (x)vrq (7.98)

where x = [x1 , x2 , x3 , x4 , x5 , x6 ]T and

⎛ ⎞
x2
⎜ − KJm x2 − TJm + nJ (irq x3 − ird x4 ) ⎟
⎜ ⎟
⎜ − 1
x + ω x + M
x + v ⎟
⎜ τs 3 dq 4 τs 5 sd ⎟
f (x) = ⎜ −ωdq x3 − τs x4 + τs x6 + vsq
1 M ⎟
⎜ ⎟
⎜ β ⎟
⎝−βx2 x4 + τ x3 + (ωdq − x2 )x6 − γ2 x5 − βvsd ⎠ (7.99)
s
β
τs x 4 + βx 2 x 3 − (ωdq − x 2 )x 5 − γ2 x 6 − βvsq
 T
ga (x) = 0 0 0 0 σ 1L r 0
 T
gb (x) = 0 0 0 0 0 σ 1L r

The active and reactive power delivered by the DFIG stator are associated to the real
and imaginary parts of the power at the stator’s terminals, i.e.,

Ps = Re{Us Is∗ } = vsd i sd + vsq i sq (7.100)

Q s = I m{Us Is∗ } = vsd i sq − vsq i sd (7.101)

7.3.3 Input–Output Linearization of the DFIG Using Lie

Algebra

The following variables have been defined:

h 1 (x) = x1 = θ
(7.102)
h 2 (x) = x32 + x42 = ψs2d + ψs2q
364 7 Differential Flatness Theory and Electric Power Generation

Next, based on h 1 , h 2 the following transformed state variables are defined:

z 1 = h 1 (x) = θ (7.103)

z 2 = L f h 1 (x)⇒
(7.104)
z 2 = f 1 ⇒z 2 = x2 ⇒z 2 = ω

Similarly, one has

z 3 = L 2f h 1 (x) = L f z 2 ⇒
z 3 = f 2 ⇒z 3 = − KJm x2 − TJm + ηJ (irq x3 − irq x4 ) ⇒ (7.105)
z 3 = − KJm x2 − TJm + ηJ (x6 x3 − x5 x4 )

Moreover, one has

ż 3 = L 3f h 1 (x) + (L ga L 2f h 1 (x))u 1 + (L gb L 2f h 2 (x))u 2 (7.106)

It holds that
L 3f h 1 (x) = L f z 3 (7.107)

L 3f h 1 (x) = − KJm [− KJm x2 − TJm + ηJ (x6 x3 − x5 x4 )]

+ ηJ x6 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]
− ηJ x5 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ] (7.108)
− J x4 [−βx2 x4 + τβs x3 + (ωdq − x2 )x6 − γ2 x5 − βvsd ]
η

+ ηJ x3 [ τβs x4 + βx2 x3 + (ωdq − x2 )x5 − γ2 x6 − βvsq ]

Equivalently, one has

L ga (L 2f h 1 (x)) = L ga z 3 ⇒
(7.109)
L ga (L 2f h 1 (x)) = − ηJ σ 1L r x4

and similarly,
L gb (L 2f h 1 (x)) = L gb z 3 ⇒
(7.110)
L gb (L 2f h 1 (x)) = ηJ σ 1L r x3

For the transformed state variable z 4 one has

z 4 = h 2 (x) = ψs2d + ψs2q = x32 + x42 (7.111)

and
z 5 = L f h 2 (x)⇒z 5 = 2x3 f 3 + 2x4 f 4 ⇒
z 5 = 2x3 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]+ (7.112)
+2x4 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]
7.3 State Estimation-Based Control of DFIGs 365

and equivalently, one has

ż 5 = L 2f h 2 (x) + L ga (L f h 2 (x))u 1 + L gb (L f h 2 (x))u 2 (7.113)

It holds that

L 2f h 2 (x) =
(− τ4s x3 − 2M τs x 5 + 2vsd )[− τs x 3 + ωdq x 4 + τs x 5 + vsd ]+
1 M

(− τs x4 + τs x6 + 2vsq )[−ωdq x3 − τs x4 + τs x6 + vsq ]+

4 2M 1 M
(7.114)
β
( 2M
τs x 3 )[−βx 2 x 4 + τs x 3 + (ωdq − x 2 )x 6 − γ2 x 5 − βvsd ]+
β
( 2M
τs x 3 )[ τs x 4 + βx 2 x 3 + (ωdq − x 2 )x 5 − γ2 x 6 − βvsq ]

Moreover, it holds that

2M 2M 1
L ga (L f h 2 (x)) = x3 ga5 ⇒L ga (L f h 2 (x)) = x3 (7.115)
τs τs σ L s

and in a similar manner

2M 2M 1
L gb (L f h 2 (x)) = x4 ga6 ⇒L ga (L f h 2 (x)) = x4 (7.116)
τs τs σ L s

Next, it is confirmed that after change in the state variables it holds that

ż 1 = z 2
ż 2 = z 3
ż 3 = L 3f h 1 (x) + L ga (L 2f )h 1 (x)u 1 + L gb (L 2f )h 1 (x)u 2 (7.117)
ż 4 = z 5
ż 5 = L 2f h 2 (x) + L ga (L f )h 2 (x)u 1 + L gb (L f )h 2 (x)u 2

It holds that z 1 = θ , ż 1 = ω = z 2 , ż 2 = ω̇ = f 2 (x) + ga2 u 1 + gb2 u 2 ⇒ż 2 =

f 2 (x) + 0u 1 + 0u 2 which finally gives ż 2 = f 2 (x). Moreover, it has been proven
that z 3 = f 2 , therefore, it holds that ż 2 = z 3 . Moreover, it holds that

∂z 3 ∂z 3 ∂z 3 ∂z 3 ∂z 3 ∂z 3
ż 3 = ẋ1 + ẋ2 + ẋ3 + ẋ4 + ẋ5 + ẋ6 (7.118)
∂ x1 ∂ x2 ∂ x3 ∂ x4 ∂ x5 ∂ x6

which in turn gives

ż 3 = ∂∂zx31 f 1 + ∂∂zx32 f 2 + ∂∂zx33 f 3 + ∂∂zx34 f 4 +

(7.119)
+ ∂∂zx35 ( f 5 + σ 1L r u 1 ) + ∂∂zx36 ( f 6 + σ 1L r u 2 )
366 7 Differential Flatness Theory and Electric Power Generation

which is also written as

ż 3 = L 3f h 1 (x) + L ga (L 2f h 1 (x))u 1 + L gb (L 2f h 2 (x))u 2 (7.120)

Similarly, one has

z 4 = x32 + x42 ⇒
ż 4 = 2x3 ẋ3 + 2x4 ẋ4 ⇒ż 4 = 2x3 f 3 + 2x4 f 4
ż 4 = 2x3 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]+ (7.121)
+2x4 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]⇒ż 4 = z 5

Additionally, it holds that

∂z 5 ∂z 5 ∂z 5 ∂z 5 ∂z 5 ∂z 6
ż 5 = ẋ1 + ẋ2 + ẋ3 + ẋ4 + ẋ5 + ẋ6 (7.122)
∂ x1 ∂ x2 ∂ x3 ∂ x4 ∂ x5 ∂ x6

which in turn gives

ż 5 = ∂∂zx51 f 1 + ∂∂zx52 f 2 + ∂∂zx53 f 3 + ∂∂zx54 f 4 +

(7.123)
+ ∂∂zx55 ( f 5 + σ 1L r u 1 ) + ∂∂zx56 ( f 6 + σ 1L r u 2 )

which subsequently gives

ż 5 = L 2f h 2 (x) + L ga (L f h 2 (x))u 1 + L gb (L f h 2 (x))u 2 (7.124)

and this is the anticipated relation about ż 5 . Consequently, Eq. (7.117) is confirmed
to hold. The system of Eq. (7.117) can be written in the input–output linearized form
 
z 1(3)
= f a + M̃u (7.125)
z̈ 4

where  
L 3f h 1 (x)
f a (x) =
L 2f h 2 (x)
  (7.126)
L ga L 2f h 1 (x) L gb L 2f h 2 (x)
M̃ =
L ga L f h 1 (x) L gb L f h 2 (x)

or equivalently one has the system’s description in the MIMO canonical form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 0 0 z1 0 0
⎜ż 2 ⎟ ⎜0 0 1 0 0⎟ ⎜ z 2 ⎟ ⎜0 0⎟  
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ż 3 ⎟ = ⎜0 0⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ 0 0 0 ⎟ ⎜ z 3 ⎟ + ⎜1 0⎟
v
(7.127)
⎝ż 4 ⎠ ⎝0 0 0 0 1 ⎝ z 4 ⎠ ⎝0
⎠ 0⎠ 2
ż 5 0 0 0 0 0 z5 0 1
7.3 State Estimation-Based Control of DFIGs 367

where
v1 = L 3f h 1 (x) + L ga (L 2f h 1 (x))u 1 + L gb (L 2f h 2 (x))u 2
(7.128)
v2 = L 2f h 2 (x) + L ga (L f h 1 (x))u 1 + L gb (L f h 2 (x))u 2

7.3.4 Input–Output Linearization of the DFIG Using

Differential Flatness Theory

7.3.4.1 Differential Flatness Properties of DFIG

The flat outputs of the system are defined as

y1 = θ or y = x1
(7.129)
y2 = ψs2d + ψs2q or y2 = x32 + x42

It holds that
ẏ1 = ω or ẏ1 = x2 ⇒
ÿ1 = ω̇ = − KJm x2 − TJm + ηJ (x6 x3 − x5 x4 ) ⇒ (7.130)
ÿ1 = ω̇ = − KJm ẏ1 − TJm + ηJ (x6 x3 − x5 x4 ) ⇒

Deriving the last row of Eq. (13.64) with respect to time one obtains

(3)
y1 = − KJm ÿ1 + ηJ (ẋ6 x3 + x6 ẋ3 − ẋ5 x4 − x5 ẋ4 ) ⇒
y1(3) = − KJm ÿ1 + ηJ x3 {[ τβs x4 + βx2 x3 + (ωdq − x2 )x5 −
−γ2 x6 − βvsq ] + σ 1L r u 1 } + ηJ x6 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ] (7.131)
− ηJ x4 {[−βx2 x4 + τβs x3 + (ωdq − x2 )x6 − γ2 x5 −
−βvsd ] + σ 1L r u 2 } − ηJ x5 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]

Moreover, about the second flat output it holds that

ẏ2 = 2x3 ẋ3 + 2x4 ẋ4 ⇒

ẏ2 = 2x3 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]+ (7.132)
+2x4 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]⇒

Consequently, it holds that

ÿ2 = 2 ẋ3 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]+

+2x3 [− τ1s ẋ3 + ωdq ẋ4 + τMs ẋ5 ]+
(7.133)
2 ẋ4 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]+
+2x4 [−ωdq ẋ3 − τ1s ẋ4 + τMs ẋ6 ]
368 7 Differential Flatness Theory and Electric Power Generation

or equivalently

ÿ2 = 2[− τ1s x3 + ωdq x4 + τMs x5 + vsd ]2 − τ2s x3 [− τ1s x3 + ωdq x4 + τMs x5 + vsd ]
−2ωdq x3 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]+
+ τs x3 {[−βx2 x4 + τβs x3 + (ωdq − x2 )x6 − γ2 x5 − βvsd ] + σ 1L r u 1 }
2M

+2[−ωdq x3 − τ1s x4 + τMs x6 + vsq ]2

−2ωdq x4 [− τs x3 + ωdq x4 + τMs x5 + vsd ] − τ2s x4 [−ωdq x3 − τ1s x4 + τMs x6 + vsq ]
1

2x4 τMs {[ τβs x4 + βx2 x3 + (ωdq − x2 )x5 − γ2 x6 − βvsq ] + σ 1L r u 2 }

(7.134)

It holds that x1 = y1 , x2 = ẏ1 . From the second row of Eq. (7.129) and considering
√
that the field orientation condition requires x4 = ψsq = 0 one obtains that x3 = y2 .
Moreover, from Eq. (7.130) it holds that

η√
ÿ1 = − KJm ẏ1 − Tm
J + J y2 x 6 ⇒
ÿ1 + KJm ẏ1 + TJm (7.135)
x6 = η√ , y2 =0
J y2

From Eq. (7.132) one obtains

ẏ2 = − τ2s x32 + 2M

τs x 3 x 5 + 2vsd x 3 ⇒
ẏ2 + ( τs x3 − 2vsd )x3 = 2M
2
τs x 3 x 5 ⇒ (7.136)
√ √
ẏ2 +( τ2s y2 −2vsd ) y2
x5 = 2M √ y2 =0
τs y2

Therefore, x5 is also a function of the flat output and its derivatives. Additionally,
by solving the system of Eqs. (7.131) and (7.134) with respect to the control inputs
u 1 and u 2 one obtains that the control inputs are functions of the flat output and its
derivatives. Therefore, the model of DFIG is a differentially flat one.
Next, to design the flatness-based controller for DFIG the following transforma-
tion of state variables is introduced: z 1 = y1 , z 2 = ẏ1 , z 3 = ÿ1 , z 4 = y2 , z 5 = ẏ2
for which holds

ż 1 = z 2
ż 2 = z 3
ż 3 = L 3f h 1 (x) + (L ga L 2f h 1 (x))u 1 + (L gb L 2f h 1 (x))u 2 (7.137)
ż 4 = z 5
ż 5 = L 2f h 2 (x) + (L ga L f h 2 (x))u 1 + (L gb L f h 2 (x))u 2

Therefore, one obtains the decoupled and linearized representation of the system as
      
(3) L 3f h 1 (x) L ga L 2f h 1 (x) L gb L 2f h 1 (x)
z1 u1
= + (7.138)
z̈ 4 L 2f h 2 (x) L ga L f h 2 (x) L gb L f h 2 (x) u2
7.3 State Estimation-Based Control of DFIGs 369

or equivalently  
(3)
z1
= f a + M̃u (7.139)
z̈ 4

where    
L 3f h 1 (x) L ga L 2f h 1 (x) L gb L 2f h 1 (x)
fa = M̃ = (7.140)
L 2f h 2 (x) L ga L f h 2 (x) L gb L f h 2 (x)

By defining the control inputs v1 = L 3f h 1 (x) + (L ga L 2f h 1 (x))u 1 + (L gb L 2f h 1 (x))u 2

and v2 = L 2f h 2 (x) + (L ga L f h 2 (x))u 1 + (L gb L f h 2 (x))u 2 one can also have the
description in the MIMO canonical form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 0 0 z1 0 0
⎜ż 2 ⎟ ⎜0 0 1 0 0⎟ ⎜ z 2 ⎟ ⎜0 0⎟  
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ż 3 ⎟ = ⎜0 0⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ 0 0 0 ⎟ ⎜ z 3 ⎟ + ⎜1 0⎟
v
(7.141)
⎝ż 4 ⎠ ⎝0 0 0 0 1 ⎝ z 4 ⎠ ⎝0
⎠ 0⎠ 2
ż 5 0 0 0 0 0 z5 0 1

The control input for the linearized and decoupled model of the DFIG is chosen as
follows:
(3) (1) (1) (1)
v1 = z 1d − k1 (z̈ 1 − z̈ 1d ) − k2 (ż 1 − ż 1d ) − k3 (z 1 − z 1d )
(2) (2) (7.142)
v2 = z̈ 4d − k1 (ż 4 − ż 4d ) − k2 (z 4 − z 4d )

and finally the control input that is applied to the system is

u = M̃ −1 (− f a + v) (7.143)

The proposed control scheme can work with the use of measurements from a small
number of sensors. That is, there is need to obtain measurements of only y1 = θ
which is the turn angle of the generator’s rotor, and of the magnetic flux y2 = ψs2 =
ψs2d + ψs2q , or due to the orientation of the magnetic field y2 = ψs2 = ψs2d . The stator
flux (ψs ) cannot be measured directly from a sensor (e.g., the use of Hall sensor in
an electric machine with a rotating part would not be efficient), however Eq. (7.87)
that relates stator flux and stator and rotor currents can be used to calculate ψs . Thus
one has:
ψsd = L s i sd + Mird
(7.144)
ψsq = 0

which means that by measuring stator and rotor currents one can obtain an indirect
measurement of the stator’s magnetic flux ψsd . Next, one can compute the dynamics
of the magnetic flux, jointly with the dynamics of the rotor’s motion through the
use of Derivative-free Nonlinear Kalman Filter. Actually, this estimation method is
based on the application of the Kalman Filter recursion to the linearized equivalent
370 7 Differential Flatness Theory and Electric Power Generation

of the generator’s model given by Eq. (7.141). Equation (7.141) can be written in the
state-space form
ż = Az + Bv
(7.145)
z meas = C z

where ⎛ ⎞ ⎛ ⎞
01000 00
⎜0 0 1 0 0 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟
A=⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 ⎟ B = ⎜ 1 0 ⎟
⎝0 0 0 0 1 ⎠ ⎝0 0⎠ (7.146)
00000 01
 
10000
C=
00010

The estimator’s dynamics is:

ẑ˙ = A · ẑ + B · v + K (z meas − C ẑ) (7.147)

where K ∈ R 5×2 is the state estimator’s gain. Defining as Ãd , B̃d , and C̃d , the
discrete-time equivalents of matrices A, B, and C respectively, the associated Kalman
Filter-based estimator is given by [31, 229, 405]
measurement update:

K (k) = P − (k)CdT [Cd · P − (k)CdT + R]−1

ẑ(k) = ẑ − (k) + K (k)[z meas (k) − Cd ẑ − (k)] (7.148)
P(k) = P − (k) − K (k)Cd P − (k)

time update:

P − (k + 1) = Ad (k)P(k)AdT (k) + Q(k)

(7.149)
ẑ − (k + 1) = Ad (k)ẑ(k) + Bd (k)v(k)

In terms of computation, the differential flatness theory-based linearization is simpler

because it does not require the calculation of Lie derivatives. Moreover, by express-
ing all state variables as functions of the flat output and its derivatives, the differential
flatness theory-based linearization enables to perform state estimation and to recon-
struct the state variables of the initial nonlinear system. This is not possible for the
Lie algebra-based approach, where to perform filtering it is necessary to compute
and invert the Jacobian matrix of the transformed state vector [441].
Compared with other control approaches for DFIGs, e.g., control of the rotor’s
speed and of the stator’s magnetic flux in cascading loops analyzed in Ref. [442], one
can use measurements of the rotor currents. The rest of the state variables of DFIG
have to be estimated with filtering methods that make use of the initial nonlinear
model of the system, such as the Extended or the Unscented Kalman Filter. Being
based on an exact linearization method, the control and state estimation approach for
7.3 State Estimation-Based Control of DFIGs 371

DFIG that is presented in Sect. 7.3.4 has the advantage of using a reduced number
of sensors while at the same time remaining robust to modeling uncertainties and
external perturbations and avoiding numerical approximation errors.

7.3.5 Kalman Filter-Based Disturbance Observer

for the DFIG Model

Next, it will be considered that additive input disturbances (e.g., due to load varia-
tions) affect the DFIG model. Thus, it is assumed that the third and fifth rows of the
state-space equations of the Doubly-Fed Induction Generator of Eq. (7.137) include
a disturbance term

ż 3 = L 3f h 1 (x) + L ga (L 2f h 1 (x))u 1 + L gb (L 2f h 2 (x))u 2 + d̃1

(7.150)
ż 5 = L 2f h 2 (x) + L ga (L f h 2 (x))u 1 + L gb (L f h 2 (x))u 2 + d̃2

Without loss of generality, the dynamics of the disturbance terms is described by

their second order derivatives and the associated initial conditions, i.e., d̃¨1 = f˜a (x)
and d̃¨2 = f˜b (x). Next, an extended state-space model of the system is defined that
comprises as additional state variables the disturbance terms z 6 = d̃1 , z 7 = d̃˙1 , while
z 8 = d̃2 , and z 9 = d̃˙2 . Thus, the extended state-space model is written as ż 1 = z 2 ,
ż 2 = z 3 , ż 3 = v1 + z 6 , ż 4 = z 5 , and ż 5 = v2 + z 8 , ż 6 = z 7 , ż 7 = f˜a , ż 8 = z 9 and
ż 9 = f˜b , or in matrix form one has

z̃˙ = Ãz̃ + B̃ ṽ
(7.151)
z̃ meas = C̃ z̃
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010000000 z1 0000
⎜ż 2 ⎟ ⎜0 0 1 0 0 0 0 0 0⎟ ⎜z 2 ⎟ ⎜0 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ż 3 ⎟ ⎜0 0 0 0 0 1 0 0 0⎟ ⎜z 3 ⎟ ⎜1 0 0 0⎟ ⎛ ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ż 4 ⎟ ⎜0 0 0 0 1 0 0 0 0⎟ ⎜z 4 ⎟ ⎜0 0 0 0⎟ v1
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ż 5 ⎟ = ⎜0 0 0 0 0 0 0 1 0⎟ ⎜z 5 ⎟ + ⎜0 1 0 0⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎝ ˜ ⎠
⎜ż 6 ⎟ ⎜0 0 0 0 0 0 1 0 0⎟ ⎜z 6 ⎟ ⎜0 0 0 0⎟ f a
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ (7.152)
⎜ż 7 ⎟ ⎜0 0 0 0 0 0 0 0 0⎟ ⎜z 7 ⎟ ⎜0 0 1 0⎟ f˜b
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ż 8 ⎠ ⎝0 0 0 0 0 0 0 0 1⎠ ⎝z 8 ⎠ ⎝0 0 0 0⎠
ż 9 000000000 z9 0001
 meas   
z1 100000000
= z
z 4meas 000100000
372 7 Differential Flatness Theory and Electric Power Generation

where z = [z 1 , z 2 , . . . , z 9 ]T . The associated state estimator is

żˆ = Ão ẑ + B̃o ṽ1 + K o (z̃ meas − C̃ ẑ) (7.153)

where ⎛ ⎞
010000000 ⎛ ⎞
⎜0 0 1 0 0 0 0 0 0 ⎟ 0000
⎜ ⎟ ⎜0 0 0 0 ⎟
⎜0 0 0 0 0 1 0 0 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜1 0 0 0 ⎟
⎜0 0 0 0 1 0 0 0 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟
Ão = ⎜ 0 0 0 0 0 0 0 1 0 ⎟ B̃o = ⎜0 0 0 0⎟
⎜ ⎟ ⎜0 1 0 0 ⎟
⎜0 0 0 0 0 0 1 0 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜0 0 0 0 ⎟ (7.154)
⎜0 0 0 0 0 0 0 0 0⎟ ⎜ ⎟
⎜ ⎟ ⎝0 0 0 0 ⎠
⎝0 0 0 0 0 0 0 0 1⎠
0000
000000000
 
100000000
C̃o =
000100000

while the estimator’s gain K o ∈ R 9×2 is obtained from the standard Kalman Filter
recursion [31, 229, 405].
Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents of matrices Ão , B̃o and
C̃o , respectively, a Derivative-free nonlinear Kalman Filter can be designed for the
aforementioned representation of the system dynamics [427, 436]. The associated
Kalman Filter-based disturbance estimator is given as
measurement update:

K (k) = P − (k)C̃dT [C̃d · P − (k)C̃dT + R]−1

ˆ
z̃(k) = ẑ − (k) + K (k)[C̃d z̃(k) − C̃d z̃ˆ − (k)] (7.155)
P(k) = P − (k) − K (k)C̃d P − (k)

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

ˆz̃ − (k + 1) = Ãd (k)z̃(k)
ˆ (7.156)
+ B̃d (k)ṽ(k)

The advantages of the proposed nonlinear feedback control method for DFIGs (based
on differential flatness theory and on the Derivative-free nonlinear Kalman Filter)
against PID type control (included in vector control loops) are obvious. In most cases
the application of PID control to electric machines is based on heuristic parameters
tuning, has no stability proof, and has limited robustness to the change in operating
points or to the effects of external perturbations. Moreover, in the case of multi-
variable systems such as DFIGs the application of PID control is known to have
questionable performance. Although the first approaches for asynchronous machines
control (e.g., Blascke 1971) made use of multiple PID loops which were implemented
in a cascaded manner, (for controlling separately the magnetic flux and the rotation
7.3 State Estimation-Based Control of DFIGs 373

angle of the machine), such methods were based on the assumption that the flux
and the rotation speed become finally decoupled at steady state. There is no proof
about that (it cannot be always assured that transients will be eliminated and the
machine will reach a steady-date) and therefore the performance of the control loop
is not always guaranteed. Therefore, although PID control is met in some cases in
asynchronous machines, it is not the recommended solution.
Regarding stability features it is confirmed that the linearized equivalent model of
the DFIG, after application of the pole placement technique has poles, exclusively in
the left complex semi-plane. Besides, the inclusion of the additional control input that
compensates for the estimated additive disturbance terms improves the robustness
features of this control loop. It is also noted that the linearized equivalent model of
the DFIG exhibits multiple poles at the origin. Thus stabilization can be succeeded by
pole placement methods. Finally, it is noted that the stability and robustness features
of the control scheme, which comprises also estimation and compensation of the
disturbances, are similar to those of LQG control. Based on the above the chapter
provides sufficient evidence about the stability and disturbance rejection capability of
the proposed feedback control scheme. On the other hand, the presented simulation
experiments demonstrated the efficiency of the control method in tracking rapidly
changing reference setpoints while also succeeding good transients. The disturbances
appearing in the simulation experiments could be met in adverse operating conditions
of the power generator. Even for the latter case the good performance of the control
loop is confirmed.

7.3.6 Simulation Tests

The structure of the proposed control scheme is depicted in Fig. 7.13. The control
loop comprises (i) the flatness-based control part which computes the control signal
for the system’s equivalent model that is transformed to the linear canonical form,
(ii) a Kalman Filter-based disturbances estimator which provides estimates for the
elements of the state vector of the DFIG, such as rotor’s speed, magnetic flux at the
stator, and disturbances affecting the generator’s model.
Indicative numerical values for the parameters of the considered doubly-fed induc-
tion generator model are given in Table 7.1.
Simulation tests were carried out for two different setpoints of the turn speed
of the generator’s rotor. The results obtained for the first setpoint are depicted in
Figs. 7.14 and 7.15. Similarly, the results obtained for the second setpoint are depicted
in Figs. 7.16 and 7.17. It can be observed that the proposed control scheme assures
that the rotor’s turn speed follows a specific setpoint, while tracking of reference
setpoints is succeeded for the components of the magnetic flux and for the rotor’s
currents. Several reference setpoints have been defined for the DFIG state variables,
i.e., rotor’s angular speed ω, rotor currents ird , irq and the magnetic flux ψsd and as it
can be observed from the associated diagrams, the proposed control scheme results
in fast and accurate convergence to these setpoints. The disturbance observer that
374 7 Differential Flatness Theory and Electric Power Generation

Fig. 7.13 Control loop of the DFIG comprising a flatness-based control element and an estimator
for disturbances compensation

Table 7.1 Ratings of the modeled DFIG

Rated power 15.5 kW
Number of pole pairs 4
Stator resistance 0.58 
Stator inductance 13 · mH
Rotor resistance 1.30 
Rotor inductance 3 · mH
Mutual inductance 10 · mH
Rotor’s inertia 20.0 kg · m2

(a) 80 (b) 3

70
2.5
60
2
50
ψs d

40 1.5
ω

30
1
20
0.5
10

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time (sec) time (sec)

Fig. 7.14 DFIG setpoint 1: a Control of state variable x2 = ω. b Control of state variable x3 = ψsd
7.3 State Estimation-Based Control of DFIGs 375

(a) 0.025 (b) 3 1.5

1
0.02 2
0.5

d/dt d1
d1
1
ird

0.015 0
-0.5
0.01 0
-1
0.005 -1 -1.5
0 5 10 15 20 25 30 35 40 0 10 20 30 40 0 10 20 30 40
time (sec) time (sec) time (sec)
30 3 1.5
25 1
2

d/dt d2
20 0.5

d2
irq

15 1 0
10 -0.5
0
5 -1
0 -1 -1.5
0 5 10 15 20 25 30 35 40 0 10 20 30 40 0 10 20 30 40
time (sec) time (sec) time (sec)

Fig. 7.15 DFIG setpoint 1: a Control of state variable x5 = ird and of state variable x6 = irq . b
Estimation of disturbance inputs di , i = 1, 2 and of their derivatives

(a) 90 (b) 3

80
2.5
70

60 2

50
ψs d

1.5
ω

40

30 1

20
0.5
10
0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time (sec) time (sec)

Fig. 7.16 DFIG setpoint 2: a Control of state variable x2 = ω. b Control of state variable x3 = ψsd

was based on the Derivative-free nonlinear Kalman Filter was capable of estimating
the unknown and time-varying input disturbances affecting the DFIG model.
The improvement in the performance of the control loop due to the use of a dis-
turbance observer based on the Derivative-free nonlinear Kalman Filter is explained
as follows: (i) compensation of the disturbance terms which is generated by para-
metric uncertainty or unknown external inputs, (ii) more accurate estimation of the
disturbance terms because the filtering procedure is based on an exact linearization
of the system’s dynamics and does not introduce numerical errors (as for example
in the case of the Extended Kalman Filter). This is shown in Figs. 7.14, 7.15, 7.16,
7.17 and 7.18.
376 7 Differential Flatness Theory and Electric Power Generation

(a) 0.025 (b) 3 1.5

1
0.02 2

d/dt d1
0.5

d1
0.015
ird

1 0
-0.5
0.01 0
-1
0.005 -1 -1.5
0 5 10 15 20 25 30 35 40 0 10 20 30 40 0 10 20 30 40
time (sec) time (sec) time (sec)
40 3 1.5
1
30 2

d/dt d2
0.5

d2
irq

20 1 0
10 -0.5
0
-1
0 -1 -1.5
0 5 10 15 20 25 30 35 40 0 10 20 30 40 0 10 20 30 40
time (sec) time (sec) time (sec)

Fig. 7.17 DFIG setpoint 2: a Control of state variable x5 = ird and of state variable x6 = irq . b
Estimation of disturbance inputs di , i = 1, 2 and of their derivatives

(a) 3 (b) 3

2.5 2.5

2 2
ψs d

ψs d

1.5 1.5

1 1

0.5 0.5

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time (sec) time (sec)

Fig. 7.18 Convergence of the stator’s magnetic flux ψsd to the reference setpoint, a without using
the disturbance observer, b when using the disturbance observer

Variables d̃1 and d̃2 appearing in Eq. (7.150) are aggregate disturbance terms which
include any type of perturbations that may be due to load variations and change of
the stator currents, change of the mechanical torque, voltage fluctuation and faults
in the grid (vsd and vsq non constant), or modeling uncertainty and changes in the
numerical values of the parameters appearing in the DFIG model. Representing the
aggregate disturbance effects as in Eq. (7.150) enables the design of a disturbances
estimator and compensator based on the Derivative-free nonlinear Kalman Filter.
7.4 Flatness-Based Control of DFIG in Cascading Loops 377

7.4 Flatness-Based Control of DFIG in Cascading Loops

7.4.1 Overview

The significance of the control problem for Doubly-fed induction generators was
analyzed in the previous section. Taking into account that the installation and main-
tenance of sensors for measuring several parameters of the generator’s state vector can
be technically difficult or costly, the need for developing sensorless control schemes
for DFIG becomes apparent. In this section, a novel sensorless control scheme for
DFIGs is developed using flatness-based control theory and a state estimation method
that is based on Extended Kalman Filtering. The controller is implemented in cas-
cading loops.
In the present section, it is shown that the doubly-fed induction generator model is
differentially flat, using a flat output other than the one presented in Sect. 7.3. Now, the
proposed flat output is the pair consisting of the rotor’s angle θ and of the orientation
of the magnetic flux ρ. It is shown that the doubly-fed induction generator model
is differentially flat since all its state variables and control inputs can be written as
functions of the flat output and of the associated derivatives. The differential flatness
property shows that the design of a DFIG controller can take the form of cascading
feedback loops [459]. The design of the DFIG controller consists of two stages: (i)
in the outer-loop the controller enables convergence of the stator’s magnetic flux
and of the rotor’s angular velocity to the associated reference setpoint. The control
input consists of the rotor currents expressed in the rotating reference frame (ii) in
the inner-loop the controller enables convergence of the rotor’s dq currents to the
associated reference setpoints, where these setpoints are the control inputs used by
the outer control loop. The control inputs for the inner loop are now the rotor voltages,
also expressed in the rotating reference frames.
Next, sensorless control for the doubly-fed induction generator is implemented. A
main difficulty in the implementation of DFIG control is that some of its state vector
elements are not directly measured and their monitoring requires the installation of
specific sensors that can be costly, or difficult to install and maintain. The Extended
Kalman Filter is used to estimate the nonmeasurable elements of the DFIG state
vector such as the rotation speed and the magnetic flux through the processing of
measurements of the rotor’s angle and of the rotor’currents. As explained, the Kalman
Filter is a special kind of observer that provides optimal filtering of process and
measurement noises if the covariances of these noises are known. For linear systems
subject to Gaussian measurement or process noise the Kalman Filter is the optimal
state estimator, since it results in minimization of the trace of the estimation error’s
covariance matrix [31, 229, 408, 573]. As noted in Chap. 4, for nonlinear systems
subject to Gaussian noise one can use the generalization of the Kalman Filter as
formulated in terms of the Extended Kalman Filter (EKF). The Extended Kalman
Filter is based on a linearization of the systems’ dynamics using a first order Taylor
expansion [405, 427]. In this section it is shown that with the use of nonlinear Kalman
378 7 Differential Flatness Theory and Electric Power Generation

Filtering it is possible to reduce the number of sensors involved in the control loops of
the doubly-fed induction generator and to implement efficient state estimation-based
control.

7.4.2 A New Proof of the Differential Flatness of the DFIG

As shown, the dynamic model of the doubly-fed induction generator can be written
in a state-space equations form by defining the following state variables: x1 = θ ,
x2 = ωr , x3 = ψsd , x4 = ψsq , x5 = ird and x6 = irq . It holds that

ẋ1 = x2 (7.157)

Km Tm η
ẋ2 = − x2 − + (i sq x3 − i sd x4 ) (7.158)
J J J
1 M
ẋ3 = − x3 + ωdq x4 + x5 + vsd (7.159)
τs τs

1 M
ẋ4 = −ωdq x3 − x4 + x6 + vsq (7.160)
τs τs

β 1
ẋ5 = −βx2 x4 + x3 + (ωdq − x2 )x6 − γ2 x5 + vr − βvsd (7.161)
τs σ Ls d

β
ẋ6 = x4 + βx2 x3 + (ωdq − x2 )x5 − γ2 x6 (7.162)
τs

which can be written also in the form

ẋ = f (x) + ga (x)vrd + gb (x)vrq (7.163)

where x = [x1 , x2 , x3 , x4 , x5 , x6 ]T and

⎛ ⎞
x2
⎜ − KJm x2 − TJm + nJ (i sq x3 − i sd x4 ) ⎟
⎜ ⎟
⎜ − 1
x + ω x + M
x + v ⎟
⎜ τs 3 dq 4 τs 5 sd ⎟
f (x) = ⎜ −ωdq x3 − τs x4 + τs x6 + vsq
1 M ⎟
⎜ ⎟
⎜ β ⎟
⎝−βx2 x4 + τ x3 + (ωdq − x2 )x6 − γ2 x5 − βvsd ⎠ (7.164)
s
β
τs x 4 + βx 2 x 3 + (ωdq − x 2 )x 5 − γ2 x 6
 T
ga (x) = 0 0 0 0 σ 1L s 0
 T
gb (x) = 0 0 0 0 0 σ 1L s
7.4 Flatness-Based Control of DFIG in Cascading Loops 379

Next, it will be shown that the doubly-fed induction generator model is differentially
flat. First the stator’s magnetic flux is written in a complex number form

ψ = ψsd + jψsq = αe jρ (7.165)

ψs
where ρ = tan −1 ( ψsb ) denotes the orientation angle of the magnetic flux vector.
a
According to Lorentz law the torque that is developed due to the magnetic flux is
given by
Te = k1 Im (ψs∗ dψ
dt )⇒Te = k1 Im (αe
s − jρ α j ρ̇e jρ )⇒
(7.166)
Te = k1 I m( jα ρ̇)⇒Te = k1 α 2 ρ̇
2

The generator’s equations of motion are

J ω˙r = Tm − bω̇r − T
e⇒
(7.167)
J θ̈ = Tm − bθ̇ − k1 α 2 ρ̇⇒α = (Tm −b θ̇−J θ̈)
k1 ρ̇

As flat output, the following pair of variables is considered y = (θ, ρ), where θ is
the rotor’s rotation angle and ρ is the orientation angle of the magnetic field. Thus,
in the doubly-fed induction generator model the magnetic flux can be written as a
function of the flat output y and of the associated derivatives. This is denoted as

ψs = f g (y, ẏ, ÿ) (7.168)

Moreover, due to the field orientation it holds that

ψsd = Re{ f g (y, ẏ, ÿ)}, ψsq = I m{ f g (y, ẏ, ÿ)} = 0 (7.169)

Considering that the stator feed voltages vsd and vsq are constant, and that due to
field orientation it holds that ψsq = 0, from Eq. (7.77) it holds

1 dψsd 1
i sd = [vs + − ωdq ψsq ]⇒i sd = f g (y, ẏ, ÿ) (7.170)
Rs d dt Rs 1

Moreover, from Eq. (7.78) one has

1 dψsq 1
i sq = [vsq − + ωdq ψsd ]⇒i sq = f g (y, ẏ, ÿ) (7.171)
Rs dt Rs 2

From Eq. (7.81) one obtains

1
i rd = [ψsd − L s i sd ]⇒ird = f g3 (y, ẏ, ÿ) (7.172)
M
380 7 Differential Flatness Theory and Electric Power Generation

From Eq. (7.82) one obtains

1
i rq = [ψsq − L s i sq ]⇒irq = f g4 (y, ẏ, ÿ) (7.173)
M
Additionally, from Eq. (7.83) it holds that

ψrd = L s ird + Mi sd ⇒ψrd = f g5 (y, ẏ, ÿ) (7.174)

while from Eq. (7.84) one has

ψrq = L s irq + Mi sq ⇒ψrq = f g6 (y, ẏ, ÿ) (7.175)

Since variables ψrd ,ψrq , ird and irq have been expressed as functions of the flat
output and of the associated derivatives the same can be done for the control inputs
vrd and vrq .
From Eq. (7.79) and using Eqs. (7.176) and (7.174) one obtains

dψrd
vrd = Rr ird + − ωdq ψrq ⇒vrd = f g7 (y, ẏ, ÿ) (7.176)
dt
From Eq. (7.80) and using Eqs. (7.177) and (7.175) one has

dψrq
vrq = Rr irq + − ωdq ψrd ⇒vrq = f g8 (y, ẏ, ÿ) (7.177)
dt
Since all elements of the state vector x = [θ, ω, ψsd , ψsq , ird , irq ] and the control
inputs vrd and vrq of the doubly-fed induction generator can be expressed as functions
of the flat output y = [θ, ρ] and of the associated derivatives, it can be concluded
that the DFIG model is differentially flat.

7.4.3 Control of the DFIG in Cascading Loops

The differential flatness property shows that the design of a DFIG controller is pos-
sible using feedforward control terms which are complemented by suitable error
feedback terms. The design of the controller of the doubly-fed induction genera-
tor consists of two stages: (i) in the outer-loop the controller enables convergence
of the stator’s magnetic flux to the associated reference setpoint. The control input
consists the rotor currents on the d and q reference axes (ii) in the inner-loop the
controller enables convergence of the rotor’s dq currents to the associated reference
setpoints, where these setpoints are the control inputs used by the outer control loop
(see Fig. 7.19). The control inputs for the inner loop are now the rotor voltages vrd
and vrq .
7.4 Flatness-Based Control of DFIG in Cascading Loops 381

Fig. 7.19 Implementation of DFIG control with cascading feedback loops

The outer control loop views as flat outputs the vector [ψsd , ψsq ]. For the outer
control loop the stator flux dynamics is given by:

dψsd 1 M
= ωdq ψsq − ψsd + ird + vsd (7.178)
dt τs τs

dψsq 1 M
=− ψs − ωdq ψsd + irq + vsq (7.179)
dt τs q τs

Inner control loop of the rotor’s currents:

The inner control loop views as flat outputs the vector [ird , irq ]. For the inner control
loop the rotor current dynamics is given as:

= −βωr ψsq + τβs ψsd +

dird
dt (7.180)
+(ωdq − ωr )irq − γ2 ird − βvsd + σ 1L r vrd

= τβs ψsq − βωr ψsd − γ2 irq −

dirq
dt (7.181)
−(ωdq − ωr )ird − βvsq + σ 1L r vrq

The following control law is applied to the inner loop to make the rotor currents
convergence to the desirable reference setpoints:

β
vrd = σ L r [βωr ψsq − τs ψsd −
dir∗ (7.182)
−(ωdq − ωr )irq + γ2 ir∗d + βvsd + dt
d
]

vrq = σ L r [− τβs ψsq − βωr ψsd + γ2 ir∗q +

dir∗q (7.183)
+(ωdq − ωr )ird + βvsq + dt ]
382 7 Differential Flatness Theory and Electric Power Generation

The inner control loop is designed so as to succeed irq →ir∗q and ird →ir∗d , where the
setpoints ir∗q and ir∗d are the control inputs that make the stator’s flux outer control
loop converge to the flux reference setpoints ψs∗d and ψs∗q .
It is also noted that in DFIGs, the rotor-side converter is used to control the
turbine output power and the voltage measured at the grid terminals. The power is
controlled in order to follow a predefined power-speed characteristic, named tracking
characteristic. The grid side converter is used to regulate the voltage of the dc bus
capacitor. In addition, by using the grid side converter it is also possible to generate
or absorb reactive power [63, 162].
Outer control loop of the stator’s flux:
For the control of the magnetic flux, from Eqs. (7.86) and (7.87) the following matrix
form is obtained:
 dψ         
sd
ω dq − 1
τ ψ s v s
M
τ 0 i dr
dt = s d
+ d
+ s M (7.184)
dψsq
− τ1 −ωdq ψsq vsq 0 τ i qr
dt s s

The magnetic flux reference setpoint is defined as ψs∗ = [ψs∗d , ψs∗q ]T . Then the
control law that assures that ψs →ψs∗ is given as

τs 1
i rd = [−vsd + ψsd − ωdq ψsq + ψ̇s∗d − r1 (ψsd − ψs∗d )] (7.185)
M τs

τs 1
i rq = [−vsq + ψsq + ωdq ψsd + ψ̇s∗q − r2 (ψsq − ψs∗q )] (7.186)
M τs

To speedup convergence of the flux tracking error to zero it suffices to select appro-
priately coefficients r1 and r2 appearing in the right part of Eqs. (7.185) and (7.186).
Thus, in the outer control loop one defines the flux reference setpoint ψs =
[ψsd , ψsq ]T which can be attained by applying an appropriate control input ird , irq .
The control input ird , irq becomes reference setpoint ir∗d , ir∗q for the inner control
loop. The latter control loop assures convergence of ird , irq to ir∗d , ir∗q .
Finally, considering that in the dq reference frame ψsq = 0, the rotation speed of
the rotor is controlled by an input associated to the magnetic flux

J Tm Km
ψsd = ( + ωr ) + ω̇r∗ − r3 (ωr ) − ωr∗ (7.187)
n p i sq J J

The control signal ψsd computed through Eq. (7.187) becomes reference setpoint for
the flux control loop shown in Fig. 7.19.
7.4 Flatness-Based Control of DFIG in Cascading Loops 383

7.4.4 EKF Implementation for Sensorless Control of the DFIG

The nonlinear state space equation of the doubly-fed induction generator is given by

ẋ = f (x) + ga vrd + gb vrq (7.188)

where the state vector has been defined as x = [θ, ω, ψsd , ψsq , ird , irq ]T while
functions f (x), ga (x) and gb (x) have been defined as
⎛ ⎞
x2
⎜ − KJm x2 − TJm + ηJ (i sq x3 − i sd x4 ) ⎟
⎜ ⎟
⎜ − τ1s x3 + ωdq x4 + τMs x5 + vsd ⎟
⎜ ⎟
f (x) = ⎜ −ωdq x3 − τs x4 + τs x6 + vsq
1 M ⎟ (7.189)
⎜ ⎟
⎜ ⎟
⎝−βx2 x4 + τβ x3 + (ωdq − x2 )x6 − γ2 x3 ⎠
s
β
τs x 4 + βx 2 x 3 + (ωdq − x 2 )x 5 − γ2 x 6

 T
1
ga (x) = 0 0 0 0 σ Ls 0 (7.190)
 
1
gb (x) = 0 0 0 0 0 σ Ls (7.191)

The measured state variables are taken to be x1 = θ , x5 = ird and x6 = irq . Then,
the associated Jacobian matrices are:
 T
Jφ = Jφ1 Jφ2 Jφ3 Jφ4 Jφ5 Jφ6 (7.192)

Km η η
where Jφ1 = [0, 1, 0, 0, 0, 0], Jφ2 = [0, J , J i sq , − J i sd , 0, 0], Jφ = [0, 0, − τs ,
3 1

ωdq , τMs , 0], Jφ4 = [0, 0, −ωdq , − τ1s , 0, τMs ], Jφ5 = [0, −βx4 −x6 , −γ2 + τβs , −βx2 , 0,
ωdq − x2 ] and Jφ6 = [0, βx3 − x5 , βx2 , τβs , ωdq − x2 , −γ2 ].
⎛ ⎞
100000
Jγ = ⎝0 0 0 0 1 0⎠ (7.193)
000001

It is noted that: (i) the discrete-time model of Eq. (4.7) can be computed from
Eq. (7.163) after applying common discretization methods (e.g., first order Euler
approximation of the continuous-time derivative), (ii) if a Jacobian matrix Jφ , asso-
ciated to the drift term of the system’s dynamics, is computed using the system’s
continuous-time description of Eq. (7.163), then in the EKF recursion of Eqs. (4.13)
and (4.14) it should be substituted by I + Ts Jφ , where Ts is the sampling period and
I ∈ R n×n is the identity matrix.
384 7 Differential Flatness Theory and Electric Power Generation

7.4.4.1 Estimation of the Wind-Generated Mechanical Torque Using

EKF

The Extended Kalman Filter can be used not only for estimation of the nonmeasur-
able state vector elements of the DFIG (such as stator’s magnetic flux and rotor’s
angular velocity) but also for estimation of the wind generated mechanical torque.
One approach to succeed this is to redefine the parameters of the state vector of the
induction generator as follows: x1 = θ , x2 = Tm , x3 = ωr , x3 = Ṫm , x5 = ψsd ,
x6 = ψsq , x7 = ird and x8 = irq . Using the relation about the dependence of the
mechanical torque on the design characteristics of the wind turbine as well as on
the rotor’s speed given in Eq. (7.76) and considering constant or piecewise constant
wind speed v, the state equations of the DFIG can be written as
⎛ ⎞
x3
⎜ x4 ⎟
⎜ ⎟
⎜ − K m x3 − x2 + η (i s x5 − i s x6 ) ⎟
⎜ J J J q d ⎟
⎜ 3 d 2 C (λ, β)v2 ⎟
⎜ 1
ρπ R q ⎟
⎜
f (x) = ⎜
2 dt 2
⎟ (7.194)
− 1
x + ω x + M
x + v ⎟
⎜ τs 5 dq 6 τs 7 sd ⎟
⎜ −ωdq x5 − τs x6 + τs x8 + vsq
1 M ⎟
⎜ ⎟
⎜ β ⎟
⎝−βx x
3 6 + x
τs 5 + (ω dq − x )x
3 8 − γ 2 5⎠
x
β
τs x 6 + βx 3 x 5 + (ωdq − x 3 )x 7 − γ2 x 8

 T
1
ga (x) = 0 0 0 0 0 0 σ Ls 0 (7.195)
 
1
gb (x) = 0 0 0 0 0 0 0 σ Ls (7.196)

Introducing to the DFIG state vector the new state variables x2 and x4 and defining
function
1 ωR
C ∗ (x3 ) = ρπ R 3 C̈q ( , β)v2 (7.197)
2 v
the Jacobian matrices Jφ and Jγ associated to the mechanical torque and used in the
updated EKF recursion are written as
 T
Jφ = Jφ1 Jφ2 Jφ3 Jφ4 Jφ5 Jφ6 Jφ7 Jφ8 (7.198)

where

Jφ1 = [0, 0, 1, 0, 0, 0, 0, 0]T Jφ2 = [0, 0, 0, 1, 0, 0, 0, 0]

1 Km n p i sq n p i sd ∂ C̈ ∗ (x3 )
Jφ3 = [0, − , − , 0, ,− , 0, 0] Jφ4 = [0, 0, , 0, 0, 0, 0, 0]
J J J J ∂ x3
7.4 Flatness-Based Control of DFIG in Cascading Loops 385

1 M 1 M
Jφ5 = [0, 0, 0, 0, − , ωdq , , 0] Jφ6 = [0, 0, 0, 0, −ωdq , − , 0, ]
τs τs τs τs
β
Jφ7 = [0, 0, −βx6 − x8 , 0, − γ2 , −βx3 , 0, ωdq ]
τs
β
Jφ8 = [0, 0, βx5 − x7 , 0, βx3 , , ωdq − x3 , −γ2 ]
τs

⎛ ⎞
1000000 0
Jγ = ⎝0 0 0 0 0 0 1 0⎠ (7.199)
0000000 1

Once again, when the Jacobian matrix Jφ , associated to the drift term of the system’s
dynamics, is computed using the system’s continuous-time description of Eq. (7.163),
then in the EKF recursion of Eqs. (4.13) and (4.14) it should be substituted by I +Ts Jφ
where Ts is the sampling period and I ∈ R n×n is the identity matrix.
The estimation of the mechanical torque Tm using the EKF recursion can be
more accurate than calculating Tm with the use of Eq. (7.76) since there may be
inaccuracies about the values of parameters ρ, R, λ, and v as well in the modeling
of the torque coefficient Cq . This type of modeling uncertainty is taken into account
and compensated by the EKF, through the use of the process and measurement noise
covariance matrices Q and R in the estimator’s recursion described in Eqs. (4.13)
and (4.14).

7.4.5 Simulation Tests

The efficiency of the proposed flatness-based control scheme in cascading loops was
tested in case of variable rotor speed setpoint assuming different wind profiles. The
obtained results for the first wind torque profile are shown in Figs. 7.20, 7.21 and
7.22, while the results for the second wind torque profile are shown in Figs. 7.23, 7.24
and 7.25. The real state variable is denoted by the dashed blue line, the estimated state
variable is denoted by the dashed green line, while the associated reference setpoint
is denoted by the continuous red line. Although using a small number of sensors
and measuring only specific elements of the DFIG state vector such as the rotor’s
angle and the stator’s currents, the proposed control scheme succeeded in accurate
tracking of the reference setpoints and accurate estimation of the nonmeasurable
parameters such as the rotation speed of the rotor, the magnetic flux of the stator, and
the wind-generated mechanical torque.
Taking into account that several variables of the induction generator’s state vector
(e.g., rotation speed and magnetic flux) are not directly measurable (due to sensors
cost and limited reliability, sensors failures, and difficulties in sensors installation)
the significance of state estimation through Kalman Filtering becomes clear. It can be
386 7 Differential Flatness Theory and Electric Power Generation

(a) 40 (b) 16

magnetic flux ψs d of the induction

rotation speed ω of the induction

35 14

12
30
generator (rad/sec)

generator (Wb)
10
25
8
20
6
15
4
10
2
5 0

0 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.20 Extended Kalman Filter-based control of the doubly-fed induction generator, a estimation
and setpoint tracking for rotor’s angular velocity ωrd , b estimation and setpoint tracking for stator’s
magnetic flux ψsd

(a) 80 (b) 5
rotor current i rq of the induction
rotor current i r d of the induction

0
60
−5
generator (A)
generator (A)

40
−10

20 −15

−20
0
−25
−20
−30

−40 −35
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.21 Extended Kalman Filter-based control of the doubly-fed induction generator, a rotor’s
control input current ırd , b rotor’s control input current irq

noticed that the Extended Kalman Filter is an efficient approach for the implementa-
tion of state estimation-based control of the sixth-order induction generator model.
The Unscented Kalman Filter can be also used in place of the Extended Kalman
Filter and in the latter case there will be no need to compute Jacobian matrices.
Finally, it is noted that to compensate for external disturbances and parameter vari-
ations in state estimation-based nonlinear control schemes, the following approaches
are possible:(i) use of adaptive Kalman Filter, (ii) redesign of the Kalman filter in
the form of a disturbance observer, and (iii) redesign of the Kalman Filter in the
form of a high-gain Extended Kalman Filter. Regarding (i), adaptive Kalman Fil-
tering is implemented through the adaptation of the process and measurement noise
covariance matrices in the Kalman Filter recursion so as to cope with variable noise
7.4 Flatness-Based Control of DFIG in Cascading Loops 387

(a) (b)
450 3000
mechanical torque Tm of the induction

electromagnetic torque Te at the rotor

400
2500
350
2000
generator (Nm)

300
250 1500

(Nm)
200 1000
150
500
100
0
50
0 −500
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.22 a Mechanical torque Tm due to wind. b Electromagnetic torque Te at the DFIG rotor

(a) 40 (b) 16
magnetic flux ψs d of the induction
rotation speed ω of the induction

35 14

12
30
generator (rad/sec)

generator (Wb)

10
25
8
20
6
15
4
10
2
5 0

0 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.23 Extended Kalman Filter-based control of the doubly-fed induction generator, a estimation
and setpoint tracking for rotor’s angular velocity ωrd , b estimation, and setpoint tracking for stator’s
magnetic flux ψsd

levels [12]. Regarding (ii), additive disturbances and parametric changes can be
identified with the use of Kalman Filters that operate as disturbance observers (see
work on Kalman Filtering and disturbance observers in [432]). Once the disturbance
affecting the nonlinear system becomes known it can be compensated through the
introduction of an additional control term in the loop. Thus, it can be assured again
that the estimation error x − x̂ will be minimal and the performance of the control loop
will be satisfactory. As previously analyzed, there are several results on disturbance
observers (e.g., Extended State Observers, Unknown Input Observers or Perturba-
tion Observers) and on their use within a Kalman Filter framework [260, 354]. Such
results can be exploited so as to make Kalman Filtering for electric power gener-
ators as well as the associated state estimation-based control loop be more robust
388 7 Differential Flatness Theory and Electric Power Generation

(a) 80 (b) 5

rotor current i rq of the induction

rotor current i rd of the induction

0
60
−5
40

generator (A)
generator (A)

−10

20 −15

−20
0
−25
−20
−30

−40 −35
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.24 Extended Kalman Filter-based control of the doubly-fed induction generator, a rotor’s
control input current ırd , b rotor’s control input current irq

(a) (b)
mechanical torque Tm of the induction

50 3000
electromagnetic torque Te at the

2500
0
2000
generator (Nm)

rotor (Nm)

−50 1500

1000
−100

500
−150
0

−200 −500
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 7.25 a Mechanical torque Tm due to wind. b Electromagnetic torque Te at the DFIG rotor

[432]. Regarding (iii) the high-gain Extended Kalman Filter can provide additional
robustness to state estimation under external disturbances and parametric variations
[38, 50].

7.5 State Estimation-Based Control of Distributed PMSGs

7.5.1 Overview

This section extends the results of Sect. 7.2 about sensorless control of stand-alone
PMSGs toward sensorless control of interconnected PMSGs (permanent magnet
synchronous generators) with the use of the Derivative-free nonlinear Kalman Filter,
7.5 State Estimation-Based Control of Distributed PMSGs 389

in the framework of a multi-area multi-machine model of distributed power gen-

eration units. A large-scale power system consists of multiple control areas where
local power generation units operate, while these areas are connected through tie
lines and power transformers. Due to change in active power loads, the frequen-
cies of the areas variate from their nominal values (synchronous speed) and power
exchange through the tie lines will also deviate from the associated setpoints [182,
441, 595]. Therefore, it is necessary to supply power generator units with local con-
trollers capable of maintaining the generators’ turn speed at its nominal value. To
stabilize the functioning of the distributed power generation system and eliminate
frequency deviations several control approaches have been implemented, such as
standard PID control, lead-lag compensators where the selection of the controller’s
parameters is performed with the use of Kharitonov’s theorem, H∞ control or adap-
tive control [93, 115, 131, 221, 230, 235, 277, 278, 425]. To control efficiently power
generators in case that the complete state vector of the electric machine cannot be
measured directly, observer-based control schemes have been proposed [60, 139,
347, 348]. Moreover, to control distributed power generators, disturbance rejection
control schemes have been developed [318, 319, 379].
First, a sensorless control scheme is developed for the local power generators, as
analyzed in Sect. 7.2. The control approach is based on a diffeomorphism (change
of coordinates) that follows differential flatness theory and enables to transform the
initial nonlinear model of the generator into a linear one. The latter model makes
possible to design a state feedback controller for the PMSG. Moreover, to estimate the
nonmeasurable elements of the PMSG state vector and to use them in the control loop,
the Derivative-free nonlinear Kalman Filter is considered. This filtering approach
exploits the previously mentioned transformation of the generator’s model in the
linear canonical (Brunovsky) form through the application of differential flatness
theory, and performs filtering on the linearized equivalent model of the generator
using the standard Kalman Filter recursion. It will be shown that the sensorless control
scheme for the local power generators can be modified so as to provide the desirable
robustness to the distributed power generation scheme. By redesigning the Kalman
Filter, initially used for estimating the nonmeasurable elements of the generator’s
state vector, in the form of a disturbances estimator it becomes possible to estimate in
real-time the aggregate disturbance that affects each local generator [107, 183, 354].
Knowing the disturbances’ value it is possible to generate a counter-disturbance
term that serves as auxiliary control input and which finally compensates for the
disturbances’ effects. With the previously described control approach each generator
keeps running at its nominal operating conditions and the distributed generation units
become robust to phenomena such as mechanical input torque variations and interarea
oscillations [11, 449, 522, 565, 578].
The proposed control method is suitable for compensating for both endogenous
and exogenous disturbances that affect the generators’ model. Endogenous distur-
bances are associated with parametric uncertainty, e.g., there can be imprecision in
the values of reactances, stator, and rotor circuit coefficients and the rotor’s moment
of inertia. Exogenous disturbances are associated with variations in the mechan-
ical input torque, as well as with perturbations in the excitation system of each
390 7 Differential Flatness Theory and Electric Power Generation

generator which are due to lack of synchronization with other generators connected
to the grid [47]. The efficient compensation of these disturbances enables to maintain
synchronism between the individual power generators, thus assuring: (i) robustness
improvement of the power grid and maintenance of the synchronization between the
distributed electric power generation units even under adverse conditions in the power
grid, (ii) improvement of the provided power quality (distortions in power quality due
to deviations of the power generation units from their nominal operating conditions
will be avoided), (iii) uninterrupted power flow in the grid, ability to incorporate more
power generation units without affecting stability, and consequently better response
to the increased demand for power supply.
The proposed distributed filtering and control method for the system of the inter-
connected power generators is a highly data-driven one. Actually, to succeed state
estimation and disturbances compensation for a set of n distributed PMSGs it is
necessary that at each sampling period each power generator controller receives n
variables associated with the turn angles of all generators in the power grid. Thus,
the total amount of data that has to be processed consists of n 2 variables. This is
a quite intensive computation load which can be managed efficiently thanks to the
distributed control structure that has been implemented (each generator computes
locally its own control signal). Moreover, taking into account that the exchange of
data has to take place in real-time one can conclude that communication links of
efficient bit rate have to be established between the generators.

7.5.2 Dynamic Model of the Distributed Power Generation

Units

7.5.2.1 Dynamics of the Interconnected Power Generators

The dynamic model of the stand-alone PMSGs has been developed in Sect. 7.2.2. The
interconnection between several local power generation units which are described
by the previously analyzed SMIB model results in a multi-area multi-machine dis-
tributed power generation system (Fig. 7.26).
A multi-machine power system (MMPS) with n machines, in which the first
machine is chosen as the reference machine can be described by the following non-
linear dynamic model:

δ̇i = ωi − ω0
Di Pm
ω̇i = − 2Ji (ωi − ω0 ) + ω0 2Jii −
 2   
−ω0 2J1 i [G ii E qi + E qi nj=1, j=i E q j G i j sin(δi − δ j − αi j )] (7.200)

  xd −xd
Ė qi = − 1 E qi + Td1 i  i Vsi cos(Δδi ) + Td1 E f i
Td oi xd oi
i Σi
7.5 State Estimation-Based Control of Distributed PMSGs 391

Fig. 7.26 A multi-machine

(3-area) distributed power
generation model

where the electric torque Pei which is associated with the active power at the ith
generator is now given as

 2 
n

Pei = G ii E qi + E qi E q j G i j sin(δi − δ j − αi j ) (7.201)
j=1, j=i

for i = 1, 2, . . . , n. For a power grid that consists of n generators the aggregate state
vector comprises the state vectors of the local machines, i.e., x = [x 1 , x 2 , . . . , x n ]T ,

where x i = [x1i , x2i , x3i ]T , with x1i = Δδi , x2i = Δωi and x3i = E qi are the state
variables for the ith machine and i = 1, 2, . . . , n.

7.5.3 Lie Algebra-Based Design of a Feedback Controller

for the PMSG

7.5.3.1 Linearization of the Distributed Power Generation System Using

Lie Algebra

The linearization of the sensorless PMSG model has been analyzed in Sect. 7.2.2.
Next, the results are generalized toward linearization of the distributed PMSG model,
again with the use of Lie algebra. As previously analyzed the dynamic model of the
distributed power generation system is defined by Eq. (7.200). The state vector of the
ith local power generator is defined as
392 7 Differential Flatness Theory and Electric Power Generation

xi = [x1i , x2i , x3i ]T = [Δδi , Δωi , E qi ]T (7.202)

The local ith power generation system is described as

ẋ i = f i (x) + g i (x)u i (7.203)

where ⎛ ⎞ ⎛ ⎞
f 1i 0
f i (x) = ⎝ f 2i ⎠ g i (x) = ⎝ 0 ⎠ (7.204)
1
f 3i Tdo i

with

f 1i (x) = x2i
ω0 2 2
f 2i (x) = − 2J x + 2J
D i
i 2
Pm i − ω0 2J1 i {G ii x3i + x3i [G ii x3 i +
 j
i
j
+x3i nj=1, j=i x3 G i j sin(x1i − x1 − ai j )]} (7.205)

xdi −xd
f 3i (x) = − 1
 x3i + 1
Tdo i xdΣ
i
Vsi cos(x1 i ).
Td i
i

Generalizing, the distributed power system is described as

ẋ = f (x) + g(x)u (7.206)

 T
where x = x 1 x 2 . . . x n . Without loss of generality a three-machine distributed
power generation model will be considered next (n = 3). The state-space equation
of the distributed power generation model is
⎛ 1⎞ ⎛ 1 ⎞ ⎛ ⎞
ẋ1 f 1 (x) 0 0 0
⎜ẋ21 ⎟ ⎜ f 21 (x)⎟ ⎜ 0 0 0⎟
⎜ 1⎟ ⎜ 1 ⎟ ⎜ ⎟
⎜ẋ ⎟ ⎜ f (x)⎟ ⎜g1 0 0⎟
⎜ 32 ⎟ ⎜ 32 ⎟ ⎜ ⎟⎛ ⎞
⎜ẋ ⎟ ⎜ f (x)⎟ ⎜ 0 0 0⎟
⎜ 12 ⎟ ⎜ 12 ⎟ ⎜ ⎟ u1
⎜ẋ ⎟ = ⎜ f (x)⎟ + ⎜ 0 0 0⎟ ⎝ ⎠
⎜ 2⎟ ⎜ 2 ⎟ ⎜ ⎟ u2 (7.207)
⎜ẋ 2 ⎟ ⎜ f 2 (x)⎟ ⎜ 0 g2 0⎟
⎜ 3⎟ ⎜ 3 ⎟ ⎜ ⎟ u3
⎜ẋ 3 ⎟ ⎜ f 3 (x)⎟ ⎜ 0 0 0⎟
⎜ 1⎟ ⎜ 1 ⎟ ⎜ ⎟
⎝ẋ 3 ⎠ ⎝ f 3 (x)⎠ ⎝ 0 0 0⎠
2 2
ẋ33 f 33 (x) 0 0 g3

The procedure for linearization with the use of Lie algebra is as follows: The fol-
lowing state vector is defined as x = [x11 , x21 , x31 , x12 , x22 , x32 , x13 , x23 , x33 ]T while the
state vector of the linearized equivalent of the distributed PMSG model is defined
as z ∈ R 9×1 . It holds that z 11 = h 1 (x) = Δδ 1 , z 21 = L f h 1 (x) = x21 and
z 31 = L 2f h 1 (x) = f 21 . In a similar manner one obtains z 12 = h 2 (x) = Δδ 2 , z 22 = x22
and z 32 = f 22 . Moreover, in a similar manner one obtains z 13 = h 3 (x) = Δδ 3 ,
z 22 = L f h 3 (x) = x22 and z 33 = L 2f h 2 (x) = f 23 .
7.5 State Estimation-Based Control of Distributed PMSGs 393

Furthermore, for the vectors defined as z 3 = [z 31 , z 32 , z 33 ], f˜ = [ f 21 , f 22 , f 23 ] and

h̃ = [h 1 , h 2 , h 3 ] it holds that

ż = L 3f h(x) + L g L 2f h(x)u (7.208)

or, equivalently
⎛∂ f1 ⎞ ⎛ ⎞
⎛ 1⎞ 2
f1 +
∂ f 21
+ ···
∂ f 21 ∂ f 21 ∂ f 21
···
∂ f 21 ⎛ ⎞
∂ x2 f 2 ∂ xn f n g1 ∂ x2 g2 ∂ xn g3
ż 3 ⎜ ∂∂ xf 12 ⎟ ⎜ ∂∂ xf 12 ⎟ u1
⎝ż ⎠ = ⎜ 2 f 1 +
2 ∂ f 22
+ ···
∂ f 22 ⎟+⎜ 2 g ∂ f 22
···
∂ f 22 ⎟ ⎝ ⎠
u
3 ⎝ ∂ x1 ∂ x2 f 2 ∂ xn f n ⎠ ⎝ ∂ x1 1 ∂ x2 g2 ∂ xn g3 ⎠ 2
ż 33 ∂ f 23 ∂ f 23 ∂ f2 3 ∂ f 23 ∂ f 23 ∂ f 23 u3
∂ x1 f 1 + ∂ x2 f 2 + ··· ∂ xn f n ∂ x1 g1 ∂ x2 g2 ··· ∂ xn g3
(7.209)
Thus, it holds that
ż 3 = f a + Mu = v (7.210)

Therefore, setting ż 3 = v, where v = f a + Mu the dynamics of the distributed power

generation system for the ith power generator i = 1, 2, 3 becomes:
⎛ i ⎞⎛ ⎞⎛ i ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż i ⎠ ⎝0 0 1⎠ ⎝z i ⎠ + ⎝0⎠ vi (7.211)
2 2
ż 3i 000 z 3i 1

In this manner the initial nonlinear power system is transformed into three decoupled
linear subsystems which are in the canonical Brunovksy form. For each of these
subsystems i = 1, 2, 3 the appropriate control law is

(3) i
vi = z d − k3 (z̈ i − z̈ di ) − k2 (ż i − ż di ) − k1 (z i − z di ) − d̃ i (7.212)

where the feedback control gains k1 , k2 , k3 are chosen such that the associated char-
acteristic polynomial of the linearized system is a Hurwitz one.

7.5.4 Differential Flatness of the Distributed PMSG Model

1. Differential flatness of the distributed power generation model

It will be proven that the multi-machine power generation system is also a differ-
entially flat one. As flat output of the distributed power generation system, con-
sisting of n PMSGs, the following vector is defined as y = [y11 , y12 , . . . , y1n ] or
y = Δδ 1 , Δδ 2 , . . . , Δδ n . For the n-machines power generation system it holds that
x11 = y 1 , x12 = y 2 , x13 = y 3 , . . ., x1n = y n and x21 = Δω1 = ẏ 1 , x22 = Δω2 = ẏ 2 ,
x23 = Δω3 = ẏ 3 , . . ., x2n = Δωn = ẏ n . Moreover, it holds that
394 7 Differential Flatness Theory and Electric Power Generation

Di i ω0
ẋ2i = − 2J x + 2J
i 2
Pm i −
ω0 2  n j
i
j (7.213)
− 2Ji [G ii x3 + x3 j=1, j=i [x3 G i j sin(x1i − x1 − αi j )]
i i

or, using the flat output variables

Di i ω0
ÿ i = − 2J ẏ + 2J Pm i −
ω0 2  n
i
j
i
(7.214)
− 2Ji [G ii x3 + x3 j=1, j=i [x3 G i j sin(y i − y j − αi j )]
i i

The external mechanical torque Pm i is considered to be a piecewise constant vari-

able. For i = 1, 2, . . . , n one obtains n equations of the form of Eq. (7.214), with
unknowns the state variables x3i , i = 1, 2, . . . , n. By solving this system of equations
with respect to x3i , i = 1, 2, . . . , n one arrives at defining the state variables x3i as
functions of the elements of the flat outputs vector y i , i = 1, 2, . . . , n and of their
derivatives. Thus one has x3i = f x3 (y 1 , y 2 , . . . , y n ). Additionally, from the relation

 xd −xd
Ė qi = − T1d E qi + Td1 xi d i Vsi cos(Δδi ) + Td1 E fi and knowing that the state vari-
i oi Σi oi

ables x1i , x2i , x3i , i = 1, 2, . . . , n can be written as functions of the flat output and its
derivatives, one can solve with respect to the control input u i , thus showing that all
control inputs u i , i = 1, 2, . . . , n can be written as functions of the flat output and
its derivatives.
By expressing all elements of the multi-machine power generation system and
the associated control inputs as functions of the flat output and its derivatives, it is
proven that the distributed power generation system stands for a differentially flat
model.
2. The case of unknown time-varying mechanical input torque
Functioning of the power generation system under unknown time-varying external
mechanical torque Pmi is the most generic condition, which also comprises the case
of piecewise constant external input torque. It will be shown how with the use of
the differential flatness theory a controller can be designed for the distributed power
system in case of unknown time-varying external mechanical torque. By deriving
the expression about ÿ i once more with respect to time one obtains

i
y (3) = a i (x) + b1i (x)g1 u 1 + b2i (x)g2 u 2 + b3i (x)g3 u 3 (7.215)

which means that one arrives at a function of the form ż 3i = a i (x) + b1i (x)g1 u 1 +
b2i (x)g2 u 2 + b3i (x)g3 u 3 + d̃ i , where in the case of the distributed power generation
that consists of n = 3 machines, and considering for instance i = 1, j = 2, 3 one has

Di 2 i Di ω0 2 n j j
ai = ( ) x2 + [G ii x3i + x3i x G i j sin(x1i − x1 − αi j )]−
2Ji (2Ji ) 2 j=1, j=i 3

ω0 n j j 1 1 xdi − xdi
− [G ii x3i + x3 G i j sin(x1i − x1 − αi j )(−  x3i + ( Vsi cos(x1i ))]−
2Ji j=1, j=i Td Td oi xd Σi 
i
7.5 State Estimation-Based Control of Distributed PMSGs 395


ω0 i n j 1 1 xdi − xdi
− x G i j sin(x1i − x1 − αi j )(−  x3i + ( Vsi cos(x1i ))−
2Ji 3 j=1, j=i Tdi Td oi xd Σi 
ω0 i n j j ω0 i n j j j
− x x G i j cos(x1i − x1 − αi j )x2i x x G i j cos(x1i − x1 − αi j )x2
2Ji 3 j=1, j=i 3 2Ji 3 j=1, j=i 3

(7.216)

while for functions b1i , b2i and b3i it holds that

ω0 n j j
b1i = − 2Ji
[2G ii x3i + i
j=1, j=i x 3 G i j sin(x 1 − x1 − αi j )] Td1
oi
ω0
b2i = − 2Ji
G i2 sin(x1i − x12 − αi2 ) Td1 (7.217)
o2
ω0
b3i = − 2Ji
G i3 sin(x1i − x13 − αi3 ) Td1
o3

Finally for the disturbance term one has d̃ i = − D2Ji ω20 Pmi + 2J
ω0 i
Ṗ
i m
i
Thus, one has the following description of the dynamics of the ith power generator

ż 1i = z 2i
ż 2i = z 3i (7.218)
ż 3i = a i (x) + b1 i g1 u 1 + b2 i g2 u 2 + b3 i g3 u 3 + d̃ i

For the complete system of the 3 generators one has

ż 31 = a 1 (x) + b1 1 g1 u 1 + b2 1 g2 u 2 + b3 1 g3 u 3 + d̃ 1
ż 32 = a 2 (x) + b1 2 g1 u 1 + b2 2 g2 u 2 + b3 2 g3 u 3 + d̃ 2 (7.219)
ż 33 = a 3 (x) + b1 3 g1 u 1 + b2 3 g2 u 2 + b3 3 g3 u 3 + d̃ 1

or in matrix form
ż 3 = f a (x) + Mu + d̃ (7.220)

where z 3 = [z 31 , z 32 , z 33 ]T , u = [u 1 , u 2 , u 3 ]T and d̃ = [d̃1 , d̃2 , d̃3 ]T while

⎛ 1 ⎞ ⎛ 1 ⎞
a (x) b1 g1 b2 1 g2 b3 1 g3
f a (x) = ⎝a 2 (x)⎠ , M = ⎝b1 2 g1 b2 2 g2 b3 2 g3 ⎠ (7.221)
a 3 (x) b1 3 g1 b2 3 g2 b3 3 g3

Setting v = f a (x) + Mu + d̃, one obtains again the linear canonical form for the ith
generator given by
⎛ i⎞ ⎛ ⎞⎛ i ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż i ⎠ = ⎝0 0 1⎠ ⎝z i ⎠ + ⎝0⎠ (vi + d̃ i ) (7.222)
2 2
ż 3i 000 z 3i 1
396 7 Differential Flatness Theory and Electric Power Generation

In this manner the initial nonlinear power system is transformed into three decoupled
linear subsystems which are in the canonical Brunovksy form. For each of these
subsystems the appropriate control law is

(3) i
vi = z d − k3 (z̈ i − z̈ di ) − k2 (ż i − ż di ) − k1 (z i − z di ) − d̃ i (7.223)

Since the disturbance term d̃ i , which is due to the time-varying mechanical input
torque at the ith generator is unknown, in the computation of the control input of
Eq. (7.223) it will be substituted by its estimate d̃ˆi , which will be provided by a
disturbance observer. For the implementation of the distributed power generation
control scheme, the controller at the ith power generator makes use of not only its
own state vector X i = [x1i , x2i , x3i ]T , but also of the state vectors of the rest n−1 power
j j j
generators, i.e., x j = [x1 , x2 , x3 ], j=i, j = 1, 2, . . . , n. The transformation of the
dynamical model of the local power generators into the linear canonical form enables
to obtain estimates of the transformed state vector of the system Y i = [y i , ẏ i , ÿ i ]T
where y i = Δδ i , through Kalman Filtering, after processing measurements of only
the turn angle difference Δδ i of the ith power generator. Therefore, one has to
compute the estimates X̂ i = [x̂1i , x̂2i , x̂3i ]T after using the estimate provided by the
Kalman Filter Ŷ i = [ ŷ i , ẏˆ i , ÿˆ i ]T . As already explained, this is a highly data-driven
application and requires the exchange and processing in real-time of a large amount
of data. It holds that
x̂1i = ŷ i x̂2i = ẏˆ i (7.224)

while for the computation of x̂3i for i = 1, 2, . . . , n one has to solve with respect to
x̂3i the system of equations

ω0 ω0
ÿˆ 1 = − 2J
D1 ˆ 1 2
ẏ + 2J Pm 1 − 2J [G 11 x31 +
 1
j
1 1
+x̂31 nj=1, j=1 [x̂3 G 1 j sin(y 1 − y j − α1 j )]
ω0 ω0
ÿˆ 2 = − 2J
Di ˆ 2 2
ẏ + 2J Pm 2 − 2J [G 22 x32 +
 2
j
2 2
+x32 nj=1, j=2 [x̂3 G 2 j sin(y 2 − y j − α2 j )] (7.225)
···
ω0 ω0
ÿˆ n = − 2J
Dn ˆ n
ẏ + n2
2Jn Pm n − 2Jn [G nn x 3 +
 n
j
+x̂3n nj=1, j=n [x̂3 G n j sin(y i − y j − αi j )]

The computation of state estimates for the initial MIMO nonlinear system of the inter-
connected power generators, out of the estimates obtained for its linearized equiva-
lent, can be also formulated as an optimization problem. To obtain state estimates for
the initial system one comes against a set of coupled nonlinear equations described
in Eq. (7.225). The latter can be solved with respect to the state vector elements of
the initial system using ordinary computation tools for optimization and nonlinear
programming [305, 370]. Moreover, by generalizing the results of Sect. 7.2.2 to the
distributed MIMO case of the interconnected electric power generators, it is possible
7.5 State Estimation-Based Control of Distributed PMSGs 397

to compute the state vector estimate for the initial system, through a filter of the
form described in Eq. (7.24) that uses the inverse Jacobian matrix of the linearization
transformation in its observation gain.
The proposed control scheme considers as output of the synchronous generator
the rotor’s difference angle Δδ. However, it is possible to avoid the use of encoder
readings about the rotor’s turn angle, and to indirectly estimate this parameter through
the processing of measurement coming from PMUs. Such measurements are the bus
voltage magnitude and the associated phase angle, the line current magnitude and
the associated angle and the electrical output power at the terminal bus [29, 120, 144,
161, 211, 368].

7.5.4.1 State Estimation with the Derivative-Free Nonlinear Kalman

Filter

It was shown that using differential flatness theory the nonlinear model of each PMSG
can be written in the canonical form of Eqs. (7.20) and (7.53). Thus one has the linear
model
ẏ f = A f y f + B f v
(7.226)
z f = Cf yf

where y f = [y1 , y2 , y3 ]T and matrices A f , B f , C f are in the canonical form

⎛ ⎞ ⎛ ⎞ ⎛ ⎞
010 0 1
A f = ⎝0 0 1⎠ B f = ⎝0⎠ C Tf = ⎝0⎠ (7.227)
000 1 0

and the measurable variable y1 = δ is associated with the turn angle of the rotor. After
carrying out discretization of matrices A f , B f , and C f with common discretization
methods one can perform linear Kalman filtering. This is the previously analyzed
Derivative-free nonlinear Kalman filtering [427, 434]. The estimation of disturbance
inputs for the individual power generators can be performed with the redesign of the
Derivative-free nonlinear Kalman Filter as a disturbance observer, as explained in
Sect. 7.2.2.
To compensate for the effects of the disturbance terms it suffices to use in the
control loop the modified control input, which actually compensates for the effects
of the external disturbance term Tm . This gives v∗ = v − T̂m or v∗ = v − ẑ 4 .

7.5.5 Simulation Tests

The procedure for estimating the disturbance inputs that affect each individual power
generator has been explained in Sect. 7.2.2. This makes use of a Kalman Filter-based
disturbance observer for each individual generator. To evaluate the performance of
398 7 Differential Flatness Theory and Electric Power Generation

the proposed nonlinear control scheme that uses Kalman Filtering in the estimation
of the nonmeasurable state vector elements of the distributed PMSGs and the exter-
nal disturbances, simulation experiments have been carried out. Different rotation
speed setpoints had been assumed. Moreover, different input torques (mechanical
input power profiles) have been assumed to affect the dynamic model of each local
PMSG. Indicative ratings of the PMSG model are as follows: direct axis synchro-

nous reactance xd = 2.1 p.u., quadrature axis synchronous reactance xd = 0.4 p.u.,
infinite bus voltage Vs = 1.0 p.u., direct axis open circuit time constant Tdo = 5s,
rotor’s moment of inertia J = 50 kgm2 . The nonlinear control scheme works effi-
ciently even for distributed synchronous generators with uneven ratings.
The case of distributed PMSG operation under unknown input power (torque)
has been examined. The input power at each local generator was considered to
be a disturbance input to the PMSG model and it was assumed that its change in
(3)
time was defined by the third derivative of the associated variable, i.e., Tm where
ω0
Tm = −ω0 (2J )2 Pm + 2J Ṗm . The disturbance dynamics was completely unknown
D

to the controller and its identification was performed in real-time by the disturbance
estimator. It is shown that the Derivative-free nonlinear Kalman Filter redesigned as
a disturbance observer is capable of estimating simultaneously the nonmeasurable

state variables of the generator (ω and E q ), as well as the unknown disturbance input
Tm . A first set of results is concerned with the tracking performance of the control
loop in case of piecewise constant disturbance input. The real value of the state vari-
able of the generator is plotted in blue, the estimated values are given in green, while
the associated setpoint is printed in red. This is shown in Figs. 7.27a, 7.28a and 7.29a.
The estimation of the piecewise constant disturbance input for the first generator is

(a) Generator 1 (b) Generator 1

0.15 0.35
Estimation of input torque Tm (p.u.)

0.3
0.1
0.25
Δω (p.u.)

0.05
0.2

0.15
0

0.1
−0.05
0.05

−0.1 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.27 Sensorless control of the PMSG No 1 under nonmeasurable (piecewise constant) mechan-
ical input torque in case of speed reference setpoint 1: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm1
7.5 State Estimation-Based Control of Distributed PMSGs 399

(a) Generator 2 (b) Generator 2

0.12 0.25

Estimation of input torque Tm (p.u.)

0.1
0.2
0.08

0.06
Δω (p.u.)

0.15
0.04

0.02 0.1

0
0.05
−0.02

−0.04 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.28 Sensorless control of PMSG No 2 under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 1: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm2

shown in Fig. 7.27b, for the second generator it is given in Fig. 7.28b, while for the
third generator it is shown in Fig. 7.29b.
Additionally, results of the tracking performance of the control loop in case of
a second piecewise constant disturbance input are shown Figs. 7.28a, 7.30a, 7.32a.
The estimation of the piecewise constant disturbance input for the first generator is
shown in Fig. 7.30b, for the second generator is given in Fig. 7.31b, while for the

(a) Generator 3 (b) Generator 3

0.25 0.7
Estimation of input torque Tm (p.u.)

0.2
0.6
0.15
0.5
0.1
Δω (p.u.)

0.05 0.4

0 0.3
−0.05
0.2
−0.1
0.1
−0.15

−0.2 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.29 Sensorless control of PMSG No 3 under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 1: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm3
400 7 Differential Flatness Theory and Electric Power Generation

(a) 0.25 Generator 1 (b) Generator 1

0.5

Estimation of input torque Tm (p.u.)

0.2
0.4
0.15

0.3
Δω (p.u.)

0.1

0.05
0.2

0
0.1
−0.05

−0.1 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.30 Sensorless control of PMSG No 1 under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 2: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm1

(a) Generator 2 (b) Generator 2

0.2 0.35
Estimation of input torque Tm (p.u.)

0.3
0.15
0.25
Δω (p.u.)

0.1
0.2

0.05 0.15

0.1
0
0.05

−0.05 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.31 Sensorless control of PMSG No 2 under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 2: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm2

third generator it is shown in Fig. 7.32b. The units of the PMSG state variables have
been expressed again in p.u. (per unit system).
The simulation experiments confirm that the proposed state estimation-based con-
trol scheme not only enables implementation of distributed PMSG control through
the measurement of a small number of variables (e.g., of only the rotor’s turn angle)
but also improves the robustness of the PMSG control loop in case of varying speed
setpoints and varying mechanical input torque.
7.5 State Estimation-Based Control of Distributed PMSGs 401

(a) 0.4
Generator 3 (b) 0.8
Generator 3

Estimation of input torque Tm (p.u.)

0.3 0.7

0.6
0.2
0.5
Δω (p.u.)

0.1
0.4
0
0.3
−0.1
0.2
−0.2 0.1

−0.3 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 7.32 Sensorless control of PMSG No 3 under nonmeasurable (piecewise constant) mechanical
input torque in case of speed reference setpoint 2: a convergence of the real and estimated values
of the angular speed difference Δω, b estimation of the external mechanical input torque Tm3

Table 7.2 RMSE for nonlinear filters

DKFa EKFa DKFb EKFb
Δδ 0.0006 0.0007 0.0011 0.0012
Δω 0.0061 0.0093 0.0069 0.0119

Eq 0.0167 0.1654 0.0376 0.2253
Tm 0.0172 0.0843 0.0184 0.0810
Ṫm 0.0137 0.0374 0.0146 0.0339
T̈m 0.0054 0.0083 0.0057 0.0080

Comparison to the Extended Kalman Filter, which is the most commonly used
nonlinear estimation technique, has shown that the Derivative-free nonlinear Kalman
Filter succeeds estimation of smaller variance for both the nonmeasurable state vector
elements of the PMSG and the term Tm expressing external torques and disturbances
inputs. Estimation with the Extended Kalman Filter was based on local linearization
of the PMSG model and on the computation of the Jacobian of Eq. (13.39). The
obtained results are summarized in Table 7.2, and correspond to experiments (a) and
(b) associated with the two turn speed setpoints given in Figs. 7.27, 7.28, 7.29, 7.30,
7.31 and 7.32 under sinusoidal variations of Tm , and considering different noise
levels.
The proposed control method is not constrained by the number of power generators
and can be applied to a power grid that comprises N interconnected power generators.
Variations in load and variations in the flow of active and reactive power in the grid
can affect the excitation of the individual power generators and can cause deviation
from the synchronous rotation speed. Such effects are modeled as disturbance terms
402 7 Differential Flatness Theory and Electric Power Generation

that affect the dynamic model of a power generator. Therefore, this approach covers
the cases of perturbations associated with power electronics in the grid (e.g., con-
verters or rectifiers of the thyristor-type) or variations of loads connected to the grid
(e.g., motor loads and any type of power consuming devices).
The proposed control method can also be applied to distributed induction gen-
erators, e.g., Doubly-fed induction generators (DFIGs). It is possible to perform
input–output linearization on the dynamic model of the DFIGs with the use of the
previously analyzed Lie algebra-based approach or with the use of the differential
flatness theory-based approach. Next, state and disturbances estimation can be per-
formed on the DFIGs transformed model using the Derivative-free nonlinear Kalman
Filter [65, 70, 95, 125].
The proposed filtering method that is used for joint state and disturbances esti-
mation in the interconnected PMSGs model exhibits specific advantages: (i) it is not
based on local linearization of the PMSGs nonlinear dynamics but on exact transfor-
mation of the generators’ dynamic model to a linear equivalent, (ii) it is not based
on a model of the generators that results from truncation of higher order Taylor
expansion terms (unlike for instance to the Extended Kalman Filter), thus preserving
the accuracy and robustness of the performed estimation, (iii) it does not require
the computation of Jacobian matrices. The Derivative-free nonlinear Kalman Filter
appears to be faster than other nonlinear filtering methods while also providing very
accurate (in terms of variance) state estimates [434, 437].
The compensation of communication delays is significant for implementing con-
trol of the distributed power generators system over a communication network. It has
been shown that the Derivative-free nonlinear Kalman Filter can be also the basis for
eliminating the effects of data transfer delays and data losses in networked control
schemes [450]. Actually, this requires redesign of the Kalman Filter recursion with
the inclusion of a modified Riccati equation for the computation of the covariance
matrix of the state vector estimation error, as well as the inclusion of a smoothing
term in the update of the state vector estimate.
Chapter 8
Differential Flatness Theory for Electric
Motors and Actuators

8.1 Introduction

In this chapter, differential flatness theory is used to develop adaptive fuzzy control
for electric motors and actuators of unknown model, or of model characterized by
nonmeasurable state variables and parametric uncertainty. First, the problem of adap-
tive fuzzy control for DC electric motors is studied. The considered electric motors
can be written in the Brunovsky (canonical) form after a transformation of their state
variables and control input. The resulting control signal is shown to consist of non-
linear elements, which in case of unknown system parameters can be approximated
using neuro-fuzzy networks. An adaptation law for the neuro-fuzzy approximators
can be computed using Lyapunov stability analysis. It is shown that the proposed
adaptation law assures stability of the closed loop. First, a nonlinear DC motor model
is used to evaluate the performance of the proposed flatness-based adaptive control
scheme.
Additionally, in this chapter it is shown that the complete 6th order model of the
induction motor satisfies differential flatness properties since all its state variables
and control inputs can be expressed as functions of the flat outputs. The flat outputs
are chosen to be the rotor’s turn angle and orientation angle of the magnetic flux.
This type of flatness-based control for the induction motor model is implemented in
cascading loops. Moreover, nonlinear Kalman Filtering methods, such the Extended
and the Unscented Kalman Filter are included in this control scheme to estimate the
state vector of the asynchronous motor using a limited number of sensors, such as
the ones measuring stator currents. In the latter case, control of the induction motor
is implemented through feedback of the estimated state vector. The efficiency of the
Kalman Filter-based control scheme is confirmed by simulation experiments.
Finally, the chapter proposes an adaptive fuzzy approach to the problem of con-
trol of electrostatically actuated MEMS (microelectromechanical systems), which is
based on differential flatness theory and which uses exclusively output feedback. It
is shown that the model of the electrostatically actuated MEMS is a differentially

© Springer International Publishing Switzerland 2015 403

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_8
404 8 Differential Flatness Theory for Electric Motors and Actuators

flat one and this permits to transform it to the so-called linear canonical form. For
the new description of the system’s dynamics the transformed control inputs contain
unknown terms which depend on the system’s parameters. To identify these terms
an adaptive fuzzy approximator is used in the control loop. Thus an indirect adaptive
control scheme is implemented in which the unknown or unmodeled system dynam-
ics is approximated by neuro-fuzzy networks and next this information is used by
a feedback controller that makes the electrostatically activated MEMS converge to
the desirable motion setpoints. This adaptive control scheme is exclusively imple-
mented with the use of output feedback, while the state vector elements which are
not directly measured are estimated with the use of a state observer that operates in
the control loop. The learning rate of the adaptive fuzzy system is suitably computed
from Lyapunov analysis, so as to assure that both the learning procedure for the
unknown system’s parameters, the dynamics of the observer and the dynamics of
the control loop will remain stable. The Lyapunov stability analysis depends on two
Riccati equations, one associated with the feedback controller and one associated
with the state observer. Finally, it is proven that for the control scheme that com-
prises the feedback controller, the state observer and the neuro-fuzzy approximator,
H-infinity tracking performance can be succeeded. The functioning of the control
loop has been evaluated through simulation experiments.

8.2 Flatness-Based Adaptive Control of DC Motors

8.2.1 Overview

This section is particularly concerned with differentially flat single-input dynami-

cal systems which can be written in the Brunovksy (canonical) form. As shown in
Chap. 2, according to the Lie-Backlünd condition for equivalence of differentially
flat systems, if a system is differentially flat then it can be transformed to the linear
canonical (Brunovsky) form. In particular, transformation into the Brunovksy form
can be succeeded for systems that admit static feedback linearization (i.e., a change
of coordinates for both the system state variables and the system’s control input).
Single-input differentially flat systems admit static feedback linearization and there-
fore can be finally written in the Brunovsky form [340]. Moreover, flatness-based
adaptive fuzzy control can be applied to multi-input dynamical systems. For MIMO
dynamical systems which are differentially flat and which admit static feedback
linearization, transformation to the canonical (Brunovsky) form can be performed.
Additionally, even for MIMO dynamical systems which are differentially flat and
admit only dynamic feedback linearization, one can succeed transformation to the
canonical (Brunovsky) form. Therefore, there exists a wide class of nonlinear dynam-
ical systems to which the proposed flatness-based adaptive fuzzy control method can
be applied [426].
8.2 Flatness-Based Adaptive Control of DC Motors 405

After transformation to the linear canonical form, the resulting control input is
shown to contain nonlinear elements which depend on the system’s parameters. If
the parameters of the system are unknown, then the nonlinear terms which appear in
the control signal can be approximated with the use of neuro-fuzzy networks. In this
chapter it is shown that a suitable learning law can be defined for the aforementioned
neuro-fuzzy approximators so as to preserve the closed-loop system stability. Lya-
punov stability analysis proves also that the proposed flatness-based adaptive fuzzy
control scheme results in H∞ tracking performance, in accordance to the results of
[407, 410, 413, 433].
Unlike other adaptive fuzzy control schemes which are based on several assump-
tions about the structure of the nonlinear system as well as about the uncertainty
characterizing the system’s model, the proposed adaptive fuzzy control scheme based
on differential flatness theory offers an exact solution to the design of adaptive con-
trollers for unknown dynamical systems. The only assumption needed for the design
of the controller and for succeeding H∞ tracking performance for the control loop
is that there exists a solution for a Riccati equation associated to the linearized error
dynamics of the differentially flat model. This assumption is quite reasonable for
several nonlinear systems (including electric motors and actuators), thus providing a
systematic approach to the design of reliable controllers for such systems [426, 433].

8.2.2 Dynamics and Linearization of the DC Motor Model

The control approach to be followed in this section is similar to the one analyzed in
Chap. 3. The dynamic model of the nonlinear DC motor has been already presented
in Chap. 4. As explained, the dynamical model of the DC-motor model can be written
as an affine in the input system: ẋ = f (x, t)+g(x, t)u, with ẋ denoting the derivative
of the motor’s state vector, x = [x1 , x2 , x3 ]T = [θ, θ̇ , i α ] (θ is the position of the
motor, θ̇ is the angular velocity of the motor and i α is the armature current) [202,
539]. Functions f (x) and g(x) are vector field functions defined as:
⎛ ⎞ ⎛ ⎞
x2 0
f (x) = ⎝k1 x2 + k2 x3 + k3 x32 + k4 T1 ⎠ , g(x) = ⎝ 0 ⎠ (8.1)
k5 x 2 + k6 x 2 x 3 + k7 x 3 k8

where k1 = −F/J , k2 = A/J , k3 = B/J , k4 = −1/J , k5 = −A/L, k6 = −B/L,

k7 = −R/L, k8 = −1/L, R and L are the armature resistance and induction
respectively, and J is the rotor’s inertia, while F is the friction. Variable A is a
constant defining the torque due to the armature’s current, while variable B is a
constant associated to the armature’s reaction. Now, the state-space equation of the
DC motor becomes
ẋ1 = x2
ẋ2 = k1 x2 + k2 x3 + k3 x32 + k4 T1 (8.2)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u
406 8 Differential Flatness Theory for Electric Motors and Actuators

where T1 the load torque and u is the terminal voltage. From the second row of
Eq. (13.2) one obtains,

ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 x3 ẋ3 ⇒

(8.3)
ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 k5 x2 x3 + 2k3 k6 x2 x32 + 2k3 k7 x32 + 2k3 k8 x3 u

Thus, one has ẍ2 = f¯(x) + ḡ(x)u where

f¯(x) = k1 ẋ2 + k2 ẋ3 + 2k3 k5 x2 x3 + 2k3 k6 x2 x32 + 2k3 k7 x32 , and

(8.4)
ḡ(x) = 2k3 k8 x3

For the considered nonlinear electric motor model described in Eq. (8.4) it is assured
that inherently ḡ(x) = 0, therefore again singularities are not going to appear in the
control law. Moreover, assuming the effects of friction k1 x2 and of the load torque
k4 T1 as external disturbances, the nonlinear DC motor model of Eq. (13.1) becomes

ẋ1 = x2
ẋ2 = k2 x3 + k3 x32 (8.5)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u

Selecting the flat output to be y = x1 one can see that all state variables xi , i = 1, 2, 3
and the control input u can be expressed as functions of the flat output and its
derivatives. Indeed it holds
x1 = y
x2 = ẏ
(8.6)
−k2 + |k22 +4k3 ÿ|
x3 = 2k3

with control input

u= ḡ(x) [y
1 (3) − f¯(y, ẏ, ÿ)]. (8.7)

The aforementioned system of Eq. (13.4) can be written in the Brunovsky form:
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (8.8)
ẏ3 000 y3 1

where y1 = y and v = f¯(x, t) + ḡ(x, t)u.

With functions f¯(x, t) and ḡ(x, t) to be given by Eq. (8.6). The stability analysis
of the adaptive fuzzy control scheme follows the stages presented in Chap. 3, for the
case of single-input dynamical systems. Using a control input as in Eq. (8.7) it was
possible to make the electric motor’s angle track any desirable setpoint. Regarding
the implementation of the flatness-based adaptive fuzzy controller, the neuro-fuzzy
approximators for functions f¯(x, t) and ḡ(x, t) have now three inputs, i.e., x, ẋ
8.2 Flatness-Based Adaptive Control of DC Motors 407

and ẍ. Taking that each fuzzy input variable consists of 3 fuzzy sets, there are now
27 fuzzy rules of the form:

R l : IF x is Al1 AND ẋ is Al2 AND ẍ is Al3 THEN fˆl is bl (8.9)

In the simulation experiments, it was assumed that at the beginning of the second
half of the simulation time an additive sinusoidal disturbance of amplitude A = 2.0
and period T = 7.5 s affected the DC motor. The position and velocity variations for
a sinusoidal setpoint are depicted in Fig. 8.1a, b, respectively. The performance of
the proposed flatness-based adaptive fuzzy control was also tested in the tracking of
a seesaw setpoint. The associated position and velocity variation are demonstrated
in Fig. 8.2a, b, respectively. The control signal in the case of tracking of a sinusoidal
setpoint is shown in Fig. 8.3a, while the control signal when tracking a seesaw setpoint
is shown in Fig. 8.3b. Finally, the approximation of function g(x, t) in the case of
tracking of a sinusoidal setpoint is shown in Fig. 8.4a (and is marked as a dashed
line), while when tracking a seesaw setpoint the approximated function g(x, t) is
shown in Fig. 8.4b.
As already explained in Chap. 3, and comparing the proposed flatness-based adap-
tive fuzzy control method to other adaptive fuzzy control approaches and to the analy-
sis on neural adaptive control methods given in the relevant bibliography (e.g., [167,
168, 205]), the following can be noted: (i) the transformation of the initial nonlinear
system into the linearized Brunovksy (canonical) form does not require the compu-
tation of partial derivatives or Lie derivatives, (ii) there is no need to make restrictive
assumptions about the number of truncated higher order terms in the linearization
of the system’s nonlinear model or about a bounded error in the linearization of the
output of the neural/fuzzy approximators, (iii) the number of adaptable parameters
that is involved in the training of the neural/fuzzy approximators remains small and

(a) 1.5 (b) 1.5

1 1
position x 1 of the motor

velocity x 2 of the motor

0.5 0.5

0 0

−0.5 −0.5

−1 −1
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 8.1 a Tracking of a sinusoidal position setpoint (red line) by the angle of the motor. b Tracking
of a sinusoidal velocity setpoint (red line) by the angular velocity of the motor
408 8 Differential Flatness Theory for Electric Motors and Actuators

(a) 1.2 (b) 1.5

1
1
position x of the motor

velocity x of the motor

0.8
0.5
0.6

0.4 0

2
1

0.2
−0.5
0
−1
−0.2

−0.4 −1.5
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 8.2 a Tracking of a seesaw position setpoint (red line) by the angle of the motor. b Tracking
of a seesaw velocity setpoint (red line) by the angular velocity of the motor (continuous line)

(a) 100 (b) 200

80
150
60
control signal of the motor
control signal of the motor

100
40
50
20

0 0

−20
−50
−40
−100
−60
−150
−80

−100 −200
0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 8.3 a Control input of the motor when tracking a sinusoidal setpoint. b Control input of the
motor when tracking a seesaw setpoint

there is no need to use an excessive number of neural/fuzzy approximators to pro-

duce the control signal, (iv) the tracking performance of the neuro-fuzzy control
loop is evaluated with the use of specific metrics (e.g., H∞ tracking performance),
(v) the proposed flatness-based control method can be also extended to MIMO actu-
ation systems, without constraining assumptions about their dynamics and structure
(e.g. triangular, affine-in-the-input, of canonical form, etc.) [426, 433].
The problem of avoidance of singularities in the proposed control scheme has been
already discussed in Chap. 3. Flatness-based adaptive fuzzy control assures stability
of the control loop and the asymptotic elimination of the tracking error. Therefore,
it can be ensured that θ̃g = θg − θg∗ →0 which means that ĝ(x)→g(x). Provided
that function g(x) does not become zero as long as x remains bounded, then ĝ(x)
8.2 Flatness-Based Adaptive Control of DC Motors 409

(a)
real vs estimated function g 1 (b) 1

real vs estimated function g

0.5 0.5

0 0

−0.5 −0.5

−1 −1

0 5 10 15 20 25 30 0 5 10 15 20 25 30
t (sec) t (sec)

Fig. 8.4 a Approximation of function ḡ(x, t) of the motor model when tracking a sinusoidal
setpoint. b Approximation of function ḡ(x, t) of the motor model when tracking a seesaw setpoint

will be also different than zero. For the cart-pole system described in Eq. (3.39) and
Eq. (3.40) it holds that g(x) = 0 for x1 = θ = 0, however, for x1 ∈(0, π ) this case is
not going to occur, even if ideally there is no function approximation error and ĝ(x)
coincides with g(x). For the considered nonlinear electric motor model described
in Eq. (8.4) it is assured that inherently ḡ(x) = 0, therefore again singularities are
not going to appear in the control law. In the generic case, to assure the avoidance
of singularities in the proposed control scheme one has to exclude singularity points
from the reference trajectory that the system’s state vector has to track.
Singularities may appear not only in the proposed adaptive fuzzy control scheme
but in all control systems which are based on static feedback linearization. For exam-
ple, the linearization of the system through the use of a new control input of the
form v = f (x) + g(x)u means that u = g(x)−1 [v − f (x)] which does exclude the
appearance of singularities. Therefore, singularities do not concern only the proposed
control method but the whole class of static feedback-based linearization schemes.
As explained in Chap. 2, some modifications can be introduced in the design of the
controller to prohibit the appearance of singularities, for example, a change in coor-
dinates that results in a new state-space representation which does not include any
points of singularity [340].

8.3 Flatness-Based Control of Induction Motors

in Cascading Loops

8.3.1 Overview

This section analyzes sensorless control of induction motors with the use of control
methods which are based on differential flatness theory. Induction motors are cur-
rently a main element of several industrial systems, as well as of motion transmission
410 8 Differential Flatness Theory for Electric Motors and Actuators

and transportation systems. The possibility to reduce the number of sensors involved
in the control of induction motors has been a subject of systematic research during
the last years [49, 100, 170, 199, 200, 280, 296, 336, 432, 538]. As a result, state
estimation-based control has become an active research area in the field of electric
machines and power electronics. Elimination of the speed and magnetic flux sensors
has the advantages of lower cost, ruggedness as well as increased reliability. Nonlin-
ear Kalman Filtering can be used to obtain accurate estimates of the induction motor’s
state vector through the processing of measurements coming from a small number
of sensors, e.g., control input currents applied to stator. A well-established nonlinear
Kalman Filtering approach is the Extended Kalman Filter (EKF), which is based on a
linearization of the nonlinear dynamics using a first-order Taylor expansion [31, 229,
405, 408]. Alternatively, the Unscented Kalman Filter (UKF) can be considered. The
Unscented Kalman Filter is a derivative-free state estimation method of high accu-
racy. The state distribution in UKF is approximated by a Gaussian random variable,
which is represented using a minimal set of suitably chosen weighted sample points.
These sigma points are propagated through the true nonlinear system, thus gener-
ating the posterior sigma-point set, and the posterior statistics are calculated. The
sample points progressively converge to the true mean and covariance of the Gaussian
random variable [418, 419]. The use of the Unscented Kalman Filter for state estima-
tion and control of nonlinear electric motor models is a relatively new and promising
topic. Indicative results on the use of the UKF for sensorless control of induction
motors and fault diagnosis of electric drives can be found in [4, 5, 6, 233, 256].
In this section, a sensorless control scheme for induction motors is developed
consisting of (i) a nonlinear Kalman Filter, such as the Extended or the Unscented
Kalman Filter, which provides estimates of the complete 6th order state vector of
the induction motor, after sequential processing of measurements from a limited
number of sensors (as the ones measuring stator currents), (ii) a nonlinear controller
that is based on the principles of the differential flatness theory, which unlike the
conventional field-oriented control approach makes no assumption about decoupling
between the rotor’s magnetic flux and the rotor’s angular speed. The performance
of the Extended Kalman Filter-based sensorless control scheme is tested through
simulation experiments and compared to an Unscented Kalman Filter-based control
loop. It is shown that both the EKF- and the UKF-based control result in fast and
accurate trajectory tracking.

8.3.2 A Cascading Loops Scheme for Control

of Field-Oriented Induction Motors

8.3.2.1 Field-Oriented Induction Motor Model

As in the case of asynchronous generators that was analyzed in Chap. 7 to derive the
dynamic model of an induction motor, the three-phase variables are first transformed
8.3 Flatness-Based Control of Induction Motors in Cascading Loops 411

Fig. 8.5 Schematic diagram of the proposed flatness-based control scheme with the use of nonlinear
Kalman filtering

to two-phase ones. This two-phase system can be described in the stator-coordinates

frame α − b and the associated voltages are denoted as vsα and vsb , while the currents
of the stator are i sα and i sb , and the components of the rotor’s magnetic flux are ψrα
and ψrb (Fig. 8.5). Then, the rotation angle of the rotor with respect to the stator is
denoted by δ, and a rotating reference frame d − q on rotor, is defined [432].
The state vector of the motor is x = [θ, ω, ψrα , ψrb , i sα , i sb ] (where θ stands for
the turn angle of the rotor and ω for the rotation speed), while the dynamic model of
the induction motor is written as [62, 412, 434]:

ẋ = f (x) + gα (x)vsα + gb (x)vsb + w(t)

(8.10)
z = h(x) + v(t)

with the first row to describe the state equation of the motor and the second row to
describe the measurement equation of the motor (where h(x) is a nonlinear vector
field of x). The elements of the induction motor’s dynamic model are:
⎛ ⎞
x2
⎜ μ1 (x3 x6 − x4 x5 ) − TL ⎟
⎜ J ⎟
⎜ −α1 x3 − n p x2 x4 + α1 M x5 ⎟
⎜
f (x) = ⎜ ⎟ (8.11)
⎟
⎜ n p x2 x3 − α1 x4 + α1 M x6 ⎟
⎝ α1 β1 x3 + n p β1 x2 x4 − γ1 x5 ⎠
−n p β1 x2 x3 + α1 β1 x4 − γ1 x6

gα = [0, 0, 0, 0, σ 1L s , 0]T gb = [0, 0, 0, 0, 0, σ 1L s ]T (8.12)

412 8 Differential Flatness Theory for Electric Motors and Actuators

where J is the rotor’s inertia, and TL is the external load torque. The rest of the model
2
parameters are σ = 1 − M 2 /L s L r , α1 = LRrr , β1 = σ LMs L r , γ1 = ( σML RLr2 + σRLss ),
s r
n M
μ1 = JpL r , where L s , L r are the stator and rotor auto-inductances, M is the mutual
inductance and n p is the number of poles.
The process noise w(k) given in Eq. (8.10) is due to model inaccuracies associ-
ated with random variations of the model’s parameters. For example, resistances,
inductances, and magnetic permeability of the electric motor can exhibit a stochastic
variation round a nominal value. The measurement noise v(k) given in Eq. (8.10) is
due to stochastic variations of the elements of the measuring devices. If the effects
of the noise signals are not compensated by a filtering procedure, the performance
of the control loop can be unsatisfactory or even the stability of the control loop can
be risked. In the sensorless control scheme of the induction motor studied in this
section, the measured variables are considered to be the a-b reference frame currents
of the stator.

8.3.2.2 Decoupling of Speed-Flux Dynamics

The classical method for induction motors control is based on a transformation of

the stator’s currents (i sα and i sb ) and of the magnetic fluxes of the rotor (ψrα and
ψrb ) to the reference frame d − q which rotates together with the rotor. In the d − q
frame there will be only one nonzero component of the magnetic flux ψrd , while the
component of the flux along the q axis equals 0. The new control inputs of the system
are considered to be vsd , vsq , and are associated to the d − q frame voltages vd and
vq , respectively. The control inputs vsd , vsq are connected to vsα , vsb of Eq. (8.10),
according to the relation


−1

vs α ψr α ψr b vs d
= ||ψ||· (8.13)
vs b ψr b ψr α vs q

where ψ = ψrd and ||ψ|| = ψs2α + ψs2b . Next, the following nonlinear feedback
control law is defined
⎛ ⎞

α Mi s q 2
vs d −n p ωi s q − ψr − αbψr d + vd
= σ Ls ⎝ d
α Mi s q i s d
⎠ (8.14)
vs q n p ωi s + bn p ωψr + + vq
d d ψr d
8.3 Flatness-Based Control of Induction Motors in Cascading Loops 413

The control signal in the coordinates system α − b is

−1
vs α ψr α ψr b
= ||ψ||σ L s ·
⎛ vs b −ψr b ψr α ⎞
α Mi 2 (8.15)
−n p ωi s q − ψr s q − αβψr d + vd
·⎝ d
α Mi i
⎠
n p ωi s d + βn p ωψr d + ψrs q s d + vq
d

Substituting Eq. (8.15) into Eq. (8.10) one obtains

d TL
ω = μψr d i s q − (8.16)
dt J
d
i s q = −γ i s q + vq (8.17)
dt
d
ψr = −αψr d + α Mi s d (8.18)
dt d
d
i s d = −γ i s d + vd (8.19)
dt

d is q
ρ = n pω + α M (8.20)
dt ψr d

The system of Eqs. (8.16)–(8.20) consists of two linear subsystems, where the first
one has as output the magnetic flux ψr d and the second has as output the rotation
speed ω, i.e.,
d
ψr = −αψr d + α Mi s d (8.21)
dt d
d
i s d = −γ i s d + vd (8.22)
dt
d TL
ω = μψr d i s q − (8.23)
dt J
d
i s q = −γ i s q + vq (8.24)
dt

If ψr d →ψr ref
d , i.e., the transient phenomena for ψr d have been eliminated and ψr d
has converged to a steady state value, then Eq. (8.23) is not dependent on ψr d , and
consequently the two subsystems described by Eqs. (8.21)–(8.22) and Eqs. (8.23)–
(8.24) are decoupled. The subsystem that is described by Eqs. (8.21) and (8.22) is
linear and has as control input vd , and can be controlled using methods of DC motor
control [336, 413, 434, 535].
414 8 Differential Flatness Theory for Electric Motors and Actuators

8.3.3 A Flatness-Based Control Approach for Induction

Motors

In [338] the voltage-fed induction machine was shown to be a differentially flat

system. It has been proven that the angle of the rotor position (rotation angle θ ) and
the angle ρ of the magnetic field (angle between flux ψra and ψrb ) constitute a flat
output for the induction motor model [117, 118, 535]. Since all state variables of
the circuits describing the induction motor dynamics can be expressed as functions
of y = (θ, ρ) and its derivatives it can be concluded that the induction motor is a
differentially flat system.
The equations of the induction motor in the d − q reference frame, given by
Eqs. (8.21)–(8.24), are now rewritten in the form of Eqs. (8.25)–(8.29):

TL
dt ω = μψr d i s q −
d (8.25)
J

dt ψr d = −αψr d + α Mi s d
d
(8.26)

α Mi s q 2
d
dt i s d = −γ i s d + αβψr d + n p ωi s q + ψr d + 1
σ L s vs d
(8.27)

α Mi s q i s d
d
dt i s q = −γ i s q − βn p ωψr d − n p ωi s d − ψrd + 1
σ L s vs q (8.28)

d α Mi s q
ρ = n pω + (8.29)
dt ψr d

The flat outputs for the voltage-fed induction motor are the angle of the rotor θ and
variable ρ, where ρ has been defined as the rotor flux angle. According to [117],
if the stator current dynamics are much faster than the speed and flux dynamics a
faster inner current control loop can be designed using only Eqs. (8.27) and (8.28)
and assuming the speed and flux as constants. For the outer speed and flux control
design, the stator currents are treated as new control inputs and the system behavior
is described by Eqs. (8.25), (8.26), and (8.29). This system of lower order is also flat
with ψrd and θ as flat outputs.
It can be shown that all state variables of the induction motor can be written
as functions of the flat outputs and their derivatives. Moreover, using Eqs. (8.27)
and (8.28) a controller that satisfies the flatness properties (and thus it can be also
expressed as a function of the flat outputs and their derivatives) is:

di s∗ α Mi sq 2
vsd = σ L s ( dt
d
+ γ i s∗d − αβψrd − n p ωi sq − ψrd + vd ) (8.30)

di s∗q α Mi sq i sd
vsq = σ L s ( dt + γ i s∗q + βn p ωψrd + n p ωi sd + ψrd + vq ) (8.31)
8.3 Flatness-Based Control of Induction Motors in Cascading Loops 415

where i s∗q and i s∗d denote current setpoints. Substituting Eqs. (8.30) and (8.31) into
Eqs. (8.27) and (8.28) one obtains the dynamics of the current tracking errors.

dΔi sd
dt = −γ Δi sd + vd (8.32)

dΔi sq
dt = −γ Δi sq + vq (8.33)

where Δi sd = (i sd − i s∗d ). For the decoupled system of Eqs. (8.32) and (8.33) one
can apply state feedback control. For example, a suitable state feedback controller
would be
vd = −γ1 Δi sd (8.34)

vq = −γ2 Δi sq (8.35)

Tracking of the reference setpoint can be also succeeded for the rotor’s speed and
flux through the application of the control law of Eqs. (8.30) and (8.31) to Eqs. (8.25)
and (8.29). The control inputs are chosen as
∗
1 dψrd
i sd = α M ( dt + αψr∗d + i d ) (8.36)

dω∗
i sq = μψrd ( dt
1
+ iq ) (8.37)

Denoting Δψrd = ψrd − ψr∗d and Δω = ω − ω∗ the tracking error dynamics are
given by
dΔψrd
dt = −αΔψrd + i d (8.38)

dΔω
dt = − TJ + i q (8.39)

The convergence of the tracking error to zero can be assured through the application
of the following feedback control laws:

i d = −α1 Δψrd (8.40)

iq = T
J − α2 Δω (8.41)

8.3.4 Implementation of the EKF for the Nonlinear Induction

Motor Model

State estimation for nonlinear systems with the use of the Extended Kalman Filter
has been explained in Eqs. (4.13) and (4.14). To implement the Extended Kalman
Filter in the induction motor’s model that is expressed in the d − q reference frame
416 8 Differential Flatness Theory for Electric Motors and Actuators

the Jacobian matrices Jφ and Jγ are calculated. Thus:

Jφ = [Jφ1 , Jφ2 , Jφ3 , Jφ4 , Jφ5 , Jφ6 , ]T (8.42)

where the rows of the above-defined Jacobian matrix are given by

Jφ1 = [0, 1, 0, 0, 0, 0], Jφ2 = [0, 0, μx5 , 0, μx3 , 0], Jφ3 = [0, 0, −α, α M, 0, 0],
α M x52
Jφ4 = [0, n p x5 , αβ − x32
, −γ , n p x2 + 2αxM3 x5 , 0], Jφ5 = [0, −βn p x3 − n p x4 ,
α M x4 x5
−βn p x2 + x2
, −n p x2 − α M x5 α M x4
x3 , −γ − x3 , 0] and Jφ6 = [0, n p , − α M
x2
x5
, 0,
3 3
αM
x3 , 0].

Moreover, considering that the motor’s state vector is x = [θ ,ω,ψsd ,i sd ,i sq ,ρ] and
that the measurable state vector elements are the stator currents, initially expressed
in the a − b reference frame as i sa and i sb , and equivalently in the d − q reference
frame as i sd and i sq , one has the measurement equation Jacobian matrix

000100
Jγ = (8.43)
000010

Since the Jacobian matrix Jφ , is associated to the drift term of the system’s dynamics,
and is computed using the system’s continuous-time description of Eq. (8.10), then
in the EKF recursion of Eqs. (4.13) and (4.14) it should be substituted by I + Ts Jφ
where Ts is the sampling period and I ∈R n×n is the identity matrix.

8.3.5 Unscented Kalman Filtering for Induction Motor

Control

Apart from Extended Kalman Filtering for sensorless control of the induction motor,
the Unscented Kalman Filter can be also used. The stages of Unscented Kalman Fil-
tering for nonlinear dynamical systems, consisting of time update and measurement
update have been given in Eqs. (4.18) and (4.19). In Unscented Kalman Filter-based
control a set of suitably chosen weighted sample points (sigma points) are propagated
through the nonlinear system and used to approximate the true value of the system’s
state vector and of the state vector’s covariance matrix. The UKF algorithm is also
summarized as follows:
The time update of the UKF is

xki = φ(xk−1
i ) + L(k − 1)U (k − 1), i = 0, 1, . . . , 2n
−  2n
x̂k = i=0 wi xki −
Px x k − = Px x k−1 + Q k
8.3 Flatness-Based Control of Induction Motors in Cascading Loops 417

The measurement update of the UKF is

z ki = h(xki − , u k ) + rk , i = 0, 1, . . . , 2n
2n
ẑ k = i=0 w zi
2n i k i
Pzz k = i=0 wi [z k − ẑ k ][z ki − ẑ k ]T + Rk
2n
Px z k = i=0 wi [xki − − xkˆ− ][z ki − ẑ k ]T
K k = Px z k Pzz k −1
x̂k = x̂k − + K k [z k − ẑ k ]
Px x k = Pk − − K k Pzz k K kT

It is noted that the Unscented Kalman Filter results in posterior approximations that
are accurate to the third order for Gaussian inputs for all nonlinearities. For non-
Gaussian inputs, approximations are accurate to at least the second order, with the
accuracy of third and higher order moments determined by the specific choice of
weights and scaling factors. Furthermore, unlike EKF no analytical Jacobians of
the system equations need to be calculated. The concept of UKF for approximating
the distribution of a system’s state is given in Fig. 8.6 [533]. It can be observed that
comparing to EKF, the UKF (sigma-point) approach succeeds improved estimation of
the state vector’s mean value and covariance (only 5 points are needed to approximate
sufficiently the 2D distribution).

Fig. 8.6 Approximation of a 2D distribution by the extended and Unscented Kalman Filter
418 8 Differential Flatness Theory for Electric Motors and Actuators

8.4 Simulation Results

The flatness-based control method for the induction motor that was presented in
Sect. 8.3.3 requires knowledge of the motor’s state vector x = [θ ,ω,ψsd ,i sd ,i sq ,ρ]. It
will be shown that it is possible to implement state estimation for the electric motor
using measurements only of the stator currents i sa and i sb . A nonlinear Kalman Fil-
ter, such as the Unscented Kalman Filter or the Extended Kalman Filter, can give
estimates of the nonmeasured state vector elements, i.e., of the rotor’s angle θ , of the
rotation speed ω, of the magnetic flux ψrd , and of the angle ρ between the flux vectors
ψra and ψrb . Using currents i sa and i sb and the estimate ρ̂ of angle ρ, the input mea-
surements i sd and i sq can be provided to the nonlinear Kalman Filters. Thus one has

i sd cos(ρ̂) sin(ρ̂) i
= · sa (8.44)
i sq −sin(ρ̂) cos(ρ̂) i sb

The performance of the proposed sensorless control scheme, which uses the nonlinear
Kalman Filtering for estimation of the nonmeasurable parameters of the motor’s state
vector is depicted in Figs. 8.7, 8.8, 8.9, 8.10 (tracking of a sinusoidal setpoint) and
in Figs. 8.11, 8.12, 8.13, 8.14 (tracking of a seesaw setpoint). Comparison between
the sensorless control loop that is based on the Extended and the Unscented Kalman
Filter is provided.
From the simulation experiments it can be observed that the Unscented Kalman
Filter-based control results in fast and accurate trajectory tracking. The performance

(a) 1.5 (b) 1.5

1 1
position x 1 of the motor

position x 1 of the motor

0.5 0.5

0 0

−0.5 −0.5

−1 −1

−1.5 −1.5
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
t (sec) t (sec)

Fig. 8.7 Angle θ of the induction motor (blue line) in sensorless control when tracking a sinusoidal
setpoint (red line) and state estimation is performed with a the Extended Kalman Filter, b Unscented
Kalman Filter
8.4 Simulation Results 419

(a) 3 (b) 8

6
2

velocity x of the motor

velocity x 2 of the motor

4
1
2

0 0

2
−2
−1
−4
−2
−6

−3 −8
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.8 Angular velocity ω of the induction motor (blue line) in sensorless control when tracking
a sinusoidal setpoint (red line) and state estimation is performed with a the Extended Kalman Filter,
b Unscented Kalman Filter

(a) 0.3 (b) 0.3

0.25
0.2
stator current i of the motor

stator current i of the motor

0.2

0.15 0.1

0.1
0
d

d
s

0.05
−0.1
0

−0.05 −0.2
−0.1
−0.3
−0.15

−0.2 −0.4
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.9 Control input current i sd of the induction motor (blue line) in sensorless control when
tracking a sinusoidal setpoint (red line) and state estimation is performed with a the Extended
Kalman Filter, b Unscented Kalman Filter

of the UKF-based control loop, when considering as measured variables only the
stator currents, was comparable to the one of the EKF-based control loop. Methods
to further enhance the robustness of the nonlinear filtering-based control loops have
been discussed in [12, 38].
420 8 Differential Flatness Theory for Electric Motors and Actuators

(a) 0.45 (b) 3

0.4
2.5
of the motor

of the motor
0.35

0.3 2

0.25
1.5
sq

sq
0.2

stator current i
stator current i

1
0.15

0.1 0.5
0.05
0
0

−0.05 −0.5
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.10 Control input current i sq of the induction motor (blue line) in sensorless control when
tracking a sinusoidal setpoint (red line) and state estimation is performed with a the Extended
Kalman Filter, b Unscented Kalman Filter

(a) 1.2 (b) 1.2

1 1

0.8 0.8
position x of the motor

position x of the motor

0.6
0.6
0.4
0.4
0.2
1

0.2
0
0
−0.2
−0.2
−0.4
−0.4 −0.6

−0.6 −0.8
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.11 Angle θ of the induction motor in sensorless control (blue line) when tracking a seesaw
setpoint (red line) and state estimation is performed with a the Extended Kalman Filter, b Unscented
Kalman Filter
8.4 Simulation Results 421

(a) 3 (b) 8

6
2
4
velocity x2 of the motor

velocity x of the motor

1
2

0 0

2
−1 −2

−4
−2
−6
−3
−8

−4 −10
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.12 Angular velocity ω of the induction motor (blue line) in sensorless control when tracking
a seesaw setpoint (red line) and state estimation is performed with a the Extended Kalman Filter,
b Unscented Kalman Filter

(a) 0.5 (b) 1

0.5
of the motor

of the motor

0
−0.5
d

d
s

−0.5
stator current i

stator current i

−1
−1

−1.5
−1.5

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.13 Control input current i sd of the induction motor (blue line) in sensorless control when
tracking a seesaw setpoint (red line) and state estimation is performed with a the Extended Kalman
Filter, b Unscented Kalman Filter
422 8 Differential Flatness Theory for Electric Motors and Actuators

(a) 1.6 (b) 1.4

1.4 1.2
of the motor

stator current is of the motor

1.2
1
1
0.8
0.8

q
sq

0.6
stator current i

0.6
0.4
0.4
0.2
0.2

0 0

−0.2 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 8.14 Control input current i sq of the induction motor (blue line) in sensorless control when
tracking a seesaw setpoint (red line) and state estimation is performed with a the Extended Kalman
Filter, b Unscented Kalman Filter

8.5 Flatness-Based Adaptive Control of Electrostatic MEMS

Using Output Feedback

8.5.1 Introduction

The previously developed results on adaptive fuzzy control of nonlinear DC motors

will be further extended toward control of microactuators. As micro and nanotech-
nology develop fast, the use of MEMS and particularly of microactuators is rapidly
deploying. One can note several systems where the use of microactuators has become
indispensable and the solution of the associated control problems has become a pre-
requisite. In [472, 487, 613, 614], electrostatic microactuators are used in adaptive
optics and optical communications. In [53, 324] microactuators are used for micro-
manipulation and precise positioning of microobjects. Several approaches to the
control of microactuators have been proposed. In [276, 292, 513] adaptive control
methods have been used. In [146, 583], solution of microactuation control prob-
lems through robust control approaches has been attempted. In [468] backstepping
control has been used, while in [513] an output feedback control scheme has been
implemented. Additional results for the stabilization and control of microactuators
have been presented in [197, 389, 513]. In such control systems, convergence of the
state vector elements to the associated reference setpoints has to be performed with
accuracy, despite modeling uncertainties, parametric variations of external perturba-
tions. Moreover, the reliable functioning of the control loop has to be assured despite
difficulties in measuring the complete state vector of the MEMS. The present section
develops a new method for the control of microelectromechanical systems (MEMS)
which is based on differential flatness theory. The considered control problem is
8.5 Flatness-Based Adaptive Control of Electrostatic … 423

a nontrivial one because of the unknown nonlinear dynamical model of the actua-
tor and because of the constraint to implement the control using exclusively output
feedback (it is little reliable and technically difficult to use sensor measurements for
the monitoring of all state variables of the microactuator). The differential flatness
theory control approach is based on an exact linearization of the MEMS dynamics
which avoids the numerical errors of the approximate linearization that is performed
by other nonlinear control methods [103, 250, 344, 452, 454].
First, the section shows that the dynamic model of the studied microactuator is a
differentially flat one. This means that all its state variables and the control input can
be written as functions of one single algebraic variable, which is the flat output, and
also as functions of the flat output’s derivatives [286, 427, 453, 465, 495]. This change
of variables (differential flatness theory-based diffeomorphism) enables to transform
the nonlinear model of the actuator into the linear canonical (Brunovsky) form [152,
351, 516, 535]. In the latter description of the MEMS the transformed control input
contains elements which are associated with the unknown nonlinear dynamics of
the system. These are identified on-line with the use of neuro-fuzzy approximators
and the estimated system dynamics is finally used for the computation of the control
signal that will make the MEMS state vector track the desirable setpoints. Thus
an adaptive fuzzy control scheme is implemented [407, 454]. The learning rate of
the neuro-fuzzy approximators is determined by the requirement to assure that the
Lyapunov function of the control loop will remain a negative definite one.
Next, another problem that has to be dealt with is that only output feedback can
be used for the implementation of the MEMS control scheme. The nonmeasurable
state variables of the microactuator have to be reconstructed with the use of a state
estimator (observer), which functions again inside the control loop. Thus, finally, the
Lyapunov function for the proposed control scheme comprises three quadratic terms:
(i) a term that describes the deviation of the MEMS state variables from the reference
setpoints, (ii) a term that describes the error in the estimation of the nonmeasurable
state vector elements of the microactuator with respect to the reference setpoints,
and (iii) a sum of quadratic terms associated with the distance of the weights of
the neuro-fuzzy approximators from the values that give the best estimation of the
unknown MEMS dynamics. It is proven that an adaptive (learning) control law can
be found assuring that the Lyapunov function will remain a negative definite one,
thus assuring that the stability of the control loop will be preserved and that accurate
tracking of the setpoints by the system’s state variables will be succeeded (H-infinity
tracking performance).

8.5.2 Dynamic Model of the Electrostatic Actuator

The considered MEMS (electrostatic microactuator) is depicted in Fig. 8.15. The

dynamic model of the MEMS has been analyzed in [180, 203, 614–616], where
model-based control approaches have been mostly developed. It is assumed that
424 8 Differential Flatness Theory for Electric Motors and Actuators

Q(t) is the charge of the device, while ε is the permitivity in the gap. Then the
capacitance of the device is
εA
C(t) = G(t) (8.45)

while the attractive electrostatic force on the moving plate is

Vn2 ∂C ε AV 2
= − Q2ε(t)
2
F(t) = 2 ∂G = − 2G 2 (t)
n
A
(8.46)

Thus, the equation of motion of the actuator is given by

m G̈(t) + b Ġ(t) + k(G(t) − G 0 ) = − Q2ε(t)

2
A
(8.47)

From Eqs. (8.46) and (8.47) it can be concluded that the electrostatic force F increases
with the inverse square of the gap, while the restoring mechanical force which is
associated with the term k(G(t) − G 0 ) increases linearly with the plate deflection. A
critical value for the voltage across the device is called pull-in voltage and is given
by [613] 
8kG 20
V pi = 27C0
(8.48)

It is assumed that the MEMS starts operating from an initially uncharged state at t =
t
0. Then the charge of the electrodes at time instant t is given by Q(t) = 0 Is (τ )dτ ,
or equivalently Q̇(t) = Is (t). By applying Kirchhoff’s voltage law one has for the
current that goes through the resistor

Q(t)G(t)
Q̇(t) = R (Vs (t) −
1
εA ) (8.49)

Next, the equations of the system’s dynamics given in Eqs. (8.47)–(8.49) undergo
a transformation which consists of a change of the time scale τ = ωt and of the
following normalization

Q
x =1− G
G0 q= Q pi
Vs Is (8.50)
u= V pi i= V pi ω0 C0 r = ω0 C0 R

εA
where C0 = G , Q pi = 23 C0 V pi is the pull-in charge corresponding to the pull-in
√
0
voltage, ω0 = k/m is the undamped natural frequency, and ζ = 2mω b
0
is the damp-
ing ratio. The normalized voltage across the actuator can be expressed in terms of
normalized deflection x of the moveable electrode, that is, u o = 23 q(1 − x), while
the dynamics of the normalized charge is q̇ = 23 i.
8.5 Flatness-Based Adaptive Control of Electrostatic … 425

Fig. 8.15 Diagram of the 1-DOF parallel-plate electrostatic actuator

After the aforementioned normalization and transformation, the dynamic model of

the actuator is written as [613]

ẋ = v
v̇ = −2ζ v − x + 13 q 2 (8.51)
q̇ = r1 q(1 − x) + 3r2 u

The model’s state variables are defined as follows: ẋ = v: is a variable denoting

the speed of deflection of the moving electrode, q is a variable denoting the ratio
between the actual change of the plates Q and the pull-in charge Q pi . It holds that
q = QQp , where Q pi = 23 Co V pi and V pi is the pull-in voltage.
i

8.5.3 Linearization of the MEMS Model Using Lie Algebra

The MEMS nonlinear dynamics given in Eq. (8.51), with state vector defined as
x = [x, v, q], is also written in the form

ẋ = f (x) + g(x)u (8.52)

where the vector fields f (x) and g(x) are defined as

⎛ ⎞ ⎛ ⎞
v 0
f (x) = ⎝−2ζ v − x + 21 q 2 ⎠ g(x) = ⎝ 0 ⎠ (8.53)
2
− r1 q(1 − x) 3r

Using the above formulation, one can arrive at a linearized description of the
MEMS dynamics using a differential geometric approach and the computation of Lie
426 8 Differential Flatness Theory for Electric Motors and Actuators

derivatives. The following state variables are defined: z 1 = h 1 (x) = x, z 2 =

L f h 1 (x) and z 3 = L 2f h 1 (x). It holds that

z 2 = L f h 1 (x)⇒z 2 = ∂h ∂h 1 ∂h 1
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 ⇒
1
(8.54)
z 2 = 1 f 1 + 0 f 2 + 0 f 3 ⇒z 2 = f 1 ⇒z 2 = v⇒z 2 = ẋ

In a similar manner one computes

z 3 = L 2f h 1 (x)⇒z 3 = ∂∂zx21 f 1 + ∂∂zx22 f 2 + ∂∂zx23 f 3 ⇒

(8.55)
z 3 = 0 f 1 + 1 f 2 + 0 f 3 ⇒z 3 = v̇⇒z 3 = ẍ

Morever, it holds that

ż 3 = x (3) = L 3f h 1 (x) + L g L 2f h 1 x·u (8.56)

where

L 3f h 1 (x) = L f z 2 ⇒L 3f h 1 (x) = ∂∂zx31 f 1 + ∂∂zx32 f 2 + ∂∂zx33 f 3 ⇒

L 3f h 1 (x) = 1 f 1 − 2ζ f 2 + 23 q f 3 ⇒L 3f h 1 (x) = v − 2ζ v̇ + 23 q(− r1 q(1 − x))⇒
L f h 1 (x) = ẏ − 2ζ ÿ + 23 q[− r1 q(1 − x)]⇒L 3f h 1 (x) = −2ζ ÿ − ẏ − r1 (1 − y) 23 q 2 ⇒
3

L 3f h 1 (x) = −2ζ ÿ − ẏ − r2 (1 − y)[ ÿ + 2ζ ẏ + y]

(8.57)

Following a similar procedure one finds

L g L 2f h 1 (x) = L g z 3 ⇒L g L 2f h 1 (x) = ∂∂zx31 g1 + ∂∂zx32 g2 + ∂∂zx33 g3 ⇒

L g L 2f h 1 (x) = 1g1 − 2ζ g2 + 23 qg3 ⇒L g L 2f h 1 (x) = 9r4 q⇒ (8.58)
√
L g L 2f h 1 (x) = 9r4 3[ ÿ + 2ζ ẏ + y]

For the linearized description of the MEMS dynamics given in Eq. (8.56), and using
that v = L 3f h 1 (x) + L g L 2f h 1 (x)u one arrives also at the state-space description
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (8.59)
ż 3 000 z3 1
⎛ ⎞
  z1
z meas = 1 0 0 ⎝z 2 ⎠ (8.60)
z3
8.5 Flatness-Based Adaptive Control of Electrostatic … 427

For the linearized description of the system given in Eq. (8.69) the design of a state
feedback controller is carried out as follows:
(3)
v = yd − k1 ( ÿ − ÿd ) − k2 ( ẏ − ẏd ) − k3 (y − yd ) (8.61)

which results in tracking error dynamics of the form

e(3) (t) + k1 ë(t) + k2 ė(t) + k3 e(t) = 0 (8.62)

By selecting the feedback gains ki , i = 1, 2, 3 such that the characteristic polynomial

of Eq. (8.62) to be a Hurwitz one, it is assured that lim t→∞ e(t) = 0.

8.5.4 Differential Flatness of the Electrostatic Actuator

8.5.4.1 Differential Flatness Properties of the Electrostatic

Microactuator

The dynamic model of the electrostatic actuator given in Eq. (8.51) is considered.
The flat output of the model is taken to be y = x. Therefore, it also holds v = ẏ.
From the second row of the state-space equations, given in Eq. (8.51) one has

ẏ − y + 13 q 2 ⇒q 2 = 3[ ÿ + 2ζ ẏ + y]
ÿ = −2ζ√
(8.63)
⇒q = 3[ ÿ + 2ζ ẏ + y]⇒q = f q (y, ẏ, ÿ)

From the third row of the state-space equations, given in Eq. (8.51) one has

u= 2 [q̇
3r
+ r1 q(1 − x)]⇒u = f u (y, ẏ, ÿ) (8.64)

Since all state variables and the control input of the system are expressed as functions
of the flat output and its derivatives, it is concluded that the model of the electrostatic
actuator is a differentially flat one.

8.5.4.2 Linearization of the MEMS Model Using Differential Flatness

Theory

From the second row of the state-space model given in Eq. (8.51) it holds that

ÿ = −2ζ ẏ − y + 13 q 2 (8.65)

By deriving once more with respect to time one gets

y (3) = −2ζ ÿ − ẏ + 23 q q̇ (8.66)

428 8 Differential Flatness Theory for Electric Motors and Actuators

By substituting the third row of the state-space model given in Eq. (8.51) one obtains

y (3) = −2ζ ÿ − ẏ + 23 q[− r1 q(1 − x) + 3r2 u]⇒

(8.67)
y (3) = −2ζ ÿ − ẏ − 3r2 (1 − x)q 2 + 9r4 qu

Next, using from Eq. (8.63) that q 2 = ÿ + 2ζ ẏ + y or equivalently that q =

√
ÿ + 2ζ ẏ + y the following relation is obtained

4√
y (3) = −2ζ ÿ − ẏ − 2e (1 − y)[ ÿ + 2ζ ẏ + y] + 9r 3[ ÿ + 2ζ ẏ + y]u (8.68)

or equivalently
y (3) = f (y, ẏ, ÿ) + g(y, ẏ, ÿ)u (8.69)

where
f (y, ẏ, ÿ) = −2ζ ÿ − ẏ − r2 (1 − y)[ ÿ + 2ζ ẏ + y] (8.70)

4 √
g(y, ẏ, ÿ) = 9r [ 3[ ÿ + 2ζ ẏ + y] (8.71)

For the linearized description of the MEMS dynamics given in Eq. (8.69), and using
the notation z 1 = y, z 2 = ẏ and z 3 = ÿ, and v = f (y, ẏ, ÿ) + g(y, ẏ, ÿ)u one
arrives also at the state-space description
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (8.72)
ż 3 000 z3 1
⎛ ⎞
  z1
z meas = 1 0 0 ⎝z 2 ⎠ (8.73)
z3

For the linearized description of the system given in Eq. (8.69) the design of a state
feedback controller is carried out as follows:
(3)
v = yd − k1 ( ÿ − ÿd ) − k2 ( ẏ − ẏd ) − k3 (y − yd ) (8.74)

which results in tracking error dynamics of the form

e(3) (t) + k1 ë(t) + k2 ė(t) + k3 e(t) = 0 (8.75)

By selecting the feedback gains ki , i = 1, 2, 3 such that the characteristic polynomial

of Eq. (8.75) to be a Hurwitz one, it assured that lim t→∞ e(t) = 0.
8.5 Flatness-Based Adaptive Control of Electrostatic … 429

8.5.5 Adaptive Fuzzy Control of the MEMS Model Using

Output Feedback

8.5.5.1 Problem Statement

In subsection 8.5.4 the model of the MEMS actuator was transformed to a form
for which it is possible to apply differential flatness theory-based adaptive fuzzy
control. The purpose for using adaptive control, is to solve the microactuator’s control
problem in case that its dynamics is unknown and the state vector is not completely
measurable. It has been shown that after applying the differential flatness theory-
based transformation, the following nonlinear SISO system is obtained:

x (n) = f (x, t) + g(x, t)u + d̃ (8.76)

where f (x, t), g(x, t) are unknown nonlinear functions and d̃ is an unknown addi-
tive disturbance. The objective is to force the system’s output y = x to follow a
given bounded reference signal xd . As explained in Chap. 3, in the presence of non-
Gaussian disturbances w, successful tracking of the reference signal is denoted by
the H∞ criterion [454]  
T T
e T Qedt ≤ ρ 2 wT wdt (8.77)
0 0

where ρ is the attenuation level and corresponds to the maximum singular value of
the transfer function G(s) of the linearized equivalent of Eq. (8.76).

8.5.5.2 Transformation of Tracking into a Regulation Problem

The H∞ approach to nonlinear systems control consists of the following steps: (i)
linearization is applied, (ii) the unknown system dynamics are approximated by
neural of fuzzy estimators, (iii) an H∞ control term, is employed to compensate for
estimation errors and external disturbances. If the state vector is not measurable, this
can be reconstructed with the use of an observer.
For measurable state vector x, desirable state vector xm and uncertain functions
f (x, t) and g(x, t) an appropriate control law for (8.76) would be

1
u= [x (n) − fˆ(x, t) + K T e + u c ] (8.78)
ĝ(x, t) m

where, fˆ and ĝ are the approximations of the unknown parts of the system dynamics
f and g respectively, and which can be given by the outputs of suitably trained neuro-
fuzzy networks. The term u c denotes a supervisory controller which compensates
for the approximation error w = [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u, as well as
for the additive disturbance d̃. Moreover the vectors K T = [kn , kn−1 , . . . , k1 ], and
430 8 Differential Flatness Theory for Electric Motors and Actuators

e T = [e, ė, ë, . . . , e(n−1) ]T are chosen such that the polynomial e(n) + k1 e(n−1) +
k2 e(n−2) + · · · + kn e is Hurwitz. The control law of Eq. (8.78) in Eq. (8.76) results
into
1 [x (n) − fˆ(x, t) − K T e + u ] + d̃ ⇒
x (n) = f (x, t) + g(x, t) ĝ(x,t) m c
1 [x (n) − fˆ(x, t) − K T e + u ] + d̃ ⇒
x (n) = f (x, t) + {ĝ(x, t) + [g(x, t) − ĝ(x, t)]} ĝ(x,t) m c
(n)
x (n) = f (x, t) + { ĝ(x,t) [x − fˆ(x, t) − K T e + u c ] + [g(x, t) − ĝ(x, t)]u + d̃ ⇒
ĝ(x,t) m
(n)
x (n) = f (x, t) + xm − fˆ(x, t) − K T e + u c + [g(x, t) − ĝ(x, t)]u + u c + d̃ ⇒
(n)
x (n) − xm = −K T e + [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + u c + d̃ ⇒
x (n) = −K T e + u c + [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d̃
(8.79)

The above relation can be written in a state equations form. The state vector is taken
to be e T = [e, ė, . . . , e(n−1) ], which yields

ė = Ae − B K T e + Bu c + B{[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d̃}

(8.80)
or equivalently

ė = (A − B K T )e + Bu c + B{[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d̃}

e1 = C T e
(8.81)
where ⎛ ⎞
0 1 0 ··· ··· 0
⎜ 0 0 1 ··· ··· 0 ⎟
⎜ ⎟
⎜· · · · · · · · · · · · · · · · · ·⎟
A=⎜ ⎜ ⎟
⎟
⎜· · · · · · · · · · · · · · · · · ·⎟
⎝ 0 0 0 ··· ··· 1 ⎠
(8.82)
0 0 0 ··· ··· 0
   
B T = 0, 0, . . . , 0, 1 , C T = 1, 0, . . . , 0, 0
 
K T = k0 , k1 , . . . , kn−2 , kn−1

where e1 denotes the output error e1 = x −xm . Equation (8.81) describes a regulation
problem.

8.5.5.3 Estimation of the State Vector

As explained in Chap. 5, the control of the system described by Eq. (8.76) becomes
more complicated when the state vector x is not directly measurable and has to be
reconstructed through a state observer. The following definitions are used:
8.5 Flatness-Based Adaptive Control of Electrostatic … 431

• error of the state vector e = x − xm

• error of the estimated state vector ê = x̂ − xm
• observation error ẽ = e − ê = (x − xm ) − (x̂ − xm )
When an observer is used to reconstruct the state vector, the control law of Eq. (8.78)
is written as
1
u= [x (n) − fˆ(x̂, t) + K T e + u c ] (8.83)
ĝ(x̂, t) m

Applying Eq. (8.83) to the nonlinear system described by Eq. (8.76), after some oper-
ations results into
(n)
x (n) = xm − K T ê + u c + [ f (x, t) − fˆ(x̂, t)]+
[g(x, t) − ĝ(x̂, t)]u + d̃

It holds e = x − xm ⇒ x (n) = e(n) + xm(n) . Substituting x (n) in the above equation

gives
e(n) + xm(n) = xm(n) − K T ê + u c + [ f (x, t) − fˆ(x̂, t)]+
(8.84)
+[g(x, t) − ĝ(x̂, t)]u + d̃ ⇒

ė = Ae − B K T ê + Bu c + B{[ f (x, t) − fˆ(x̂, t)]+

(8.85)
+[g(x, t) − ĝ(x̂, t)]u + d̃}

e1 = C T e (8.86)

˙ ê,
where e = [e, ė, ë, . . . , e(n−1) ]T , and ê = [ê, ê, ¨ . . . , ê(n−1) ]T .

The state observer is designed according to Eqs. (8.85) and (8.86) and is given by
[454]:
ê˙ = Aê − B K T ê + K o [e1 − C T ê] (8.87)

ê1 = C T ê (8.88)

The observation gain K o = [ko0 , ko1 , . . . , kon−2 , kon−1 ]T is selected so as to assure

the convergence of the observer.

8.5.5.4 Additional Control Term for Disturbances Compensation

The additional term u c which appears in Eq. (8.83) is used in the observer-based
control to compensate for:
• The external disturbances d̃
• The state vector estimation error ẽ = e − ê = x − x̂
432 8 Differential Flatness Theory for Electric Motors and Actuators

• The approximation error of the nonlinear functions f (x, t) and g(x, t), denoted
as w = [ f (x, t) − fˆ(x̂, t)] + [g(x, t) − ĝ(x̂, t)]u
The control signal u c consists of 2 terms, namely:
• the H∞ control term, u a = − r1 B T P ẽ for the compensation of d and w
• the control term u b for the compensation of the observation error ẽ

8.5.5.5 Dynamics of the Observation Error

The observation error is defined as ẽ = e − ê = x − x̂. Substructing Eq. (8.87) from

Eq. (8.85) as well as Eq. (8.88) from Eq. (8.86) one gets

ė − ê˙ = A(e − ê) + Bu c + B{[ f (x, t) − fˆ(x̂, t)]+

+[g(x, t) − ĝ(x̂, t)]u + d̃} − K o C T (e − ê)

e1 − ê1 = C T (e − ê)

i.e.,

ẽ˙ = Aẽ + Bu c + B{[ f (x, t) − fˆ(x̂, t)]+

+[g(x, t) − ĝ(x̂, t)]u + d̃} − K o C T ẽ

ẽ1 = C T ẽ

which can be written as

ẽ˙ = (A − K o C T )ẽ + Bu c + B{[ f (x, t) − fˆ(x̂, t)]+

(8.89)
+[g(x, t) − ĝ(x̂, t)]u + d̃

ẽ1 = C ẽ (8.90)

8.5.5.6 Approximation of the Unknown MEMS Dynamics

Neuro-fuzzy networks can be trained on-line to approximate parts of the dynamic

equation of nonlinear systems, as well as to compensate for external disturbances.
The approximation of functions f (x, t) and g(x, t) of Eq. (8.83) can be carried out
with Takagi–Sugeno neuro-fuzzy networks of zero or first order (Fig. 3.1). These
consist of rules of the form:

x̂˙ is Al2 AND · · · AND x̂ (n−1) is Aln

R l : IF x̂ is Al1 AND
n
THEN ȳ = i=1
l wli x̂i + bl , l = 1, 2, . . . , L
8.5 Flatness-Based Adaptive Control of Electrostatic … 433

The output of the neuro-fuzzy model is calculated by taking the average of the
consequent part of the rules
L 
l n μ ( x̂ )
l=1 ȳ i=1 Ali i
ŷ =  L n (8.91)
l=1 i=1 μ Ali ( x̂ i )

where μ Al is the membership function of xi in the fuzzy set Ali . The training of
i
the neuro-fuzzy networks is carried out with 1st order gradient algorithms, in pat-
tern mode, i.e., by processing only one data pair (xi , yi ) at every time step i. The
estimation of f (x, t) and g(x, t) can be written as

fˆ(x̂|θ f ) = θ Tf φ(x̂)ĝ(x̂|θg ) = θgT φ(x̂) (8.92)

n
i=1 μ Al ( x̂i )
where φ(x̂) are kernel functions with elements φ l (x̂) = L n i l =
l=1 i=1 μ Al ( x̂i )
i
1, 2, . . . , L. It is assumed that that the weights θ f and θg vary in the bounded areas
Mθ f and Mθg which are defined as

Mθ f = {θ f ∈ R h : ||θ f || ≤ m θ f }
(8.93)
Mθg = {θg ∈ R h : ||θg || ≤ m θg }

with m θ f and m θg positive constants. The values of θ f and θg for which optimal
approximation is succeeded are:

θ ∗f = arg min θ f ∈Mθ f [supx∈Ux ,x̂∈Ux̂ | f (x) − fˆ(x̂|θ f )|]

θg∗ = arg min θg ∈Mθg [supx∈Ux ,x̂∈Ux̂ |g(x) − ĝ(x̂|θg )|]

The variation ranges of x and x̂ are the compact sets

Ux = {x ∈ R n : ||x|| ≤ m x < ∞},

(8.94)
Ux̂ = {x̂ ∈ R n : ||x̂|| ≤ m x̂ < ∞}

The approximation error of f (x, t) and g(x, t) is defined as

w = [ fˆ(x̂|θ ∗f ) − f (x, t)] + [ĝ(x̂|θg∗ ) − g(x, t)]u ⇒

w = {[ fˆ(x̂|θ ∗f ) − f (x|θ f )] + [ f (x|θ f ) − f (x, t)]}+ (8.95)
{[ĝ(x̂|θg∗ ) − g(x̂|θg )] + [g(x̂|θg ) − g(x, t)]}u

where
• fˆ(x̂|θ ∗f ) is the approximation of f for the best estimation θ ∗f of the weights’ vector
θf.
434 8 Differential Flatness Theory for Electric Motors and Actuators

• ĝ(x̂|θg∗ ) is the approximation of g for the best estimation θg∗ of the weights’ vector
θg .
The approximation error w can be decomposed into wa and wb , where

wa = [ fˆ(x̂|θ f ) − fˆ(x̂|θ ∗f )] + [ĝ(x̂|θg ) − ĝ(x̂|θg∗ )]u

wb = [ fˆ(x̂|θ ∗f ) − f (x, t)] + [ĝ(x̂|θg∗ ) − g(x, t)]u

Finally, the following two parameters are defined:

θ̃ f = θ f − θ ∗f , θ̃g = θg − θg∗ (8.96)

8.5.6 Lyapunov Stability Analysis

The adaptation law of the neuro-fuzzy approximators weights θ f and θg as well

as of the supervisory control term u c is derived from the requirement for negative
definiteness of the Lyapunov function

1 T 1 1 T 1 T
V = ê P1 ê + ẽ T P2 ẽ + θ̃ θ̃ f + θ̃ θ̃g (8.97)
2 2 2γ1 f 2γ2 g

The selection of the Lyapunov function is based on the following principle of indirect
adaptive control ê : limt→∞ x̂(t) = xd (t) and ẽ : limt→∞ x̂(t) = x(t). This yields
limt→∞ x(t) = xd (t). Substituting Eqs. (8.85), (8.86) and Eqs. (8.89), (8.90) into
Eq. (8.97) and differentiating results into

1 T ˙ 1 T ˙
V̇ = 21 ê˙ T P1 ê + 21 ê T P1 ê˙ + 21 ẽ˙ T P2 ẽ + 21 ẽ T P2 ẽ˙ + γ1 θ̃ f θ̃ f + γ2 θ̃g θ̃g
(8.98)

which in turn gives

V̇ = 21 {(A − B K T )ê + K o C T ẽ}T P1 ê + 21 ê T P1 {(A − B K T )ê + K o C T ẽ}+

+ 21 {(A − K o C T )ẽ + Bu c + Bd + Bw}T P2 ẽ + 21 ẽ T P2 {(A − K o C T )ẽ+}
+Bu c + Bd + Bw + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g
(8.99)
or, equivalently

V̇ = 21 {ê T (A − B K T )T + ẽ T C K oT }P1 ê + 21 ê T P1 {(A − B K T )ê + K o C T ẽ}+

+ 21 {ẽ T (A − K o C T )T + B T u c + B T w + B T d}P2 ẽ + 21 ẽ T P2 {(A − K o C T )ẽ+
+Bu c + Bw + Bd} + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g
(8.100)
8.5 Flatness-Based Adaptive Control of Electrostatic … 435

V̇ = 21 ê T (A − B K T )T P1 ê + 21 ẽ T C K oT P1 ê+
+ 21 ê T P1 (A − B K T )ê + 21 ê T P1 K o C T ẽ+
+ 2 ẽ (A − K o C T )T P2 ẽ + 21 B T P2 ẽ(u c + w + d)+
1 T
(8.101)
+ 21 ẽ T P2 (A − K o C T )ẽ + 21 ẽ T P2 B(u c + w + d)+
+ γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙ g

Assumption 1: For given positive definite matrices Q 1 and Q 2 there exist positive
definite matrices P1 and P2 , which are the solution of the following Riccati equations
[454]
(A − B K T )T P1 + P1 (A − B K T ) + Q 1 = 0 (8.102)
T
(A − K o C T ) P2 + P2 (A − K o C T )−
(8.103)
−P2 B( r2 − ρ12 )B T P2 + Q 2 = 0

The conditions given in Eqs. (8.102)–(8.103) are related to the requirement that the
systems described by Eqs. (8.87), (8.88), (8.89), and (8.90) exhibit stable dynamics.
Substituting Eqs. (8.102)–(8.103) into V̇ yields

V̇ = 21 ê T {(A − B K T )T P1 + P1 (A − B K T )}ê + ẽ T C K oT P1 ê+

+ 2 ẽ {(A − K o C T )T P2 + P2 (A − K o C T )}ẽ + B T P2 ẽ(u c + w + d)+
1 T
(8.104)
+ γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g

which is also written as

V̇ = − 21 ê T Q 1 ê + ẽ T C K oT P1 ê − 21 ẽ T {Q 2 − P2 B( r2 − ρ12 )B T P2 }ẽ+
(8.105)
+B T P2 ẽ(u c + w + d) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g

The supervisory control u c is decomposed in two terms, u a and u b

u a = − r1 p1n ẽ1 = − r1 ẽ T P2 B + r1 ( p2n ẽ2 + · · · + pnn ẽn ) =

(8.106)
= − r1 ẽ T P2 B + Δu a

where p1n stands for the last (nth) element of the first row of matrix P2 , and

u b = −[(P2 B)T (P2 B)]−1 (P2 B)T C K oT P1 ê (8.107)

• u a is an H∞ control used for the compensation of the approximation error w

and the additive disturbance d̃. Its first component − r1 ẽ T P2 B has been chosen so
as to compensate for the term r1 ẽ T P2 B B T P2 ẽ, which appears in Eq. (8.105). By
subtracting the second component − r1 ( p2n ẽ2 + · · · + pnn ẽn ) one has that u a =
− r1 p1n ẽ1 , which means that u a is computed based on the feedback the measurable
variable ẽ1 . Equation (8.106) is finally rewritten as u a = − r1 ẽ T P2 B + Δu a .
436 8 Differential Flatness Theory for Electric Motors and Actuators

Fig. 8.16 The proposed H∞ control scheme

• u b is a control used for the compensation of the observation error (the control
term u b has been chosen so as to satisfy the condition ẽ T P2 Bu b = −ẽ T C K oT P1 ê
(Fig. 8.16).
The control scheme is depicted in Fig. 10.18. Substituting Eqs. (8.106) and (10.212)
in V̇ , one gets

V̇ = − 21 ê T Q 1 ê + ẽ T C K oT P1 ê − 21 ẽ T Q 2 ẽ + r1 ẽ T P2 B B T P2 ẽ−
− 2ρ1 2 ẽ T P2 B B T P2 ẽ + ẽ T P2 Bu b − r1 ẽ T P2 B B T P2 ẽ + B T P2 ẽ(w + d + Δu a )+
+ 1 θ̃ T θ̃˙ + 1 θ̃ T θ̃˙
γ1 f f γ2 g g
(8.108)
or equivalently,

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 1 T
2ρ 2
ẽ P2 B B T P2 ẽ + B T P2 ẽ(w + d + Δu a )+
+ γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g
(8.109)
8.5 Flatness-Based Adaptive Control of Electrostatic … 437

It holds that θ̃˙ f = θ̇ f − θ˙∗f = θ˙f and θ̃˙g = θ̇g − θ˙g∗ = θ˙g . The following weight
adaptation laws are considered:

−γ1 ẽ T P2 Bφ(x̂) i f ||θ f || < m θ f
θ̇ f = (8.110)
0 ||θ f || ≥ m θ f

−γ2 ẽ T P2 Bφ(x̂)u c i f ||θg || < m θg
θ̇g = (8.111)
0 ||θg || ≥ m θg

To set θ̇ f and θ̇g equal to 0, when ||θ f ≥ m θ f ||, and ||θg ≥ m θg || the projection
operator is employed [427]:

P{γ1 ẽ T P2 Bφ(x̂)} = −γ1 ẽ T P2 Bφ(x̂) +

θ f θ Tf
+γ1 ẽ T P2 B ||θ 2 φ( x̂)
f ||

P{γ1 ẽ T P2 Bφ(x̂)u c } = −γ1 ẽ T P2 Bφ(x̂)u c +

θ f θ Tf
+γ1 ẽ T P2 B ||θ 2 φ( x̂)u c
f ||

The update of θ f follows a gradient algorithm on the cost function 21 ( f − fˆ)2 [31,
405]. The update of θg is also of the gradient type, while u c implicitly tunes the
adaptation gain γ2 . Substituting Eqs. (8.110) and (8.111) in V̇ gives

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 1 T
2ρ 2
ẽ P2 B B T P2 ẽ + B T P2 ẽ(w + d + Δu a )+
+ γ11 θ̃ Tf (−γ1 ẽ T P2 Bφ(x̂)) + γ2 θ̃g (−γ2 ẽ
1 T T P2 Bφ(x̂)u)
(8.112)
which is also written as

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ −1 T
2ρ 2
ẽ P2 B B T P2 ẽ + ẽ T P2 B(w + d + Δu a )−
−ẽ P2 B θ̃ f φ(x̂) − ẽ T P2 B θ̃gT φ(x̂)u
T T

(8.113)
and using Eqs. (8.92) and (8.96) results into

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 B B T P2 ẽ + ẽ T P2 B(w + d + Δu a )−
−ẽ T P2 B{[ fˆ(x̂|θ f ) + ĝ(x̂|θ f )u] − [ fˆ(x̂|θ ∗f ) + ĝ(x̂|θg∗ )u]}
(8.114)
where [ fˆ(x̂|θ f ) + ĝ(x̂|θ f )u] − [ fˆ(x̂|θ ∗f ) + ĝ(x̂|θg∗ )u] = wa . Thus setting w1 =
w + wa + d + Δu a one gets

V̇ = − 21 ê T Q 1 ê 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 B B T P2 ẽ + B T P2 ẽw1 ⇒
V̇ = − 21 ê T Q 1 ê 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 B B T P2 ẽ + 21 w1T B T P2 ẽ + 21 ẽ T P2 Bw1
(8.115)
438 8 Differential Flatness Theory for Electric Motors and Actuators

Lemma: The following inequality holds

1 T
2 ẽ P2 Bw1 + 21 w1T B T P2 ẽ − 1 T
2ρ 2
ẽ P2 B B T P2 ẽ ≤ 21 ρ 2 w1T w1 (8.116)

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ2a2 + ρ2
b − 2ab ≥ 0 ⇒ 21 ρ 2 a 2
1 2
+ 1 2
2ρ 2
b − ab ≥ 0 ⇒ ab − 1 2
2ρ 2
b ≤ 21 ρ 2 a 2 ⇒
1
2 ab + 2 ab − 2ρ 2 b ≤ 2 ρ a
1 1 2 1 2 2

The following substitutions are carried out: a = w1 and b = ẽ T P2 B and the previous
relation becomes
1 T
2 w1 B T P2 ẽ + 21 ẽ T P2 Bw1 − 2ρ1 2 ẽ T P2 B B T P2 ẽ
(8.117)
≤ 21 ρ 2 w1T w1

The above inequality is used in V̇ , and the right part of the associated inequality is
enforced
1 1 1
V̇ ≤ − ê T Q 1 ê − ẽ T Q 2 ẽ + ρ 2 w1T w1 (8.118)
2 2 2
Thus, Eq. (8.118) can be written as

1 1
V̇ ≤ − E T Q E + ρ 2 w1T w1 (8.119)
2 2
where

ê Q1 0
E= , Q= = diag[Q 1 , Q 2 ] (8.120)
ẽ 0 Q2

Hence, the H∞ performance criterion is derived. For ρ sufficiently small Eq. (8.118)
will be true and the H∞ tracking criterion will be satisfied. In that case, the integration
of V̇ from 0 to T gives
T T T
0 ||E||2 dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
V̇ (t)dt ≤ − 21 0
T T
2V (T ) − 2V (0) ≤ − 0 ||E||2Q dt + ρ 2 0 ||w1 ||2 dt ⇒
T T
2V (T ) + 0 ||E||2Q dt ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt
∞
It is assumed that there exists a positive constant Mw > 0 such that ||w1 ||2 dt ≤
T 0
Mw . Therefore for the integral 0 ||E||2Q dt one gets
 ∞
||E||2Q dt ≤ 2V (0) + ρ 2 Mw (8.121)
0
8.5 Flatness-Based Adaptive Control of Electrostatic … 439
∞
Thus, the integral 0 ||E||2Q dt is bounded and according to Barbalat’s Lemma

limt→∞ ê(t) = 0
limt→∞ E(t) = 0 ⇒
limt→∞ ẽ(t) = 0

Therefore limt→∞ e(t) = 0.

8.5.7 Simulation Tests

The performance of the proposed output feedback-based adaptive fuzzy control

approach for MEMS (microactuator) was tested in the case of tracking of several
reference setpoints. The only measurable variable used in the control loop was the
microactuator’s deflection variable x. The dynamic model of the MEMS, as well as
the numerical values of its parameters were considered to be completely unknown.
The control loop was based on simultaneous estimation of the unknown MEMS
dynamics (this was performed with the use of neuro-fuzzy approximators) and of the
nonmeasurable elements of the microactuator’s state vector, that is, of the deflections
change rate ẋ and of the charge of the plates q (this was performed with the use of
the state observer). The obtained results are presented in Figs. 8.17, 8.18, 8.19, 8.20,
8.21. The real values of the monitored parameters (state vector variables) are denoted
with blue line, the estimated variables are denoted with green line, and the reference
setpoints are plotted as red lines. It can be noticed that differential flatness theory-
based adaptive fuzzy control of the MEMS, succeeded fast and accurate tracking of
the reference setpoints.

(a) 4 (b) 4
1
state x

2 2
y

0 0
0 5 10 15 20 0 5 10 15 20
time time
2 2
2
state x

0 0
y

−2 −2
0 5 10 15 20 0 5 10 15 20
time time
4 5
3
state x

3 0
y

2 −5
0 5 10 15 20 0 5 10 15 20
time time

Fig. 8.17 Output feedback-based adaptive fuzzy control of MEMS (microactuator)—Test 1: a

state variables xi , i = 1, . . . , 3 of the initial nonlinear system, b transformed state variables yi ,
i = 1, . . . , 3 (blue line real value, red line setpoint)
440 8 Differential Flatness Theory for Electric Motors and Actuators

(a) 1
state x
4 (b) 4

1
2 2

y
0 0
0 5 10 15 20 0 5 10 15 20
time time
2 2
2
state x

y2
0 0

−2 −2
0 5 10 15 20 0 5 10 15 20
time time
4 2
3
state x

3
3 0

y
2 −2
0 5 10 15 20 0 5 10 15 20
time time

Fig. 8.18 Output feedback-based adaptive fuzzy control of MEMS (microactuator)—Test 2: a

state variables xi , i = 1, . . . , 3 of the initial nonlinear system, b transformed state variables yi ,
i = 1, . . . , 3 (blue line real value, red line setpoint)

(a) 10 (b) 10
1
state x

y1

5 5

0 0
0 5 10 15 20 0 5 10 15 20
time time
2 2
2
state x

1 1
y

0 0
0 5 10 15 20 0 5 10 15 20
time time
10 1
3
state x

y3

5 0

0 −1
0 5 10 15 20 0 5 10 15 20
time time

Fig. 8.19 Output feedback-based adaptive fuzzy control of MEMS (microactuator)—Test 3: a

state variables xi , i = 1, . . . , 3 of the initial nonlinear system, b transformed state variables yi ,
i = 1, . . . , 3 (blue line: real value, red line: setpoint)

The implementation of the proposed control scheme requires that the two algebraic
Riccati equations which have been defined in Eqs. (8.102) and (8.103) are solved in
each iteration of the control algorithm. These provide the positive definite matrices P1
and P2 which are used for the computation of the control signals u a and u b which have
been defined in Eqs. (8.106) and (8.107). The transients of the state vector elements
xi , i = 1, . . . , 3 observed while tracking the reference setpoints, are determined by
the values given to the positive definite matrices Q i , i = 1, 2, as well as by the value
of the parameter r and of the H-infinity coefficient (attenuation level) ρ. Moreover,
8.5 Flatness-Based Adaptive Control of Electrostatic … 441

(a) 20
state x 1 (b) 20

y1
10 10

0 0
0 5 10 15 20 0 5 10 15 20
time time
5 5
2
state x

2
0 0

y
−5 −5
0 5 10 15 20 0 5 10 15 20
time time
10 20
state x 3

3
5 0

y
0 −20
0 5 10 15 20 0 5 10 15 20
time time

Fig. 8.20 Output feedback-based adaptive fuzzy control of MEMS (microactuator)—Test 4: a

state variables xi , i = 1, . . . , 3 of the initial nonlinear system, b transformed state variables yi ,
i = 1, . . . , 3 (blue line: real value, red line: setpoint)

(a) 10 (b) 10
state x1

y1

5 5

0 0
0 5 10 15 20 0 5 10 15 20
time time
5 5
2
state x

0 0
y

−5 −5
0 5 10 15 20 0 5 10 15 20
time time
10 20
state x3

5 0
y

0 −20
0 5 10 15 20 0 5 10 15 20
time time

Fig. 8.21 Output feedback-based adaptive fuzzy control of MEMS (microactuator)—Test 5: a

state variables xi , i = 1, . . . , 3 of the initial nonlinear system, b transformed state variables yi ,
i = 1, . . . , 3 (blue line: real value, red line: setpoint)

the values of the feedback control gains K and K o also affected the convergence
characteristics of the controller and of the observer. It has been confirmed that the
variations of both xi , i = 1, . . . , 3 and of the control input u were smooth.
From the simulation tests it can be noticed that the proposed adaptive control
scheme that was based on differential flatness theory assured the stability of the
microactuator’s control loop, as well as good transient performance in the tracking
of the reference setpoints.
Chapter 9
Differential Flatness Theory in Power
Electronics

9.1 Introduction

The chapter analyzes the use of flatness-based control and differential flatness
theory-based filtering in power electronics, such as voltage source converters and
inverters. The chapter is first concerned with proving differential flatness of the
three-phase voltage source converter (VSC) model and its resulting description in
the Brunovksy (canonical) form. Voltage source converters are widely used in the
electricity network, as for example in the case of the control of asynchronous power
generators or in the case of connecting AC power generation units with DC transmis-
sion lines. For the linearized canonical model of the converter, a feedback controller is
designed. At the second stage, the Derivative-free nonlinear Kalman Filter is applied.
The proposed Kalman Filtering method is redesigned as disturbance observer for esti-
mating additive input disturbances to the VSC model. These estimated disturbance
terms are finally used by a feedback controller that enable the DC output voltage to
track the desirable setpoints. The efficiency of the proposed state estimation-based
control scheme is tested through simulation experiments.
Next, the problem of control and filtering for inverters is analyzed. The use of
three-phase voltage inverters (DC to AC converters) is frequently met in the electric
power system, such as in the connection of photovoltaics with the rest of the grid.
The chapter proposes a nonlinear feedback control method for three-phase inverters,
which is based on differential flatness theory and the new nonlinear filtering method
under the name Derivative-free nonlinear Kalman Filter. First, it is shown that the
inverter’s dynamic model is a differentially flat one. This means that all its state vari-
ables and the control inputs can be written as functions of a single algebraic variable
which is the flat output. By exploiting differential flatness properties, it is shown that
the inverter’s model can be transformed to the linear canonical (Brunovsky’s) form.
For the latter description, the design of a state feedback controller becomes possible,
e.g., using pole placement methods. Moreover, to estimate the nonmeasurable state

© Springer International Publishing Switzerland 2015 443

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_9
444 9 Differential Flatness Theory in Power Electronics

variables of the linearized equivalent of the inverter, the Derivative-free nonlinear

Kalman Filter is used. As already explained in previous chapters, this consists of the
Kalman Filter recursion applied on the linearized inverter’s model and of an inverse
transformation that is based on differential flatness theory, which enables to com-
pute estimates of the state variables of the initial nonlinear system. Furthermore, by
redesigning the aforementioned filter as a disturbance observer it also becomes possi-
ble to estimate disturbance terms that affect the inverter and subsequently to compen-
sate for them. The performance and disturbance rejection capability of the proposed
nonlinear feedback control scheme is evaluated through simulation experiments.
Finally, the chapter examines the problem of synchronization between distributed
inverters, and to this end control and filtering based on differential flatness theory
is applied again. To assure power quality and stability of the electricity network it
is important to perform control and synchronization of the inverters which are used
in the connection of distributed DC power generation units to the grid. To this end,
decentralized control for parallel inverters connected to the power grid is developed
using differential flatness theory. In this case, in the dynamics of each inverter, one
has also to compensate for interaction terms which are due to the coupling with other
inverters. It is proven that the model of the inverters is a differentially flat one. This
means that all its state variables and the control inputs can be written as functions of
a single algebraic variable which is the flat output. By exploiting differential flatness
properties, it is shown that the multiple inverters system can be transformed into a set
of local inverter models which are decoupled and linearized. For each local inverter,
the design of a state feedback controller becomes possible, e.g., using pole placement
methods. Such a controller processes measurements not only coming from the indi-
vidual inverter but also coming from other inverters which are connected to the grid.
Moreover, to estimate the nonmeasurable state variables of each local inverter, the
Derivative-free nonlinear Kalman Filter is used. Finally, by redesigning the afore-
mentioned filter as a disturbance observer it becomes also possible to estimate and
compensate for disturbance terms that affect each local inverter. The performance and
disturbance rejection capability of the proposed nonlinear feedback control scheme
for the multiple inverters model is evaluated through simulation experiments.

9.2 Three-Phase Voltage Source Converters Control

9.2.1 Overview

The present section proposes a nonlinear control scheme for three-phase voltage
source converters where estimation of disturbances and variations of the load cur-
rent is performed with the use of a new nonlinear filtering approach, the so-called
derivative-free nonlinear Kalman Filter. Voltage source converters, are three-phase
filtered rectifiers, that are widely used in the electric power grid (mainly for power
9.2 Three-Phase Voltage Source Converters Control 445

flow control). Voltage source converters are the main building blocks of power
flow controllers in transmission lines. For example, VSCs are contained in Unified
Power Flow Controllers (UPFCs), or distribution-static synchronous compensators
(D-STATCOMs). VSCs enable control of the amplitude and phase angle of the AC
terminal voltages. Moreover, their bidirectional power flow capabilities allow VSCs
to perform real and/or reactive power flow control in AC transmission lines [612].
The dynamic model of voltage source converters is a nonlinear one and several
nonlinear control methods have been proposed for it. Linearization round certain
operating points with the computation of Jacobian matrices and control with the
use of linear feedback control methods has limited efficiency [58, 528]. Operation
range is restricted and a relatively big capacitor is needed for keeping a constant DC
voltage in presence of a varying load. Initial nonlinear control approaches consider
the representation of the voltage source converter dynamics in the dq reference
frame and use PI compensators. Passivity-based control methods have been proposed
in [141]. Neural/fuzzy control methods for VSCs have been analyzed in [72–75].
Backstepping control approaches have been proposed in [9]. State estimation-based
control for voltage source converters has been studied in [212, 272, 279]. Other
control methods for voltage source converter’s that use indirect estimation of the
converter’s parameters and functional characteristics can be found in [66, 67, 321,
322].
Additional control approaches consider input-output linearization, as well as
input-state linearization [343]. The latter methods transform the nonlinear system
into a decoupled linear one. It is also known that one can attempt transformation of
the nonlinear VSC model into the linear canonical (Brunovsky) form through the
application of Lie algebra theory. With the use of Lie algebra methods, it is possible
to arrive at a description of the system in the linear canonical form if the relative
degree of the system is equal to the order of the system. However, this linearization
procedure requires the computation of Lie derivatives (time derivatives on the vector
fields describing the system dynamics) and this can be a cumbersome computation
procedure. Moreover, differential flatness theory has been proposed for VSC control
[172, 173, 447, 500].
This section is concerned with proving differential flatness of the three-phase volt-
age source converter model and its resulting description in the Brunovksy (canonical)
form [340]. It is shown that for the linearized converter’s model it is possible to design
a feedback controller. At a second stage, the Derivative-free nonlinear Kalman Filter
is proposed for estimating the non directly measurable elements of the state vector
of the linearized system. With the redesign of the proposed Kalman Filter as a dis-
turbance observer, it becomes possible to estimate disturbance terms which are due
to variations of the load current and to use these terms in the feedback controller.
By avoiding linearization approximations, the proposed filtering method, improves
the accuracy of estimation, and results in smooth control signal variations and in
minimization of the tracking error of the associated control loop [55, 434, 436, 437].
446 9 Differential Flatness Theory in Power Electronics

9.2.2 Linearization of the Converter’s Model Using Lie

Algebra

9.2.2.1 Dynamic Model of the Voltage Source Converter

The voltage source converter model in the rotating dq reference frame is given by
[279, 500]:
Vdc
L i̇ d = −Ri d + Lωi q + vd − 2 u1
Vdc
L i̇ q = −Lωi d − Ri q + vq − 2 u 2 (9.1)
Cdc V̇dc = − R1c Vdc + 43 i d u 1 + 43 i q u 2

where i d , i q are the line currents (i a , i b , i c ) after transformation in the dq reference

frame, and equivalently vd , vq are the phase voltages va , vb , vc after transformation in
the dq reference frame. Variable Vdc denotes the DC voltage output of the converter,
u 1 = ηd and u 2 = ηq stand for control inputs. The line losses and the transformer
conduction losses are modeled by R and the inverter switching losses are modeled
by Rc . Moreover, vq is taken to be 0.
In Fig. 9.1 the electric circuit of the voltage source converter is depicted. Denoting
x = [i d , i q , Vdc ]T as the state vector, y = [ec , i q ]T as the output and u = [ηd , ηq ]T
as the input vector, the MIMO nonlinear one of the VSC is written in the state-space
form

ẋ = f (x) + G(x)u
(9.2)
y = h(x)

Fig. 9.1 Electrical circuit of the voltage source converter

9.2 Three-Phase Voltage Source Converters Control 447

where ⎛ vd ⎞ ⎛ x3 ⎞
− RL x1 + ωx2 + L
− 2L 0
f = ⎝ −ωx1 − RL x2 ⎠ G = [g1 g2 ] = ⎝ 0 − 2L3 ⎠
x
(9.3)
3x1 3x1
− Cdxc3Rc 4Cdc 4Cdc

    3 Cdc 2

h1 e L(x12 + x22 ) + 2 x3
h= = c = 4 (9.4)
h2 iq x2

9.2.2.2 Use of Lie Algebra for the Linearization of the Converter’s Model

Linearization of the converter’s model can be performed using Lie algebra and with
the computation of the associated Lie derivatives [238, 327]. The following state
variables are defined: z 1 = h 1 (x), z 2 = L f h 1 (x), and z 3 = h 2 (x). Thus one
gets L f h 1 (x) = ∂h ∂h 1 ∂h 1 vd
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 ⇒L f h 1 (x) = 2 L x 1 (− L x 1 + ωx 2 + L ) +
1 3 R
x3
2 L x 2 (−ωx 1 − L x 2 )+C dc x 3 (− Cdc Rc )⇒z 2 = L f h 1 (x) = − 2 R(x 1 +x 2 )+ 2 vd x 1 −
3 R 1 3 2 2 3
1 2
Rc x 3
Similarly, one has

ż 2 = L 2f h 1 (x) + L g1 L f h 1 (x)u 1 + L g2 L f h 2 (x)u 2 (9.5)

where
∂z 21
∂z 21 ∂z 21
L 2f h 1 (x) = ∂ x1 f 1 +
∂ x2 f 2 + ∂ x3 f 3 ⇒L f h 1 (x) = (−3Rx 1 + 2 vd )(− L x 1 + ωx 2 +
2 3 R

vd x3 2
L ) + (−3Rx 2 )(−ωx 1 − L x 2 ) + (− Rc x 3 )(− Cdc Rc )⇒L f h 1 (x) = L (x 1 + x 2 ) −
R 2 2 3R 2 2

2x32

L vd x 1 + Cdc Rc2
3R

while it also holds1

∂z ∂z 1 ∂z 1 x3
L g1 L f h 1 (x) = ∂ x21 g11 + ∂ x22 g21 + ∂ x23 g31 ⇒L g1 L f h 1 (x) = (−3Rx1 + 23 vd )(− 2L )+
3x1 3x1 x3
(−3Rx2 )0 + (− R2c x3 )( 4C dc
)⇒L g1 L f h 1 (x) = 3R
2L x 1 x 3 − 3
4L vd x 3 − 2Rc Cdc
and also
∂z 1 ∂z 1 ∂z 1
L g2 L f h 1 (x) = ∂ x21 g12 + ∂ x22 g22 + ∂ x23 g32 ⇒
x3 3x2
L g2 L f h 1 (x) = (−3Rx1 + 23 vd )0+(−3Rx2 )(− 2L )+(− R2c x3 )( 4C dc
)⇒L g2 L f h 1 (x)
3Rx2 x3 3x2 x3
= 2L − 2Rc Cdc
Therefore, one has

ż 1 = z 2
(9.6)
ż 2 = L 2f h 1 (x) + L g1 L f h 1 (x)u 1 + L g2 L f h 1 (x)u 2

It can be confirmed that it holds ż 1 = z 2 . Indeed one has z 1 = h 1 (x) therefore

∂z 1 ∂z 1 ∂z 1
ż 1 = ∂ x11 ẋ1 + ∂ x12 ẋ2 + ∂ x13 ẋ3 ⇒ż 1 = 23 L x1 (− RL x1 +ωx2 + vLd )+ 23 L x2 (−ωx1 − RL x2 )+
Cdc x3 (− Cdcx3Rc )⇒ż 1 = − 23 Rx12 + 23 Lωx1 x2 + 23 vd x1 − 23 Rx22 − 23 Lωx1 x2 − R1C x32
448 9 Differential Flatness Theory in Power Electronics

Thus, one finally obtains

ż 1 = − 23 R(x12 + x22 ) + 23 vd x1 − 1 2
Rc x 3 (9.7)

Equation (9.5) can be also confirmed. Indeed, using that z 2 = L f h 1 (x) = − 23 R(x12 +
x22 ) + 23 vd x1 − R1c x32 one obtains
∂z 21 ∂z 21 ∂z 21 vd x3
ż 2 =
∂ x1 ẋ 1 + ∂ x2 ẋ 2 + ∂ x3 ẋ 3 ⇒ż 2 = (−3Rx 1 + 2 vd )(− L x 1 + ωx 2 + L − 2L u 1 ) +
3 R
x3 x3 3x1 3x2
(−3Rx2 )(−ωx1 − L x2 − 2L u 2 ) + (− Rc x3 )(− Cdc Rc + 4Cdc u 1 + 4Cdc u 2 )⇒
R 2

3R 2 2 2x32
ż 2 = L (x 1 + x 2 ) − L vd x 1 + Cdc Rc2 +
2 3R

3x1 x3 3Rx2 x3 3x2 x3

(9.8)
2L x 1 x 3 −
+( 3R 4L vd x 3 − 2Rc Cdc )u 1 + ( 2L
3
− 2R c C dc
)u 2

which through the previous relations about L 2f h 1 (x), L g1 L f h 1 (x), and L g2 L f h 1 (x)
confirms finally Eq. (9.5).
The third state variable is also examined, that is z 3 = h 2 (x) = i q . It holds that

L f h 2 (x) = ∂h ∂h 2 ∂h 2
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 ⇒
2

L f h 2 (x) = 0 f 1 + 1 f 2 + 0 f 3 ⇒ (9.9)
L f h 2 (x) = −ωx1 − RL x2

Equivalently, one gets

L g1 h 2 (x) = ∂h ∂h 2 ∂h 2
∂ x1 g11 + ∂ x2 g21 + ∂ x3 g31 ⇒
2

L g1 h 2 (x) = 0g11 + 1g21 + 0g31 ⇒ (9.10)

L g1 h 2 (x) = 0

and similarly one has

L g2 h 2 (x) = ∂h ∂h 2 ∂h 2
∂ x1 g12 + ∂ x2 g22 + ∂ x3 g32 ⇒
2

L g2 h 2 (x) = 0g12 + 1g22 + 0g32 ⇒ (9.11)

x3
L g2 h 2 (x) = − 2L

Therefore, it holds

ż 3 = L f h 2 (x) + L g1 h 2 (x)u 1 + L g2 h 2 (x)u 2 ⇒

x3
ż 3 = (−ωx1 − RL x2 ) + 0u 1 − 2L u2 (9.12)
x
ż 3 = (−ωx1 − L x2 ) − 2L u 2
R 3

Equation (9.12) corresponds to the second line of the state-space equations of Vdc
and therefore confirms the relation

ż 3 = L f h 2 (x) + L g1 h 2 (x)u 1 + L g1 h 2 (x)u 2 (9.13)

9.2 Three-Phase Voltage Source Converters Control 449

Therefore, one arrives at the following equations

ż 1 = z 2
ż 2 = L 2f h 1 (x) + L g1 L f h 1 (x)u 1 + L g2 L f h 2 (x)u 2 (9.14)
ż 3 = f 21 + g22 u 2

where the new control inputs are defined as

v1 = L 2f h 1 (x) + L g1 L f h 1 (x)u 1 + L g2 L f h 2 (x)u 2

(9.15)
v2 = f 21 + g22 u 2

Consequently, one gets

ż 1 = z 2
ż 2 = v1 (9.16)
ż 3 = v2

or in state-space description
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 z1 0 0  
⎝ż 2 ⎠ = ⎝0 v
0 1⎠ ⎝ z 2 ⎠ + ⎝ 1 0⎠ 1
v2
ż 3 0 0 0 z3 0 1
⎛ ⎞ (9.17)
    z
y1 1 0 0 ⎝ 1⎠
= z2
y2 0 0 1
z3

Therefore, the initial nonlinear system of the voltage source converter is transformed
into the linear canonical form.

9.2.3 Differential Flatness of the Voltage Source Converter

9.2.3.1 Differential Flatness of the Voltage Source Converter’s Model

It will be shown that the dynamic model of the voltage source converter is a differen-
tially flat one, i.e., it holds that all state variables and its control inputs can be written
as functions of the flat outputs and their derivatives. Moreover, it will be shown that
by expressing the elements of the state vector as functions of the flat outputs and their
derivatives, one obtains a transformation of the VSC model into the linear canonical
(Brunovsky) form.
The following flat outputs are defined
Cdc 2
y f1 = h 1 = 4 (x 1
3L 2
+ x22 ) + 2 x3
(9.18)
450 9 Differential Flatness Theory in Power Electronics

y f2 = h 2 = iq (9.19)

It holds that
ẏ f1 = − 23 R(x12 + x22 ) + 23 vd x1 − 1 2
Rc x 3 (9.20)

while deriving once more with respect to time one gets

3R 2 2 2x32
ÿ f1 = 2 (x 1 + x 2 ) − L vd x 1 + Cdc Rc2 +
2 3R

3x1 x3 3Rx2 x3 3x2 x3

(9.21)
2L x 1 x 3 −
+(− 3R 4L vd x 3 − 2Rc Cdc )u 1 + ( 2L
3
− 2R c C dc
)u 2

Moreover,
x3
ẏ f2 = (−ωx1 − L x 2 ) − 2L u 2
R (9.22)

It can be confirmed that all state variables and the control inputs of the VSC model
can be written as functions of the flat outputs y f1 , y f2 and of their derivatives. It holds
that y f2 = x2 = i q . Moreover, using the definition of the flat outputs it holds

ẏ f1 = − 23 R(x12 + x22 ) + 23 vd x1 − 1 2
Rc x 3 (9.23)

y f2 = x2 (9.24)

Solving Eq. (9.23) with respect to x32 one obtains

x32 = Rc [− 23 Rx12 − 23 R y 2f2 + 23 vd x1 − ẏ f1 ] (9.25)

By substituting Eq. (9.25) into Eq. (9.18), one gets

y f1 = 43 L x12 + 43 L y 2f2 + C2dc Rc [− 23 Rx12 − 23 R y22 + 23 vd x1 − ẏ f1 ]⇒

( 3L−3C4dc Rc R )x12 + 3Cdc4Rc vd x1 + ( 43 L y 2f2 − 3Cdc4Rc R y 2f2 − Cdc2 Rc ẏ f1 − y f1 ) = 0
(9.26)
By computing the roots of the binomial given in the above equation, it becomes pos-
sible to express state variable y1 as a function of the flat outputs and their derivatives.
Thus, by keeping the maximum of the binomial’s root’s one obtains

x1 = f a (y f1 , ẏ f1 , y f2 ) (9.27)

Using the relation for x1 described in Eq. (9.27) into Eq. (9.25) one has

x32 = Rc [− 23 R f a (y f 1 , ẏ f 1 , y f 2 ) − 23 R y 2f + 23 vd f a (y f 1 , ẏ f 1 , y f 2 ) − ẏ1 ]⇒
2
1
x3 = Rc [− 23 R f a (y f 1 , ẏ f 1 , y f 2 ) − 23 R y 2f + 23 vd f a (y f 1 , ẏ f 1 , y f 2 ) − ẏ1 ] 2 ⇒ (9.28)
2
x3 = f b (y f 1 , ẏ f 1 , y f 2 )
9.2 Three-Phase Voltage Source Converters Control 451

From the first line of the state-space description of the system given in Eqs. (9.2) and
(9.3)–(9.4) it holds that
vd x3
ẋ1 = (− RL x1 + ωx2 + L ) − 2L u 1 ⇒
vd
(− RL x1 +ωx2 + )−ẋ1
u1 = x3
L
⇒ (9.29)
2L

u 1 = f c (y f1 , ẏ f1 , y f2 )

Similarly, from the second line of the state-space description of the system given in
Eqs. (9.2) and (9.3)–(9.4) it holds that
x3
ẋ2 = (−ωx1 − L x 2 ) − 2L u 2 ⇒
R

−ωx1 − RL x2 −ẋ2
u2 = x3 ⇒ (9.30)
2L

u 2 = f d (y f1 , ẏ f1 , y f2 )

Consequently, all state variables and the control inputs in the model of the voltage
source converter can be written as functions of the flat outputs and their derivatives.
Thus, the VSC model is a differentially flat one.

9.2.3.2 Linearization of the Converter’s Model Using Differential

Flatness Theory

Using the definitions of the flat outputs y f1 and y f2 for the VSC model, and consid-
ering the new state variables z 1 = y f1 , z 2 = ẏ f1 , and z 3 = y f2 , one obtains

ż 1 = z 2

3R 2 2 2x32
ż 2 = 2 (x 1 + x 2 ) − L vd x 1 + Cdc Rc2 +
2 3R

3x1 x3 3Rx2 x3 3x2 x3

(9.31)
2L x 1 x 3 −
+( 3R 4L vd x 3 − 2Rc Cdc )u 1 + ( 2L
3
− 2R c C dc
)u 2

x3
ż 3 = (−ωx1 − L x 2 ) − 2L u 2
R

or equivalently
ż 1 = z 2
ż 2 = v1 (9.32)
ż 3 = v2

which can be also written in state-space form

452 9 Differential Flatness Theory in Power Electronics
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0 0  
⎝ż 2 ⎠ = ⎝0 0 0⎠ ⎝z 2 ⎠ + ⎝1 v
0⎠ 1
v2
ż 3 000 z3 0 1
⎛ ⎞ (9.33)
    z
ẏ1 1 0 0 ⎝ 1⎠
= z2
ẏ2 001
z3

As already noted, the linearized system is controllable and observable. The new
control inputs vi , i = 1, 2 are the same as the ones defined in the Lie algebra-based
approach. The linearized model of the VSC can be described by the following two
equations
z̈ 1 = v1
(9.34)
ż 3 = v2

The control inputs which enable convergence to the desirable setpoints are

v1 = z̈ 1d − k11 (ż 11 − ż 1d ) − k21 (z 1 − z 1d )

(9.35)
v2 = ż 3d − k12 (z 3 − z 3d )

The control law of Eq. (9.35) succeeds the following tracking error dynamics

ë1 + k11 ė1 + k21 e1 = 0

(9.36)
ė3 + k12 e3 = 0

which results into limt→∞ e1 (t) = 0 and limt→∞ e3 (t) = 0. It can be noticed that the
linearized model obtained after the application of differential flatness theory is the
same with the one obtained with the use of the Lie derivative approach. It holds that

v1 = L 2f h 1 (x) + L g1 L f h 1 (x)u 1 + L g2 L f h 1 (x)u 2

(9.37)
v2 = L f h 2 (x) + L g1 h 2 (x)u 1 + L g2 h 2 (x)u 2

or, in matrix form

   2    
v1 L f h 1 (x) L g1 L f h 1 (x) L g2 L f h 1 (x) u1
= + (9.38)
v2 L f h 2 (x) L g1 h 2 (x) L g2 h 2 (x) u2

It holds that
   3R 2 2 2x32

2 (x 1 + x 2 ) − L vd x 1 +
v1 2 3R
= Cdc Rc2 +
v2 −ωx − R x 1 L 2
(9.39)
 3R 3x1 x3 3Rx2 x3 3x2 x3   
2L x 1 x 3 − 3
4L vd x 3 − − 2R u1
+ 2Rc Cdc 2L
x3
c C dc
0 − 2L u2
9.2 Three-Phase Voltage Source Converters Control 453

According to the above, the VSC dynamic model can be also written in a more
compact form as
ṽ = f˜ + M ũ⇒
(9.40)
ũ = M −1 (ṽ − f˜)

9.2.4 Kalman Filter-Based Disturbance Observer

for the VSC Model

The simultaneous estimation of the nonmeasurable elements of the VSC state vector
(e.g., ẏ f1 ) as well as the estimation of additive disturbance terms (e.g., associated with
variations of the load current i L ) is possible with the use of a disturbance estimator
[85, 107, 108, 183, 354].
Next, it will be considered that additive input disturbances (e.g., due to load
variations) affect the VSC model. Thus, it is assumed that the third row of the state-
space equations of the voltage source converter of Eq. (9.2) includes a disturbance
term
ẋ3 = − Cdcx3Rc + 4C
3x1
dc
3x2
u 1 + 4C dc
u 2 + d̃ (9.41)

By describing the system’s dynamics using the differential flatness theory approach
and using the definition of the input v1 given in Eq. (9.39), the disturbances’ effects
appear as follows:
ÿ f1 = v1 + (− R2c x3 d̃) (9.42)

thus one has that the additive disturbance term is now described by

T̃d = − R2c x3 d̃ (9.43)

The disturbance T̃d may also comprise additional uncertainty terms associated with
the numerical values of the parameters of the VSC model. It is assumed that the
aggregate dynamics of term T̃d is described by its third-order derivative
(3)
T̃d = f L (y f1 , ẏ f1 , y f2 ) (9.44)

Thus, it holds that z 1 = y f1 , z 2 = ẏ f1 , z 3 = y f2 , z 4 = T̃d , z 5 = T̃˙d , and z 6 = T̃¨d .

The dynamics of the extended state-space model is written as

ż 1 = z 2 ż 2 = v1 + T̃d ż 3 = v2
(9.45)
ż 4 = z 5 ż 5 = z 6 ż 6 = f L (y f1 , ẏ f1 , y f2 )

or in matrix form one has

z̃˙ = Ãz̃ + B̃ ṽ
(9.46)
ỹ = C̃ z̃
454 9 Differential Flatness Theory in Power Electronics
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 0 0 0 z1 0 0 0
⎜ż 2 ⎟ ⎜0 0 0 1 0 0⎟ ⎜ z 2 ⎟ ⎜1
0⎟ ⎛ 0 ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ż 3 ⎟ ⎜0 0 0 0 0 0⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟=⎜ ⎟ ⎜z 3 ⎟ + ⎜0
0⎟ ⎝
v2
1 ⎠
⎜ż 4 ⎟ ⎜0 0 0 0 1 0⎟ ⎜
⎟ 0⎟
⎟ ⎜
⎜ ⎟ ⎜ ⎜z 4 ⎟ ⎜0 0
⎟ f L (y f , ẏ f , y f )
⎝ż 5 ⎠ ⎝0 0 0 0 0 1 ⎠ ⎝ 0⎠
z 5 ⎠ ⎝0 1
0 1 2

ż 6 0 0 0 0 0 0 z6 1⎛ ⎞0 0
(9.47)
z1
⎜z 2 ⎟
   ⎜ ⎟
y1 100000 ⎜ ⎟
⎜z 3 ⎟
=
y2 0 0 1 0 0 0 ⎜z 4 ⎟
⎜
⎟
⎝z 5 ⎠
z6

The associated state estimator is

z̃ˆ˙ = Ão z̃ˆ + B̃o ṽ1 + K o ( ỹ − C̃ z̃)

ˆ (9.48)

where ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 000
⎜0 0 0 1 0 0⎟ ⎜1 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0⎟⎟ ⎜0 1 0⎟
Ão = ⎜
⎜0 B̃ = ⎜ ⎟
⎜ 0 0 0 1 0⎟⎟
o ⎜0 0 0⎟
⎜ ⎟
⎝0 0 0 0 0 1⎠ ⎝0 0 0⎠ (9.49)
0 0 0 0 0 0 000
 
100000
C̃o =
001000

while the estimator’s gain K o ∈R 6×2 is obtained from the standard Kalman Filter
recursion [31, 192, 229, 414].
Defining as Ãd , B̃d , and C̃d , the discrete-time equivalents of matrices Ão , B̃o , and
C̃o , respectively, a Derivative-free nonlinear Kalman Filter can be designed for the
aforementioned representation of the system dynamics [427, 436]. The associated
Kalman Filter-based disturbance estimator is given by the measurement update and
time update stages which were already described in Eqs. (4.5) and (4.6).
measurement update:

K (k) = P − (k)C̃dT [C̃d ·P − (k)C̃dT + R]−1

ˆ
z̃(k) = ẑ − (k) + K (k)[C̃d z̃(k) − C̃d z̃ˆ − (k)] (9.50)
P(k) = P − (k) − K (k)C̃d P − (k)
9.2 Three-Phase Voltage Source Converters Control 455

time update:

P − (k + 1) = Ãd (k)P(k) ÃdT (k) + Q(k)

(9.51)
z̃ˆ − (k + 1) = Ãd (k)z̃(k)
ˆ + B̃d (k)ṽ(k)

9.2.5 Simulation Tests

To evaluate the performance of the proposed nonlinear control scheme, that uses
Kalman Filtering to estimate the nonmeasurable disturbances of the VSC model,
simulation experiments have been carried out. Different DC voltage setpoints Vdc
have been assumed. Moreover, different external disturbance terms d̃ (e.g., due to
load perturbations) have been considered to affect the VSC dynamic model. The
control loop used in the VSC control is shown in Fig. 9.2.
Several cases of VSC operation under different perturbation terms have been
presented. The disturbance dynamics was completely unknown to the controller
and its identification was performed in real time by the disturbance estimator. It is
shown that the derivative-free nonlinear Kalman Filter redesigned as a disturbance
observer is capable of estimating with accuracy the unknown disturbance input d̃.
The associated results are presented in Figs. 9.3, 9.4, 9.5, and 9.6. Several reference
setpoints have been defined for the VSC state variables, i.e., currents i d , i q and
the output voltage Vdc and as it can be observed from the associated diagrams, the
proposed control scheme resulted in fast and accurate convergence to these setpoints.
The disturbance observer that was based on the Derivative-free nonlinear Kalman
Filter was capable of estimating the unknown and time-varying input disturbances
affecting the VSC model.

Fig. 9.2 Control loop for the VSC comprising a flatness-based nonlinear controller and a Kalman
Filter-based disturbances estimator
456 9 Differential Flatness Theory in Power Electronics

(a) (b)

Tm1 − Tm1
20 0.5

−est
id

10 0

0 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
1 0

u1
iq

0.5 −50

0 −100
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
400 0
Vdc

u2
200 −2

0 −4
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time

Fig. 9.3 a Control of the state variables of the VSC in case of reference setpoint 1, b estimation of
disturbance input and variation of the control inputs

(a) (b)
Tm1 − Tm1

20 0.5
−est
id

0 0

−20 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
1 0
u1
iq

0.5 −20

0 −40
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
400 0
Vdc

u2

200 −1

0 −2
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time

Fig. 9.4 a Control of the state variables of the VSC in case of reference setpoint 2, b estimation of
disturbance input and variation of the control inputs

The simulation experiments have confirmed that the proposed state estimation-
based control scheme not only enables implementation of VSC control through the
measurement of a small number of variables (e.g., the ones appearing in the flat
outputs y f1 and y f2 ) but also improves the robustness of the VSC control loop in
case of disturbances.
9.2 Three-Phase Voltage Source Converters Control 457

(a) (b)

Tm1 − Tm1
20 1

−est
id

0 0

−20 −1
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
1 100

u1
iq

0.5 0

0 −100
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
400 5
Vdc

u2
200 0

0 −5
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time

Fig. 9.5 a Control of the state variables of the VSC in case of reference setpoint 3, b estimation of
disturbance input and variation of the control inputs

(a) (b)
Tm1 − Tm1

20 1
−est
id

0 0

−20 −1
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
1 200
u1
iq

0.5 0

0 −200
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time
400 50
Vdc

u2

200 0

0 −50
0 2 4 6 8 10 0 2 4 6 8 10
time (sec) time

Fig. 9.6 a Control of the state variables of the VSC in case of reference setpoint 4, b estimation of
disturbance input and variation of the control inputs

The improvement in the performance of the control loop that is due to the use
of a disturbance observer based on the Derivative-free nonlinear Kalman Filter is
explained as follows: (i) compensation of the disturbance terms which are generated
by parametric uncertainty or unknown external inputs (ii) more accurate estimation of
the disturbance terms because the filtering procedure is based on an exact linearization
458 9 Differential Flatness Theory in Power Electronics

(a) (b)
350 350

300 300

250 250

200 200
x3

x3
150 150

100 100

50 50

0 0
0 2 4 6 8 10 0 2 4 6 8 10
t t

Fig. 9.7 Convergence of the output voltage Vdc to the reference setpoint (a) without using the
disturbance observer (b) when using the disturbance observer

of the system’s dynamics and does not introduce numerical errors (as for example in
the case of the Extended Kalman Filter). This is shown in Fig. 9.7.

9.3 Inverters Control

9.3.1 Overview

Next, it is shown how differential flatness theory-based control and filtering can be
used for improving the performance of voltage inverters. A nonlinear control scheme
is developed for three-phase inverters (DC to AC converters) using differential
flatness theory and the Derivative-free nonlinear Kalman Filter [427, 453, 460].
Efficient control of inverters is important for the reliable operation of the electric
power grid and the connection to it of distributed DC power generation units (such
as photovoltaic arrays providing power to the main grid or local photovoltaic cells
connected to a microgrid infrastructure) [71, 73, 78, 84, 130, 246, 315]. Moreover,
inverters find extensive use in HVDCs (High Voltage DC) lines and in their connec-
tion with the rest of the AC electric power network [214, 309, 521]. Indicative results
on nonlinear control of inverters with the use of feedback linearization methods can
be found in [45, 328, 540]. In the same direction, this chapter proposes a method
for inverters’ control, based on differential flatness theory and on a new nonlinear
filtering method under the name Derivative-free nonlinear Kalman Filter.
The dynamic model of the inverter is a nonlinear multi-variable one [130, 214].
The control inputs are the voltages generated by an H-bridge circuit while the state
vector comprises as elements the voltages at the load’s side and the currents at the
9.3 Inverters Control 459

H-bridge circuit output. It is shown that this model is a differentially flat one, which
means that all state variables and the control inputs can be written as functions of a
single algebraic variable which is the so-called flat output (actually for the inverter’s
model the flat output is the vector of the voltages at the load side expressed in the dq
reference frame) [214, 427]. By applying differential flatness theory it is proven that
the inverter’s model can be transformed into the linear canonical (Brunowsky) form
[55, 152, 263, 286, 340, 464, 465, 495]. This transformation enables the design of a
state feedback controller, e.g., using pole placement methods.
Other problems that have to be dealt with in the design of the feedback control
scheme, are the estimation of nonmeasurable variables of the inverter’s state vector
(such as currents and rate of change of voltages), as well as the compensation of
disturbance terms that affect the inverter’s model. To solve the first problem, the
Derivative-free nonlinear Kalman Filter is used. As already explained, this consists
of the standard Kalman Filter recursion applied on the linearized equivalent of the
system and of an inverse transformation that is based on differential flatness the-
ory and which enables to obtain the estimates of the state variables of the initial
nonlinear model [436, 447, 448, 456]. To solve the second problem, the Derivative-
free nonlinear Kalman Filter is redesigned as a disturbance observer. This permits
to obtain simultaneously estimates of the inverter’s state vector and of the additive
perturbation terms that affect its inputs. Since in the case of the inverter’s model, the
Derivative-free nonlinear Kalman Filter keeps the optimality properties of the linear
Kalman Filter algorithm, it succeeds fast and accurate estimation of the perturbation
and modeling uncertainty terms [31, 408, 414, 437, 460].

9.3.2 Dynamic Model of the Inverter

The inverter’s (DC to AC converter’s) circuit is depicted in Fig. 9.8. By applying

Kirchoff’s voltage and current laws, one obtains

dt i I = L f V I − L f VL
d 1 1
(9.52)
dt VL = C f i I − C f i L
d 1 1

For the representation of the voltage and current variables, denoted as X = {I, V }
in the ab static reference frame one has
j2π j4π
X ab = X a e j0 + X b e 3 + Xce 3 (9.53)

which finally gives a complex variable of the form

X ab = X a + j X b (9.54)
460 9 Differential Flatness Theory in Power Electronics

Fig. 9.8 Circuit of the three-phase voltage inverter

Next, the voltage and current variables are represented in the rotating dq reference
frame [130]. It holds that X dq = xd + j xq and

X dq = X ab e− jθ ⇒X
 t ab = X dq e
jθ
(9.55)
where θ (t) = 0 ω(t)dt + θ0

By differentiating with respect to time, one obtains the following description of the
system’s dynamics
Ẋab = dtd
X dq + jωX dq (9.56)

Thus, one has for the current and voltage variables, respectively,

i̇ I,ab = dt
d
i I,dq + ( jω)i I,dq
(9.57)
V̇L ,ab = dt VL ,dq + ( jω)VL ,dq
d

By substituting Eq. (9.57) into Eq. (9.52), one obtains

dt i I,dq + jωi I,dq = L f V I,dq − L f VL ,dq

d 1 1
(9.58)
dt VL ,dq + jωVL ,dq = C f i I,dq − C f i L ,dq
d 1 1
9.3 Inverters Control 461

Using Eq. (9.58) and by rearranging rows, one finally obtains the inverter’s dynamic
model expressed in the dq reference frame:

dt VL ,d = ωVL ,q + C f i I,d − C f i L ,d
d 1 1

dt VL ,q = −ωVL ,d + C f i I,q − C f i L ,q
d 1 1
(9.59)
dt i I,d = ωi I,q + L f V I,d − L f VL ,d
d 1 1

dt i I,q = −ωi I,d + L f V I,q − L f VL ,q

d 1 1

The state vector of the system is taken to be X̃ = [VL d , VL q , i I,d , i I,q ]T while the
control input is taken to be the vector Ũ = [VI,d , VI,q ]. The load currents i L ,d and
i L ,q are taken to be unknown parameters which can be considered as perturbation
terms. Alternatively, these currents can be expressed as functions of the inverter’s
active and reactive power. In the latter approach, one has that the active power of the
inverter is [246]

p f = VL d i L d + VL q i L q (9.60)

while the reactive power, consisting of reactive power at the load, reactive power at
the capacitor, and reactive power at the inductance is given by

q f = VL q i L d − VL d i L q − ωC f (VL2d + VL2q ) + ωL f (i I,d

2 + i2 )
I,q (9.61)

By solving Eqs. (9.60) and (9.61), with respect to the currents i L d and i L q one obtains

p f VL d +q f VL q ωL f VL q (i I2 +i I2q )
i Ld = VL2 +VL2q
+ ωC f VL q − d
(VL2 +VL2q )
(9.62)
d d

p f VL q −q f VL d ωL f VL d (i I2 +i I2q )
i Lq = VL2 +VL2q
− ωC f VL d + d
(VL2 +VL2q )
(9.63)
d d

Using Eqs. (9.59) and (9.62), Eq. (9.63) one obtains the state-space description of the
inverter’s dynamics
⎛ ⎞
ωL f VL q (i I2 +i I2q )
1 1 p f VL d +q f VL q + ωC V
⎜ ωVL q + C f i Id − C f V 2 +V 2 f Lq −
d
⎛ ⎞ (VL2 +VL2q ) ⎟
VL d ⎜ L Lq ⎟
⎜ d
ωL f VL d (i I2 +i I2q ) ⎟
d
⎜ ⎟ ⎜
d ⎜ VL q ⎟ = ⎜−ωVL + 1 i I − 1
p f VL q −q f VL d ⎟
− ωC f VL d + d
⎟+
dt ⎝ i I ⎠ ⎜ d Cf q Cf VL2 +VL2q (VL2 +VL2q ) ⎟
d ⎜ d d ⎟
i Iq ⎜ ωi − 1 V ⎟
⎝ Iq L f Ld ⎠
1
−ωi Id − L VL q
f
462 9 Differential Flatness Theory in Power Electronics
⎛ ⎞
0 0
⎜ 0 0 ⎟ 
⎜ ⎟ V Id
+⎜ 1 0 ⎟ (9.64)
⎝Lf ⎠ V Iq
1
0 Lf

while the measurement equation of the inverter’s model is

⎛ ⎞
      VL d
y1 VL d 1000 ⎜ ⎟
⎜ VL q ⎟
= = ⎝ (9.65)
y2 VL q 0100 i Ld ⎠
i Lq

and by using the state variables notation x1 = VL d , x2 = VL q , x3 = i L d , and x4 = i L q

one has
⎛ p x +q x ωL f x2 (x32 +x42 )
⎞
⎛ ⎞ ωx2 + C1 x3 − C1 f 21 2f 2 + ωC f x2 −
x1 ⎜ f f x1 +x2 (x1 +x2 )
2 2
⎟
⎜ p f x2 −q f x1 ωL f x1 (x32 +x42 ) ⎟
d ⎜ ⎟ ⎜ 1 1
⎜x2 ⎟ = ⎜−ωx1 + C f x4 − C f x 2 +x 2 − ωC f x1 +
⎟
⎟+
⎝ ⎠ ⎜ 1 2 (x12 +x22 ) ⎟
dt x 3 ⎜ 1 ⎟
⎝ ωx4 − L x1 ⎠
x4 f
−ωx3 − L1 x2 (9.66)
f
⎛ ⎞
0 0
⎜ 0 0 ⎟ 
⎜ ⎟ u
+⎜ 1 0 ⎟ 1
⎝Lf ⎠ u2
0 L1
f

while the measurement equation of the inverter’s model is

⎛ ⎞
      x1
y1 VL d 1000 ⎜
⎜x2 ⎟
⎟
= = ⎝ (9.67)
y2 VL q 0100 x3 ⎠
x4

thus, the inverter’s model is written in the nonlinear state-space form

ẋ = f (x) + G(x)u
(9.68)
y = h(x)

where f (x)∈R 4×1 , G(x)∈R 4×1 and h(x)∈R 2×4 .

9.3 Inverters Control 463

9.3.3 Lie Algebra-Based Control of the Inverter’s Model

The following functions are defined

y1 = h 1 (x) = x1
(9.69)
y2 = h 2 (x) = x2

It holds that

z 1 = h 1 (x) = x1
z 2 = L f h 1 (x)
ż 2 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2
(9.70)
z 3 = h 2 (x) = x2
z 4 = L f h 2 (x)
ż 4 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

Using the above, one has

∂h 1 ∂h 1 ∂h 1 ∂h 1
z 2 = L f h 1 (x)⇒z 2 = ∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 + ∂ x4 f 4 ⇒
1 p f x1 +g f x2 ωL f x2 (x32 +x42 )
z 2 = 1 f 1 ⇒z 2 = ωx2 + C f x 3 − C f { x 2 +x 2
1
+ ωC f x2 − (x }
1 +x 2 )
2 2
1 2
(9.71)
In a similar manner, one computes
∂z 1 ∂z 1 ∂z 1 ∂z 1
L 2f h 1 (x) = L f z 2 ⇒L 2f h 1 (x) = ∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 + ∂ x4 f 4 ⇒ (9.72)

Thus, it holds that

p f (x12 +x22 )−( p f x1 +g f x2 )(2x1 ) ωL f x2 (x32 +x42 )(2x1 )

L 2f h 1 (x) = − C1f { 2 + (x12 +x22 )2
}ẋ1 +
(x12 +x22 )

g f (x12 +x22 )−( p f x1 +g f x2 )2x2 ωL f (x32 +x42 )(x12 +x22 )+ωL f x2 (x32 +x42 )2x2
+[ω − 1
Cf { 2 + ωC f − (x12 +x22 )2
}]ẋ2 (9.73)
(x12 +x22 )

ωL x 2x −ωL f x2 2x4
[ C1f − 1
Cf {− (x 2f+x2 2 )23 }](ωx4 − 1
Lf x1 ) + [− C1f { (x12 +x22 )
}](−ωx3 − 1
Lf x2 )
1 2

and after intermediate operations one arrives at the same formulation as in the case
of differential flatness theory. Thus, one has
( p f ẋ 1 +q f ẋ 2 )(x 12 +x 22 )−( p f x 1 +q f x 2 )(2x 1 ẋ 1 +2x 2 ẋ 2 )
L 2f h 1 (x) = ω ẋ2 + 1
Cf (ωx4 − 1
Lf x1 ) − 1
Cf { (x 12 +x 22 )2
}

ωL f ẋ 2 (x 32 +x 42 )(x 12 +x 22 )−ωL f x 2 (x 32 +x 42 )(2x 1 ẋ 1 +2x 2 ẋ 2 )

−ω ẋ2 + 1
Cf { (x 12 +x 22 )2
}

1 1
ωL f x2 2x3 (ωx4 − L f x1 ) ωL f x2 2x1 (−ωx3 − L f x2 )
+ Cf · (x +x )
2 2 + Cf · (x +x )
2 2
(9.74)
1 2 1 2
464 9 Differential Flatness Theory in Power Electronics

In a similar manner one computes

∂z 2 ∂z 2 ∂z 2 ∂z 2
L ga L f h 1 (x) = L ga z 1 ⇒L ga L f h 1 (x) = ∂ x1 ga1 + ∂ x2 ga2 + ∂ x3 ga3 + ∂ x4 ga4 ⇒

1 ∂z 2 ωL x (2x )
L ga L f h 1 (x) = L f ∂ x3 ⇒L ga L f h 1 (x) = 1
Cf { (x 2 +x
f 2
2 )L
3
+ 1
Lf } (9.75)
1 2 f

and also
∂z 2 ∂z 2 ∂z 2 ∂z 2
L gb L f h 1 (x) = L ga z 2 ⇒L gb L f h 2 (x) = ∂ x1 gb1 + ∂ x2 gb2 + ∂ x3 gb3 + ∂ x4 gb4 ⇒

1 ∂z 2 ωL f x2 2x4
L gb L f h 1 (x) = L f ∂ x4 ⇒L gb L f h 1 (x) = 1
C f (x 2 +x 2 ) L f (9.76)
1 2

Following a similar procedure, one obtains

∂h 2 ∂h 2 ∂h 2 ∂h 2
z 4 = L f h 2 (x)⇒z 4 = ∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 + ∂ x4 f 4 ⇒
1 p f x2 −q f x1 ωL f x1 (x32 +x42 ) (9.77)
z 4 = f 2 ⇒z 4 = −ωx1 + C f x 4 − C f { (x 2 +x 2 ) − ωC f +
1
(x 2 +x 2 )
}
1 2 1 2

Moreover, one computes

L 2f h 2 (x) = L f z 4 ⇒L 2f h 2 (x) = ∂∂zx41 f 1 + ∂∂zx42 f 2 + ∂∂zx43 f 3 + ∂z 4

∂ x4 f 4 ⇒ (9.78)
L 2f h 2 (x) = ∂∂zx41 ẋ1 + ∂∂zx42 ẋ2 + ∂∂zx43 ωx4 + ∂∂zx44 −ωx3

which is also written as

1 −g f (x12 + x22 ) − ( p f x2 − q f x1 )(2x1 )

L 2f h 2 (x) = [ − ω − { − ωC f }+
Cf (x12 + x22 )2
(x32 + x42 )(x12 + x22 ) − ωL f x1 (x32 + x42 )2x1
+ ωL f ]ẋ1
(x12 + x22 )2
1 p f (x12 + x22 ) − ( p f x2 − q f x4 )2x2 ωL f x1 (x32 + x42 )2x2
[− { − }]ẋ2
Cf x2 + x2
2 (x12 + x22 )2
1 2
1 ωL f x1 2x3 1
[− { 2 }](ωx4 − x1 )
C f (x1 + x2 )
2 Lf
1 1 ωL f x1 2x4 1
[ − { }](−ωx3 − x2 ) (9.79)
Cf C f (x12 + x22 ) Lf
9.3 Inverters Control 465

which after intermediate operations gives

( p ẋ −g ẋ )(x +x )−( p x −g x )(2x1 ẋ1 +2x2 ẋ2 )

2 2
L 2f h 2 (x) = −ω ẋ1 + C1 (−ωx3 ) − C1 { f 2 f 1 1 2 2 f 22 2 f 1 }+
f f (x1 +x2 )

ωL ẋ (x +x )(x +x )−ωL x (x +x )(2x1 ẋ1 +2x2 ẋ2 )

2 2 2 2 2 2
ω ẋ1 − C1 { f 1 3 4 1 2 2 f2 12 3 4 }
f (x1 +x2 )

ωL f x1 2x3 (ωx4 − L1 x1 ) ωL f x1 2x4 (−ωx3 − L1 x2 )

− C1 f
− C1 f (9.80)
f (x12 +x22 ) f (x12 +x22 )

Following a similar procedure one, finds

∂z 4 ∂z 4 ∂z 4 ∂z 4
L ga L f h 2 (x) = L ga z 4 ⇒L ga L f h 2 (x) = ∂ x1 ga1 + ∂ x2 ga2 + ∂ x3 ga3 + ∂ x4 ga4 ⇒

1 ∂z 4 ωL f x1 2x3 1
L gb L f h 2 (x) = L f ∂ x3 ⇒L ga L f h 2 (x) = − C1f (x12 +x22 ) L f
(9.81)

and equivalently

∂z 4 ∂z 4 ∂z 4 ∂z 4
L gb L f h 2 (x) = L gb z 4 ⇒L gb L f h 2 (x) = ∂ x1 gb1 + ∂ x2 gb2 + ∂ x3 gb3 + ∂ x4 gb4 ⇒

−1 ∂z 4
L gb L f h 2 (x) = L f ∂ x4 ⇒

ωL f x1 2x4 1
L gb L f h 2 (x) = − C1f { (x12 +x22 ) L f
+ 1
Lf } (9.82)

After the Lie algebra-based linearization procedure, the dynamics of the inverter is
written in the form

ẍ1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

(9.83)
ẍ2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

The following new control inputs are defined

v1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

(9.84)
v2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

This results in the linearized system’s dynamics

ẍ1 = v1
(9.85)
ẍ2 = v2

which is also written in a state-space form, after defining the state variables z 1 = x1 ,
z 2 = ẋ1 , z 3 = x2 , and z 4 = ẋ2
466 9 Differential Flatness Theory in Power Electronics
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 0 z1 0 0  
⎜ż 2 ⎟ ⎜0 0 0 0⎟ ⎜ z 2 ⎟ ⎜1 0⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟ v1 (9.86)
⎝ż 3 ⎠ ⎝0 0 0 1⎠ ⎝z 3 ⎠ ⎝0 0⎠ v2
ż 4 0 0 0 0 z4 0 1
⎛ ⎞
    z1
m
z1 1000 ⎜ ⎟
⎜z 2 ⎟
= (9.87)
z 2m 0 0 1 0 ⎝z 3 ⎠
z4

9.3.4 Differential Flatness of the Inverter’s Model

The flat output of the inverter is taken to be the vector

y = [y1 , y2 ] = [VL d , VL q ] (9.88)

It will be shown that the system is a differentially flat one, which means that all its
state variables and the control inputs can be written as functions of the flat output
and their derivatives. It holds that

x1 = VL d ⇒x1 = y1
(9.89)
x2 = VL q ⇒x2 = y2

The first row of the state-space equations is

1 p f x1 +q f x2 1 ωL f x2 (x3 +x4 )
2 2
ẋ1 = ωx2 + 1
Cf x3 − Cf x12 +x22
− ωx2 + Cf x12 +x22
(9.90)

The second row of the state-space equations is

1 p f x2 +q f x1 1 ωL f x1 (x3 +x4 )
2 2
ẋ2 = ωx1 + 1
Cf x4 − Cf x12 +x22
+ ωx1 + Cf x1 +x2
2 2 (9.91)

Both equations are solved with respect to their last term

1 ωL f x2 (x3 +x4 ) 1 p f x1 +q f x2
2 2

Cf x12 +x22
= ẋ1 − ωx2 − 1
Cf x3 + C f (x 2 +x 2 ) + ωx2 (9.92)
1 2

1 ωL f x1 (x3 +x4 ) 1 p f x2 −q f x1
2 2

Cf x1 +x2
2 2 = ẋ2 + ωx1 − 1
Cf x4 + C f (x 2 +x 2 ) − ωx1 (9.93)
1 2
9.3 Inverters Control 467

By dividing the above two equations, one obtains

p f x1 +q f x2
ẋ1 −ωx2 − C1 x3 + C1 +ωx2
f f(x12 +x22 )
− xx21 = p x −q x (9.94)
ẋ2 +ωx1 − C1 x4 + C1 f 22 2f 1 −ωx1
f f (x1 +x2 )

or using in the notation the elements of the flat output vector

p f y1 +q f y2
ẏ1 −ωy2 − C1 x3 + C1 +ωy2
f f(y12 +y22 )
− yy21 = p y −q y (9.95)
ẏ2 +ωy1 − C1 x4 + C1 f 22 2f 1 −ωy1
f f (y1 +y2 )

which equivalently gives

1 y2 p f y2 −q f y1
− yy21 ẏ2 − ωy2 + C f ( y1 ) (y 2 +y 2 ) + ωy2 =
1 2
1 p f y1 +q f y2 (9.96)
= ẏ1 − ωy2 − C f x 3 + C f (y 2 +y 2 ) + ωy2
1
1 2

Equation (9.96) is solved with respect to x3 . It holds that

1 (y2 p f y2 −q f y1
x3 = C f { (y
y1 ) ẏ2 + ωy2 −
2 1
Cf ( yy21 )x4 + C f y1 ) y12 +y22
− ωy2
1 p f y1 +q f y2 (9.97)
+ ẏ1 − ω ẏ2 + Cf y12 +y22
+ ωy2 }

which equivalently gives

1 (y2 p f y2 −q f y1
x3 = − yy21 x4 + C f { yy21 ẏ2 + ωy2 + C f y1 ) (y 2 +y 2 ) − ωy2
1 p f y1 +q f y2
1 2 (9.98)
+ ẏ1 − ωy2 + C f (y 2 +y 2 ) + ωy2 }
1 2

which is also written as

x3 = −( yy21 )x4 + f a (y1 , ẏ1 , y2 , ẏ2 ) (9.99)

Next, Eq. (9.99) is substituted into Eq. (9.90). Thus, one obtains

1 p f x2 −q f x1
ẋ2 = −ωx1 + 1
Cf x4 −
C f (x 2 +x 2 ) + ωx 1 −
1 2
(y2 (9.100)
1 ωL f x1 {[− y1 )x4 + f a (y1 , ẏ1 ,y2 ,dot y2 )] +x4 }
2 2
−Cf (x12 +x22 )

or equivalently

1 p f y2 −q f y1
ẏ2 = −ωy1 + C f x 4 − C f (y 2 +y 2 ) + ωy1 +
1
1 2
y2 (9.101)
1 ωL f y1 {[−( y1 )x4 + f a (y1 , ẏ1 ,y2 , ẏ2 )] +x4 }
2 2
−Cf (y1 +y2 )
2 2
468 9 Differential Flatness Theory in Power Electronics

Equation (9.101) is a binomial which can be solved with respect to x4 . Thus, one
obtains that
x4 = f b (y1 , ẏ1 , y2 , ẏ2 ) (9.102)

Substituting Eq. (9.102) into Eq. (9.99) one gets

x3 = −( yy21 ) f b (y1 , ẏ1 , y2 , ẏ2 ) + f 4 (y1 , ẏ1 , y2 , ẏ2 ) (9.103)

From Eqs. (9.102) and (9.103) one has that state variables x3 and x4 can be written as
functions of the flat outputs and their derivatives. Moreover, from the last two rows
of the state-space equations one has that

ẋ3 = ωx4 − 1
Lf x1 + 1
Lf u1 (9.104)

ẋ4 = −ωx3 − 1
Lf x2 + 1
Lf u2 (9.105)

From Eq. (9.104) one gets

u 1 = L f {ẋ3 − ωx4 + 1
Lf x1 }⇒u 1 = f c (y1 , ẏ1 , y2 , ẏ2 )} (9.106)

From Eq. (9.105) one gets

u 2 = L f {ẋ4 − ωx3 + 1
Lf x2 }⇒u 2 = f d (y1 , ẏ1 , y2 , ẏ2 )} (9.107)

Consequently, the control inputs are also functions of the flat outputs and their deriv-
atives. Thus, the inverter’s model is a differentially flat one.

9.3.5 Flatness-Based Control of the Inverter

Next, a flatness-based controller will be designed. To this end Eqs. (9.90) and (9.91)
are used again. By considering p f , q f as piecewise constant and by deriving
Eq. (9.90) with respect to time, one has

1 {( p f ẋ1 +q f ẋ2 )(x1 +x2 )−( p f x1 +q f x2 )(2x1 ẋ1 +2x2 ẋ2 )

2 2
ẍ1 = ω ẋ2 + 1
Cf ẋ3 − Cf (x 2 +x 2 )2 }
−
1 2

ωL f ẋ2 (x32 +x42 )(x12 +x22 )+ωL f x2 (2x3 ẋ3 +2x4 ẋ4 )(x12 +x22 )
−ω ẋ2 + 1
Cf { (x12 +x22 )2
} (9.108)

ωL f x2 (x32 +x42 )(2x1 ẋ1 +2x2 ẋ2 )

− (x12 +x22 )2
9.3 Inverters Control 469

The time derivatives are substituted from the associated rows of the state-space
equations. Next, substitutions of the terms ẋ3 and ẋ4 are performed. It holds that

{( p f ẋ1 +q f ẋ2 )(x12 +x22 )−( p f x1 +q f x2 )(2x1 ẋ1 +2x2 ẋ2 )

ẍ1 = ω ẋ2 + C1 (ωx4 − L1 x1 + L1 u 1 ) − C1 −
f f f f (x 2 +x 2 )2 } 1 2

2 2 2 2
ωL ẋ (x +x )(x +x )+ωL x (x +x )(2x1 ẋ1 +2x2 ẋ2 ) 2 2
−ω ẋ2 + C1 { f 2 3 4 1 2 2 f2 22 3 4 }
f (x1 +x2 )

ωL x2 2x3 (x12 +x22 ) ωL x 2x (x 2 +x 2 )

+ Cf (ωx4 − L1 x1 + L1 u 1 ) + C f 2 42 1 2 22 (−ωx3 − L1 x2 + L1 u 2 )
f (x12 +x22 )2 f f f (x1 +x2 ) f f
(9.109)

The previous relation is rewritten by regrouping terms which multiply the control
inputs u 1 and u 2 . Thus one obtains

( p f ẋ1 +q f ẋ2 )(x12 +x22 )−( p f x1 +q f x2 )(2x1 ẋ1 +2x2 ẋ2 )

ẍ1 = ω ẋ2 + 1
Cf (ωx4 − 1
Lf x1 ) − 1
Cf { (x12 +x22 )2
}

ωL f ẋ2 (x32 +x42 )(x12 +x22 )−ωL f x2 (x32 +x42 )(2x1 ẋ1 +2x2 ẋ2 )
−ω ẋ2 + 1
Cf { (x12 +x22 )2
}

1 1
ωL f x2 2x3 (ωx4 − L f x1 ) ωL f x2 2x4 (−ωx3 − L f x2 )
+ Cf · (x 2 +x 2 )
+ Cf · (x 2 +x 2 )
+
1 2 1 2

ωL x (2x ) ωL f x2 2x4
+ C1f { (x 2 +x
f 2
2 )L
3
+ 1
Lf }u 1 + 1
C f (x 2 +x 2 ) L f u2 (9.110)
1 2 f 1 2

The previous relation can be also written in the form

ẍ1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2 (9.111)

In a similar manner, by differentiating the second row of the state-space equations

with respect to time one has

1 ( p f ẋ2 −g f ẋ1 )(x1 +x2 )−( p f x2 −g f x1 (2x1 ẋ1 +2x2 ẋ2 ))

2 2
ẍ2 = −ω ẋ1 + 1
Cf ẋ4 − Cf (x 2 +x 2 )2
+
1 2

ωL f ẋ1 (x32 +x42 )(x12 +x22 )+ωL f x1 (2x3 ẋ3 +2x4 ẋ4 )(x12 +x22 )
+ω ẋ1 − 1
Cf { (x12 +x22 )2
− (9.112)

−ωL f x1 (x32 +x42 )(2x1 ẋ1 +2x2 ẋ2 )

(x12 +x22 )2
}

By substituting the derivative terms for x3 and x4 , according to the third and fourth
rows of the state-space equations one has
470 9 Differential Flatness Theory in Power Electronics

ẍ2 = −ω ẋ1 + 1
Cf (−ωx3 − 1
Lf x2 + 1
Lf u 2 )−

( p f ẋ2 −g f ẋ1 )(x12 +x22 )−( p f x2 −g f x1 )(2x1 ẋ1 +2x2 ẋ2 )

1 { 2 }+
Cf (x12 +x22 )

ωL f ẋ1 (x32 +x42 )(x12 +x22 )−ωL f x1 (x32 +x42 )(2x1 ẋ1 +2x2 ẋ2 )
+ ω ẋ1 − 1
Cf { (x12 +x22 )2
} (9.113)

ωL f x1 2x3 (x12 +x22 )

− C1f 2 (ωx4 − 1
Lf x1 + 1
Lf u 1 )−
(x12 +x22 )

ωL f x1 2x4 (x12 +x22 )

− C1f (x12 +x22 )2
(−ωx3 − 1
Lf x2 + 1
Lf u2)

By grouping the terms that multiply the control inputs u 1 and u 2 one obtains

( p f ẋ2 −g f ẋ1 )(x12 +x22 )−( p f x2 −g f x1 )(2x1 ẋ1 +2x2 ẋ2 )

ẍ2 = −ω ẋ1 + 1
Cf (−ωx3 ) − 1
Cf { (x12 +x22 )2
}+

ωL f ẋ1 (x32 +x42 )(x12 +x22 )−ωL f x1 (x32 +x42 )(2x1 ẋ1 +2x2 ẋ2 )
ω ẋ1 − 1
Cf { 2 }
(x12 +x22 )

ωL f x1 2x3 (ωx4 − L1 x1 ) ωL f x1 2x4 (−ωx3 − L1 x2 )

− C1f (x12 +x22 )
f
− 1
Cf (x12 +x22 )
f
−

ωL f x1 2x3 1 ωL f x1 2x4 1
− C1f (x12 +x22 ) L f
u1 − 1
Cf { (x12 +x22 ) L f
+ 1
Lf }u 2 (9.114)

The previous equation can be also written in the form

ẍ2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2 (9.115)

Thus, one obtains an input-output linearized description of the inverter in the form

ẍ1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

(9.116)
ẍ2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

The following control inputs are defined

v1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

(9.117)
v2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

Thus, the inverter’s dynamics becomes

ẍ1 = v1
(9.118)
ẍ2 = v2
9.3 Inverters Control 471

For this form of the system’s dynamics, the design of a state feedback controller is
easy. This takes the form:

v1 = ẍ1d − kd1 (ẋ1 − ẋ1d ) − k 1p (x1 − x1d )

(9.119)
v2 = ẍ2d − kd2 (ẋ2 − ẋ2d ) − k 2p (x2 − x2d )

The control input that is actually applied to the system is computed as:

ṽ = f˜ + M̃u (9.120)

or equivalently
   2   
v1 L f h 1 (x) L ga L f h 1 (x) L gb L f h 1 (x) u1
= + (9.121)
v2 L 2f h 2 (x) L ga L f h 2 (x) L gb L f h 2 (x) u2

which means that the control input that is finally applied to the system is

ũ = M̃ −1 (ṽ − f˜) (9.122)

To estimate the nonmeasurable state variables of the input-output linearized system

the Derivative-free nonlinear Kalman Filter can be used. By defining the new state
variables z 1 = x1 , z 2 = ẋ1 , z 3 = x2 and z 4 = ẋ2 , the following state-space
description of the system can be used
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 1 0 0 z1 0 0  
⎜ż 2 ⎟ ⎜0 0 0 0⎟ ⎜ z 2 ⎟ ⎜1 0⎟
⎜ ⎟=⎜ ⎟⎜ ⎟ + ⎜ ⎟ v1 (9.123)
⎝ż 3 ⎠ ⎝0 0 0 1⎠ ⎝z 3 ⎠ ⎝0 0⎠ v2
ż 4 0 0 0 0 z4 0 1
⎛ ⎞
    z1
m
z1 1000 ⎜ ⎟
⎜z 2 ⎟
= ⎝ (9.124)
z 2m 0010 z3⎠
z4

The associated state estimator has the form

żˆ = Aẑ + Bv + K f (z m − C ẑ) (9.125)

For the linearized equivalent of the system it is possible to perform state estimation
using the Derivative-free nonlinear Kalman Filter. The previously defined matrices
A, B, and C are substituted by their discrete-time equivalents Ad , Bd , and Cd . This
is done through common discretization methods. The recursion of the Kalman Filter,
as already described in Eqs. (4.5) and (4.6) is [31, 408, 414, 437]:
472 9 Differential Flatness Theory in Power Electronics

measurement update:

K f (k) = P − (k)CdT [Cd P − (k)CdT + R(k)]−1

x̂(k) = x̂ − (k) + K f (k)[Cd z(k) − Cd ẑ(k)] (9.126)
P(k) = P − (k) − K f (k)Cd P − (k)

time update:
P − (k + 1) = Ad P(k)AdT + Q(k)
(9.127)
x̂ − (k + 1) = Ad x̂(k) + Bd v(k)

The linearized equivalent model of the inverter can thus provide the state estimates
x̂1 , ẋˆ1 , x̂2 , and ẋˆ2 . Finally, using Eqs. (9.102) and (9.103), one can obtain estimates
for all state variables of the initial nonlinear model.

9.3.6 State and Disturbances Estimation with Nonlinear

Kalman Filtering

The previous estimator can be also redesigned in the form of a disturbance observer.
It is considered that the linearized model of the inverter is affected by additive input
disturbances. The disturbance terms can describe both modeling uncertainties and
external perturbations.
ẍ1 = v1 + d̃1
(9.128)
ẍ2 = v2 + d̃2

(n) (n)
These take the form d̃1 = f d1 and d̃2 = f d2 . The disturbances’ dynamics is
represented by the nth order derivative of the disturbances’ variables together with
the associated initial conditions (however, since the disturbances are going to be esti-
mated by the Kalman Filter, the knowledge of the initial conditions finally becomes
obsolete).
Without loss of generality it is assumed that n = 2. The state vector is extended by
including as additional state variables the disturbances and their derivatives. Thus,
one has z 1 = x1 , z 2 = ẋ1 , z 3 = x2 , z 4 = ẋ2 , z 5 = d̃1 , z 6 = d̃˙1 , z 7 = d̃2 and z 8 = d̃˙2 .
Therefore, one has the extended state estimator of the form

żˆ e = Ae ẑ e + Be v + K fe (z em − Ce ẑ e ) (9.129)

where
9.3 Inverters Control 473
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 0 00 10
⎜0 0 0 0 1 0 0 0⎟ ⎜1 0 ⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0 ⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 1 0⎟⎟ ⎜ ⎟
0 1⎟ T ⎜0 0⎟
Ae = ⎜
⎜0 ⎟ Be = ⎜
⎜ ⎟ C e = ⎜ ⎟
⎜0 0⎟ (9.130)
⎜ 0 0 0 0 1 0 0⎟ ⎜0 0 ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0⎟⎟ ⎜ ⎟ ⎜0 0⎟
⎜ ⎜0 0 ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 1 ⎠ ⎝ 00 ⎠ ⎝0 0⎠
0 0 0 0 0 0 0 0 00 00

For the extended state-space description of the system, it is possible again to perform
state estimation using the Derivative-free nonlinear Kalman Filter. The previously
defined matrices Ae , Be , and Ce are substituted by their discrete-time equivalents
Aed ,Bed and Ced . This is done again through common discretization methods. The
recursion of the Kalman Filter in this case is:
measurement update:

K f (k) = P − (k)CdT [Ced P − (k)CeTd + R(k)]−1

x̂(k) = x̂ − (k) + K f (k)[Ced z(k) − Ced ẑ(k)] (9.131)
P(k) = P − (k) − K (k)Ced P − (k)

time update:
P − (k + 1) = Aed P(k)AeTd + Q(k)
(9.132)
x̂ − (k + 1) = Aed x̂(k) + Bed v(k)

After identifying the disturbance terms, the control input of the inverter is modified
as follows:
v1 = ẍ1 − kd1 (ẋ1 − ẋ1d ) − k 1p (x1 − x1d ) − ẑ 5
(9.133)
v2 = ẍ2 − kd2 (ẋ2 − ẋ2d ) − k 2p (x2 − x2d ) − ẑ 7

9.3.7 Simulation Tests

The efficiency of the inverter’s nonlinear control scheme was tested in the tracking of
several setpoints. The associated results are presented in Figs. 9.9, 9.10, and 9.11. It
can be noticed that through the proposed control approach, fast and accurate tracking
of the voltage reference setpoints can be succeeded. This is shown in Figs. 9.9a
and 9.10a. Moreover, it can be noticed that the Derivative-free nonlinear Kalman
Filter enables precise estimation and compensation of the modeling uncertainties
and the external perturbation terms that affect the inverter’s model. This is shown in
Figs. 9.9b, 9.10, and 9.11b.
Using the state-space model of Eq. (9.64) as well as the linearized equivalent of
Eq. (9.128), the disturbance terms affecting the inverter’s model can be as follows:
(i) parametric uncertainty associated with the values of the capacitance C f and of the
inductance L f , (ii) variations of the grid’s frequency ω, (iii) uncertainty or variation
474 9 Differential Flatness Theory in Power Electronics

(a) (b)

d/dt d1 (p.u.)
1 1

d1 (p.u.)
1 1

d/dt VL (p.u.)
0.8 0.5 0.5
VL (p.u.)

0.5
0.6

d
0 0 0
0 10 20 30 40 0 10 20 30 40
d

0.4
−0.5
0.2 time time

d/dt d2 (p.u.)
1 1

d2 (p.u.)
0 −1
0 2 4 0 2 4
0.5 0.5
time time
1 1 0 0
0 10 20 30 40 0 10 20 30 40
d/dt V (p.u.)

0.5 0.5 time time

VL q (p.u.)

d/dt d3 (p.u.)
Lq

0 0 1 1

d3 (p.u.)
−0.5 −0.5 0.5 0.5
−1 −1
0 2 4 0 2 4 0 0
0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 9.9 a Convergence of the state vector elements VL ,d and VL ,q of the three-phase inverter to the
associated reference setpoints 1, b Estimation of disturbance inputs affecting the inverter’s model,
with the use of the Derivative-free nonlinear Kalman Filter

(a) (b)

d/dt d1 (p.u.)
1 1
d1 (p.u.)

1 1
d/dt VL (p.u.)

0.5 0.5
VL (p.u.)

0.8 0.5
0.6 0 0
0
d

0 10 20 30 40 0 10 20 30 40
d

0.4
0.2 −0.5 time time
d/dt d2 (p.u.)

1 1
d2 (p.u.)

0 −1
0 2 4 0 2 4
0.5 0.5
time time
0 0
1 1 0 10 20 30 40 0 10 20 30 40
d/dt VL (p.u.)

time time
VL (p.u.)

0.5 0.5
d/dt d3 (p.u.)

0 0 1 1
q

d3 (p.u.)
q

−0.5 −0.5 0.5 0.5

−1 −1 0 0
0 2 4 0 2 4 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 9.10 a Convergence of the state vector elements VL ,d and VL ,q of the three-phase inverter
to the associated reference setpoints 2. b Estimation of disturbance inputs affecting the inverter’s
model, with the use of the Derivative-free nonlinear Kalman Filter

of the active and reactive power coefficients p f and q f , respectively, (iv) additive
input disturbance terms affecting the control inputs Vid and Viq (these can be due to
parametric changes in the inverter’s circuit). All these perturbations are denoted by
the cumulative disturbance terms d̃1 and d̃2 which appear in Eq. (9.153). Therefore,
by estimating these disturbance inputs with the use of the Derivative-free nonlinear
Kalman Filter, it becomes possible to compensate for them and to assure the quality
of the AC power provided by the inverter either to the stand-alone load or to the grid.
9.3 Inverters Control 475

(a) 300 (b) 1

pf
0.9 qf

200
0.8
0.7
100

pf − qf (p.u.)
0.6
Vabc

0 0.5
0.4
−100
0.3
0.2
−200
0.1
−300 0
1 1.2 1.4 1.6 1.8 2 5 10 15 20 25 30 35 40
time (sec) t

Fig. 9.11 a Three-phase voltage variables at the output of the inverter, b Active and reactive power
of the inverter

In the simulation experiments, the disturbance inputs d̃1 and d̃2 of Eq. (9.153)
were due to (i), (iii), and (iv) but it is also possible to obtain similar results in case
that disturbance in the grid’s frequency ω is introduced. In all cases, the proposed
control scheme is proven to be robust and assures convergence of the output voltage
to the desirable setpoints. The disturbance inputs contained also sinusoidal terms
of variable amplitude and frequency. It was confirmed that in all cases the control
scheme succeeded elimination of the setpoints’ tracking error.

9.4 Distributed Inverters Synchronization

9.4.1 Overview

The previous results on control and filtering for stand-alone inverters are generalized
towards flatness-based control and filtering for synchronized distributed inverters.
The integration of distributed DC power generation units in microgrids, including
for instance photovoltaics and fuel cells source, requires the interfacing of such
devices with the electricity network through inverters. Another problem that arises
is the synchronization between multiple parallel inverters. There has been signifi-
cant effort in solving control and synchronization problems for voltage inverters, so
as to enable connection of a large number of power sources to the grid. In [226]
a state-space model is developed for the dynamics of distributed power generation
units connected to the grid through inverters. Stability analysis is performed and a
simple state feedback control scheme is proposed. In [310] decentralized frequency
control of inverters is proposed. Information about the total active power deficiency
of the microgrid is shared between the local units. Frequency control is performed to
476 9 Differential Flatness Theory in Power Electronics

achieve a cooperative frequency recovery. In [116], a Lyapunov theory-based feed-

back controller is proposed for voltage inverters connected to the grid. In [182],
stability of parallel DC sources connected to the grid through inverters is studied and
simple feedback control is proposed using the equivalence of the inverters’ model
to that of a synchronous generator. In [223], it is considered that each inverter is
equivalent to an electronic circuit with oscillatory dynamics and thus the problem
of control of parallel inverters becomes a problem of synchronization of coupled
electrical oscillators. In [328], Lie algebra is used to transform the nonlinear model
of three-phase photovoltaic units into linear systems and based on the linearized
description a state feedback controller is designed. In [606], the concept of syn-
chronverters is introduced. This is based on the equivalence between the voltage
inverter’s model and that of a synchronous generator. It is shown that the controller
is in principle a power controller with integrated capability of voltage and frequency
regulation. In [45] Lie algebra-based input-output feedback linearization is used to
solve the synchronization problem for nonlinear inverters connected in parallel to
the grid. It is considered that the power exchange of each local inverter towards the
grid is modeled equivalently by the dynamics of a synchronous generator. In [473] a
feedback control and synchronization method for distributed DC power generation
units is developed using the equivalence between the inverter’s model and the model
of the synchronous power generator. In [474, 475], stability conditions are provided
for distributed DC power generation units using a Lyapunov function approach.
By solving the problem of synchronization between DC distributed power gen-
eration units which are connected to the microgrid through voltage inverters a more
efficient operation of the power grid can be succeeded in terms of power quality and
stability. To this end, this section is proposing a nonlinear feedback control method
for parallel inverters which is based on differential flatness theory. Moreover, it is
proposing the previously analyzed Derivative-free nonlinear Kalman Filtering which
enables to estimate the nonmeasurable state vector elements of each local inverter, as
well as disturbance terms affecting each inverter (the latter can be due to their inter-
action with other power generation units in the grid or due to external perturbations
and modeling uncertainty) [427, 453, 456].
The electrical model of the stand-alone inverter and its differential flatness prop-
erties were analyzed in Sect. 9.3. In this section, the problem of control and synchro-
nization of parallel inverters connected to the grid is examined. As far as the power
exchange with the grid is concerned, it is shown that the dynamics of each inverter
can be written in a form that is equivalent to the model of the synchronous power
generator. Using the latter description, one can compute the active power that each
local inverter should inject to the network so as to remain synchronized with the rest
of the grid. Next, the frequency of this virtual generator model is used to compute the
output voltage setpoints for the electrical part of each local inverter. Thus, finally the
synchronization problem of each local inverter is turned into a problem of nonlinear
feedback control for the associated inverter’s electrical model. It is also shown that
in the proposed decentralized control scheme each local controller processes mea-
surements not only coming from the individual inverter but also coming from other
inverters which are connected to the grid.
9.4 Distributed Inverters Synchronization 477

9.4.2 The Synchronization Problem for Parallel Inverters

9.4.2.1 Equivalence Between Inverters and Synchronous Generators

Feedback linearizing control for the inverter, with the use of differential flatness
theory has been analyzed in the previous sections. Next, the problem of synchro-
nization of distributed inverters will be solved, making again use of filtering and
control methods which are based on differential flatness theory.
By applying appropriate control inputs, inverters can be made to function in a
manner equivalent to synchronous generators. Actually, with the regulation of the
active power produced by the inverters one can succeed the synchronization between
the voltage output of the inverters and the voltage of the grid [607, 608, 610]. This
synchronization enables better integration of DC renewable energy units to the grid.
The functioning of the ith inverter is shown to be equivalent to a synchronous
generator with turn speed denoted as ωi [45, 473, 475, 606, 607, 609]. In this modeling
approach one has
Δδi = ωi − ωd = −k pi (Pim − Pid ) (9.134)

where Pim is the measured active power of the ith power generation unit and Pid
is the desirable active power. Coefficient k pi is a “droop” gain which is practically
computed by dividing the range of variation of the inverter’s frequency (ωmax −ωmin )
by the maximum active power Pimax that the inverter can produce. Equation (9.134)
denotes that when representing the inverter as an equivalent synchronous generator,
the turn speed should be proportional to the active power that the inverter provides.
Next, a relation is obtained for the measured active power Pim and the real active
power Pi of the inverter Pi . It holds that Pim (s) = e−sτ pi P(s) (delay relation in
frequency domain) which after intermediate operations gives the differential equation

τ pi Ṗim = −Pim + Pi (9.135)

Thus the inverter’s dynamics is expressed as

Δδ̇i = Δωi
(9.136)
τ pi Ṗim = −Pim + Pi

By deriving Eq. (9.134) with respect to time one obtains

Δω̇i = −k pi Ṗim + k pi Ṗid (9.137)

Moreover, by solving Eq. (9.135) with respect to Ṗim one has

Ṗim = − τ1p Pim + 1

τ pi Pi (9.138)
i
478 9 Differential Flatness Theory in Power Electronics

By substituting Eq. (9.138) into Eq. (9.137) the following relation is obtained

Δω̇i = −k pi (− τ1p Pim + 1

τ pi Pi ) + k pi Ṗid (9.139)
i

and by considering that Ṗid = 0 one gets

k pi k pi
Δω̇i = τ pi Pim − τ pi Pi (9.140)

or equivalently
Ji Δω̇i = Pim − Pi (9.141)

where the virtual moment of inertia is defined as Ji = τ pi /k pi . Additionally, from

Eq. (9.134) one has

ωi − ωd − k pi Pid = −k pi Pim ⇒Pim = − k1p ωi + 1

k pi ωd + Pid ⇒
i (9.142)
Pim = − k1p Δωi + Pid
i

By substituting Eq. (9.142) into Eq. (9.141) the following relation is obtained

Ji Δω̇i = − k1p Δωi + Pid − Pi (9.143)

i

The damping coefficient D pi = k1p is defined. Using this coefficient in Eq. (9.142)
i
gives
Ji Δω̇i = −D pi Δωi + Pid − Pi (9.144)

which is the equation that describes the rotational motion of a synchronous genera-
tor. In ideal conditions there is no interaction (power exchange) between distributed
power units connected to the same electricity grid. However, frequently such inter-
action exists and in the latter case Eq. (9.144) should be enhanced by including an
interaction term
n
Ji Δω̇i = −D pi Δωi + (Pid − Pi ) + j=1, j=i G i j sin(δi − δj) (9.145)

where δi is the virtual turn angle that is associated with the ith power generation unit
(inverter). About the coupling coefficients G i j these are functions of the conductance
of the grid line which connects the ith to the jth power generation unit, as well as
of the grid voltage that is measured at points i and j, respectively.
Thus, finally the dynamics of the ith power generation unit is described as

Δδ̇i (t) = Δωi (t)

(9.146)
Ji Δω̇i (t) = −D pi Δωi (t) + (Pid (t) − Pi (t)) + Nj=1 G i j sin(δi − δ j )
9.4 Distributed Inverters Synchronization 479

In the design of the control and synchronization system for the distributed power
generation units, it is considered that the parameters Ji , D pi , and G i j are either
known or can be computed from measurements. Moreover, it is considered that the
ith local controller not only processes measurements coming from the associated
power generation unit, but also uses measurements coming from the other power
units which are connected to the grid (that is the virtual turn angles δ j ).

9.4.2.2 Control for Parallel Inverters Connected to the Grid

In Eq. (9.146), it has be shown that under certain conditions, the dynamics of the
inverter becomes equivalent to that of the synchronous power generator. Then one has
that the dynamics of the inverter is composed of two parts (i) the rotation (mechanical)
part and (ii) the electrical part.
The rotation part is given by:

Δδ̇i (t) = Δωi (t)

Ji Δω̇i (t) = −D pi Δωi (t) + (Pid (t) − Pi (t)) + Nj=1, j=i G i j sin(δi − δ j )
(9.147)
Ji is the virtual inertia, D pi is the virtual damping coefficient, Pid is the desirable
active power of the ith unit and P i and the measured active power, G i j is a coupling
coefficient between the ith and the jth power generation unit and sin(δi − δ j ) is a
synchronization index between the ith and the jth power generation unit [45, 473,
474, 475, 606, 607]. The electrical part has been defined in Sect. 9.3.2 and is given
by:
⎛ ⎞
ωL f VL q (i I2 +i I2q )
1 1 p f VL d +q f VL q + ωC V
⎜ ωVL q + C f i Id − C f V 2 +V 2 f Lq −
d
⎛ ⎞ (VL2 +VL2q ) ⎟
VL d ⎜ L Lq ⎟
⎜ d
ωL f VL d (i I2 +i I2q ) ⎟
d
⎜ ⎟ ⎜
d ⎜ VL q ⎟ = ⎜−ωVL + 1 i I − 1
p f VL q −q f VL d ⎟
− ωC f VL d + d
⎟+
dt ⎝ i I ⎠ ⎜ d Cf q Cf VL2 +VL2q (VL2 +VL2q ) ⎟
d ⎜ d d ⎟
i Iq ⎜ ωi − 1 V ⎟
⎝ Iq L f Ld ⎠
1
−ωi Id − L VL q
f

⎛ ⎞
0 0
⎜ 0 0 ⎟ 
⎜ ⎟ V Id
+⎜ 1 0 ⎟ (9.148)
⎝Lf ⎠ V Iq
0 L1f

The synchronizing control approach for the ith inverter makes use of Eq. (9.147) and
of the linearized inverter model given in Eq. (9.148). First, the value of Pi , that is the
active power that the ith inverter should inject to the grid, is found from the solution
of the control problem of Eq. (9.147). Subsequently, Pi is used in the computation
of the solution of the control problem for the system of Eq. (9.148). This is depicted
in Fig. 9.12.
480 9 Differential Flatness Theory in Power Electronics

Fig. 9.12 Control loops for the virtual synchronous generator model and for the electrical part of
the inverter

The synchronizing control method can be generalized to the case of multiple

inverters connected to the grid (see Fig. 9.13). It can be shown again that the model
of N distributed power units which are connected to the grid through inverters is also
differentially flat. The latter is proven following the procedure described in Sect. 9.3.4
and by defining as flat output a generalization of the vector described in Eq. (9.88),
which now becomes

Y = [y11 , y21 , y31 , y12 , y22 , y32 . . . , y1N , y2N , y3N ] or

Y = [δ 1 , VL1d , VL1q , δ 2 , VL2d , VL2q , . . . , δ N , VLNd , VLNq ]

All state variables and control inputs for the model of the N coupled inverters can
be expressed as functions of the aforementioned flat output Y and of its derivatives.
Using the previous flat output definition, and the state variables z 1i = y1 , z 2i = ẏ1 ,
z 3i = y2 , z 4i = ẏ2 , z 5i = y3 , z 6i = ẏ3 , one has:
⎛ i⎞ ⎛ ⎞⎛ i ⎞ ⎛ ⎞
ż 1 0 1 0 0 0 0 0 z1 0 0 0
⎜ż i ⎟ ⎜0 0 0 0 0 0 0⎟ ⎜ z i ⎟ ⎜1 0 0⎟ ⎛ ⎞
⎜ 2i ⎟ ⎜ ⎟ ⎜ 2i ⎟ ⎜ ⎟ vi
⎜ż ⎟ ⎜0 0 0 1 0 0 0⎟ ⎜ ⎟ ⎜ 0⎟ 1
⎜ 3⎟ = ⎜ ⎟ ⎜z 3 ⎟ + ⎜0 0 ⎟ ⎝vi ⎠ (9.149)
⎜ż i ⎟ ⎜0 0 0 0 0 0 0⎟ ⎜ i⎟ ⎜ 0⎟ 2
⎜ 4⎟ ⎜ ⎟ ⎜z 4 ⎟ ⎜0 1 ⎟ vi
⎝ż i ⎠ ⎝0 0 0 0 0 1 0⎠ ⎝z 5i ⎠ ⎝0 0 0⎠ 3
5
ż 6i 0 0 0 0 0 0 0 z 6i 0 0 1
9.4 Distributed Inverters Synchronization 481

Fig. 9.13 Distributed DC power generation units connected through inverters to the grid

where the control inputs of this model are defined as

N
v1i = 1
Ji [−D pi Δωi (t) + (Pid (t) − Pi (t)) + j=1 G i j sin(δi − δ j )] (9.150)

and using Eq. (9.117) it holds for the ith inverter’s model

v2i = L 2f h 1 i (x) + L ga L f h i1 (x)u i1 + L gb L f h i1 (x)u i2

(9.151)
v3i = L 2f h 2 i (x) + L ga L f h i2 (x)u i1 + L gb L f h i2 (x)u i2

If the coupling coefficients G i j , i = 1, . . . , n, j = 1, . . . , n are known then the

solution of the control problem for Eq. (9.147) can be obtained, by communicating
to each power generation unit i the rotation angles of the rest N −1 power generation
units so as to compute the sum Nj=1 G i j sin(δi −δ j ). In such a case, and by denoting
x1,i = δi , x2,i = ωi the control input (power) applied to the ith power generation
unit is
d − D x + P d + N G sin(x
Pi = −Ji ẍ1,i p1 2 i j=1 i j 1,i − x 1, j )+
(9.152)
+Ji K di (ẋ1,i − ẋ1,i
d ) + J K (x
i pi 1,i − x d )
1,i
482 9 Differential Flatness Theory in Power Electronics

9.5 State and Disturbances Estimation of Parallel Inverters

with Nonlinear Kalman Filtering

A state estimator for each local power generation unit can be also designed in the
form of a disturbance observer. It is considered that the linearized model of the ith
inverter is affected by additive input disturbances (which are taken to be modeling
uncertainties and external perturbations such as load changes, power injected due to
the connection of other power generation units to the grid, faults in the grid etc.).
With reference to the state-space description given in Eq. (9.149), the disturbance
terms can describe both modeling uncertainties and external perturbations.

z̈ 1i = v1i + d̃1i
z̈ 3i = v2i + d̃2i (9.153)
z̈ 5i = v3i + d̃3i

The disturbances’ dynamics can be represented by the nth order derivative of the
disturbances’ variables together with the associated initial conditions (however, since
the disturbances are going to be estimated by the Kalman Filter, the knowledge of
the initial conditions finally becomes obsolete). Thus the additive disturbances are
equivalently described in the form d̃1(n) = f d1 , d̃2(n) = f d2 and d̃3(n) = f d3 .
(n)
Without loss of generality it is assumed that in the relation d̃i = f di , the deriv-
ative’s order is n = 2. The state vector is extended by including as additional state
variables the disturbances and their derivatives. Thus, one has z 1i = x1 , z 2i = ẋ1 ,
z i = x , z i = ẋ , z i = x , z i = ẋ , z i = d̃ , z i = d̃˙ , z i = d̃ , z i = d̃˙ , z i = d̃
3 2 4 2 5 3 6 3 7 1 8 1 9 2 10 2 11 3
and z 12 = d̃˙3 .
i

It holds that ż 1i = z 2i , ż 2i = z 7 + v1i , ż 3i = z 4i , ż 4i = z 9 + v2i , ż 5i = z 6i , ż 6i = z 11

i + vi ,
3
ż 7 = z 8 , ż 8 = f d1 , ż 9 = z 10 , ż 10 = f d2 , ż 11 = z 12 , ż 12 = f d3 . Therefore, one has
i i i i i i i i i

the system’s dynamics in the extended spate-space form

ż e = Ae z e + Be ve (9.154)

where the extended control inputs vector is ve = [v1i , v2i , v3i , f d1 , f d2 , f d3 ]T and
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
010000000000 000000 100
⎜0 0 0 0 0 0 1 0 0 0 0 0 ⎟ ⎜1 0 0 0 0 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0 0 0 0 0 ⎟ ⎜0 0 0 0 0 0 ⎟ ⎜0 1 0 ⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 1 0 0 0 ⎟ ⎜0 1 0 0 0 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 1 0 0 0 0 0 0 ⎟ ⎜0 0 0 0 0 0 ⎟ ⎜0 0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 1 0 ⎟ ⎜ ⎟ ⎜ ⎟
Ae = ⎜ ⎟ Be = ⎜0 0 1 0 0 0⎟ C T = ⎜0 0 0⎟ (9.155)
⎜0 0 0 0 0 0 0 1 0 0 0 0 ⎟ ⎜0 0 0 0 0 0 ⎟ e ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 1 0 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 1 0 0 ⎟ ⎜0 0 0 0 0 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 1 0 ⎟ ⎜0 0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 0 0 0 0 1⎠ ⎝0 0 0 0 0 0 ⎠ ⎝0 0 0⎠
000000000000 000001 000
9.5 State and Disturbances Estimation of Parallel Inverters … 483

For the extended state-space description of the system, the state observer becomes

ẑ˙ e = Ao ẑ e + Bo ve + K f (Co z e − Co ẑ e ) (9.156)

where for matrices Ae and Ce holds Ao = Ae and Co = Ce while for matrix Bo

differs from Be in the elements of the 10th and 12th row, where all elements are set
to zero.
For the extended state-space description of the system it is possible to perform state
estimation using the Derivative-free nonlinear Kalman Filter. As already mentioned,
this filter consists of the standard Kalman Filter recursion on the linearized equivalent
model of the ith inverter, and of an inverse transformation based on differential
flatness theory (as explained in Sect. 9.3.4) which enables to obtain estimates for the
state variables of the initial nonlinear model of the inverter (if the latter is necessary).
In the filter’s algorithm, the previously defined matrices Ae , Be , and Ce are substituted
by their discrete-time equivalents Aed , Bed , and Ced . This is done through common
discretization methods. The recursion of the Kalman Filter in this case is [31, 408,
414, 456]:
measurement update:

K f (k) = P − (k)CdT [Ced P − (k)CeTd + R(k)]−1

x̂(k) = x̂ − (k) + K f (k)[Ced z(k) − Ced ẑ(k)] (9.157)
P(k) = P − (k) − K f (k)Ced P − (k)

time update:
P − (k + 1) = Aed P(k)AeTd + Q(k)
(9.158)
x̂ − (k + 1) = Aed x̂(k) + Bed v(k)

After identifying the disturbance terms, the control input of the inverter is modified
as follows:
v1i = z̈ 1 − kd1 (ż 1 − ż 1d ) − k 1p (z 1 − z 1d ) − ẑ 7
v2i = z̈ 3 − kd2 (ż 3 − ż 3d ) − k 2p (z 3 − z 3d ) − ẑ 9 (9.159)
v3i = z̈ 5 − kd3 (ż 5 − ż 5d ) − k 3p (z 5 − z 5d ) − ẑ 11

The inclusion of the disturbance estimation terms ẑ 7 , ẑ 9 , and ẑ 11 in the feedback

control inputs enables to compensate for effects of the perturbations d̃1 , d̃2 and d̃3 .

9.6 Simulation Tests

The performance of the proposed distributed control scheme for the synchronization
of parallel inverters was tested through simulation experiments. A model of N = 3
distributed power generation units was considered, while each one of these units
was connected to the grid through an inverter (see Fig. 9.14). The power exchange
between each inverter and the grid was considered to be described by the model of
a synchronous generator.
484 9 Differential Flatness Theory in Power Electronics

Fig. 9.14 Distributed DC power generation units connected through inverters

Table 9.1 Parameters of the Inv1 Inv2 Inv3

inverters
L f (mH) 10.5 10.3 10.1
C f (mF) 0.04 0.03 0.02
p f (kW) 21.1 22.3 23.6

The three interconnected inverters, shown in Fig. 9.14, are assumed to have differ-
ent model parameters which are described in Table 9.1. Moreover, the three invert-
ers are considered to be subjected to different perturbation inputs. Synchronization
means that given the desirable reference rotation speed ωi , one can compute the
amount of active power Pi contributed to the grid by each local inverter which in
turn results into ωi . Next, knowing Pi one can compute for the inner control loop of
Eq. (9.148) the associated voltage reference setpoints VL d and VL q and can solve the
control problem for the electrical part of the inverter’s dynamics. This means that
the synchronization problem of each local inverter is finally turned into a problem
of nonlinear feedback control for the associated electrical model of the inverter.
The main relation that the chapter considers for synchronization of the inverters
with a reference frequency of the grid is Eq. (9.147). According to this relation, the
objective is that all inverters finally attain the same frequency ωi . By knowing the
active power Pi which results into virtual rotation speed ωi , one can compute also the
setpoints for the output voltages VL d and VL q (see Fig. 9.12) which have to be used
in the control problem of the inverter’s electrical part described by Eq. (9.148). One
9.6 Simulation Tests 485

can compute these setpoints through Eqs. (9.60) and (9.61). In Eq. (9.60) the active
power should be given the value that is computed from the solution of the control
problem for the virtual synchronous generator model of Eq. (9.147). Additionally,
in Eq. (9.61) a reference value for the reactive power can be used. The currents
i L ,d and i L ,q are considered to be measurable. Thus one finally has a set of two
equations with unknowns the output voltages VL d and VL q . Solving this system
with respect to VL d and VL q provides reference setpoints for the output voltages
which finally result in the synchronization of the inverters with the grid. Moreover,
to estimate the nonmeasurable state variables in the inverter models as well as to
estimate and compensate for external perturbations affecting the inverter’s dynamics
the Derivative-free nonlinear Kalman Filter has been used.
The obtained results are depicted in Figs. 9.15, 9.16, 9.17, 9.18, 9.19, 9.20, 9.21,
9.22 and 9.23. The state variables of the distributed power generators are actually
measured in SI units. However, in the simulation experiments presented in Figs. 9.15,
9.16, 9.17, 9.18, 9.19, 9.20, 9.21, 9.22 and 9.23, these variables have been measured
in the per unit (p.u.) system. It can be noticed that the proposed control and state
estimation scheme succeeded both satisfactory transients and good tracking perfor-
mance of the reference setpoints. Moreover, fast synchronization between the dis-
tributed power generation units was succeeded. It is noted that the proposed control
and synchronization approach is scalable and can be applied to a larger number of
N interconnected DC power generation units. The simulation diagrams show con-
vergence of the virtual turn speed ωi of each inverter, to the associated reference
value ωi∗ .
Finally, it is noted that the section’s results confirm the usefulness of the synchron-
verters concept for control and stabilization of distributed power generation units,
e.g., in the case of renewable energy systems [606]. The methodology developed

(a) 1.5 (b) 1

0.8
sync error DG1 −DG2 (p.u.)

0.6
0.4
Gen 1 ω (p.u.)

1
0.2
0
−0.2
0.5
−0.4
−0.6
−0.8
0 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 9.15 a Angular speed of power generation unit 1 (red line setpoint, blue line real value), b
synchronization error between power generation units 1 and 2
486 9 Differential Flatness Theory in Power Electronics

(a) (b)

d/dt d1 (p.u.)
1 1

d1 (p.u.)
1 1

d/dt VL (p.u.)
0.8 0.5 0.5 0.5
VL (p.u.)

0.6
0 0

d
0
0.4 0 10 20 30 40 0 10 20 30 40
d

−0.5
0.2 time time

d/dt d3 (p.u.) d/dt d2 (p.u.)

0 −1 1 1

d2 (p.u.)
0 2 4 0 2 4
0.5 0.5
time time
1 1 0 0
d/dt VL q (p.u.)
0 10 20 30 40 0 10 20 30 40
VL q (p.u.)

0.5 0.5
time time
0 0 1 1

d3 (p.u.)
−0.5 −0.5 0.5 0.5
−1 −1
0 2 4 0 2 4 0 0
0 10 20 30 40 0 10 20 30 40
time time
time time

Fig. 9.16 Inverter of power generation unit 1 (a) Voltage vector components VL d and VL q and their
derivatives, (red line setpoint, blue line real value, green line estimated value) (b) estimation (green
line) of disturbance inputs (blue line) and of their derivatives with the use of the Derivative-free
nonlinear Kalman Filter

(a) 300 (b) 1

pf
0.9 qf

200
0.8
0.7
100
pf − qf (p.u.)

0.6
Vabc

0 0.5
0.4
−100
0.3
0.2
−200
0.1
−300 0
1 1.2 1.4 1.6 1.8 2 5 10 15 20 25 30 35 40
time (sec) t

Fig. 9.17 Inverter of power generation unit 1 (a) Three-phase voltage variables VL , (red line Va ,
blue line Vb , green line Vc ) (b) active and reactive power of the inverter

in this chapter enables a power corporation to control a synchronverter in the same

way as controlling synchronous generators, which considerably facilitates the con-
nection of renewable energy sources to the electricity network and the deployment
of the smart grid.
Finally, in Table 9.2, the RMSE of setpoint tracking by the state variables ω, VL d ,
VL q of the distributed inverters j = 1, 2, 3 is presented. It can be noticed that the
proposed control and state estimation scheme assures good tracking performance.
9.6 Simulation Tests 487

(a) 1.5 (b) 1

0.8

sync error DG2 −DG3 (p.u.)

0.6
0.4
Gen 2 ω (p.u.)

1
0.2
0
−0.2
0.5
−0.4
−0.6
−0.8
0 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 9.18 a Angular speed of power generation unit 2 (red line setpoint, blue line real value), b
synchronization error between power generation units 2 and 3

(b)

d/dt d1 (p.u.)
(a) 1 1
d1 (p.u.)

1 1
d/dt VL d (p.u.)

0.8 0.5 0.5

VL d (p.u.)

0.5
0.6 0 0
0 0 10 20 30 40 0 10 20 30 40
0.4
0.2 −0.5 time d/dt d2 (p.u.) time
0 −1 1 1
d2 (p.u.)

0 2 4 0 2 4
time time 0.5 0.5

1 1 0 0
d/dt VL q (p.u.)

0 10 20 30 40 0 10 20 30 40
0.5 0.5
VL q (p.u.)

time time
0 0
d/dt d3 (p.u.)

1 1
d3 (p.u.)

−0.5 −0.5
0.5 0.5
−1 −1
0 2 4 0 2 4
0 0
time time 0 10 20 30 40 0 10 20 30 40
time time

Fig. 9.19 Inverter of power generation unit 2 (a) Voltage vector components VL d and VL q and their
derivatives, (red line setpoint, blue line real value, green line estimated value) (b) estimation (green
line) of disturbance inputs (blue line) and of their derivatives with the use of the Derivative-free
nonlinear Kalman Filter

The presented simulation experiments demonstrated the efficiency of the control

method in tracking rapidly changing reference setpoints while also succeeding good
transients. The disturbances appearing in the simulation experiments could be met in
adverse operating conditions of the distributed power generation system. Even for the
latter case, the good performance of the control loop is confirmed. Such disturbances
can be due to modeling errors (e.g., parametric changes in the inverters’ model) or
488 9 Differential Flatness Theory in Power Electronics

(a) 300 (b) 1

pf
0.9 qf

200
0.8
0.7
100

pf − qf (p.u.)
0.6
Vabc

0 0.5
0.4
−100
0.3
0.2
−200
0.1
−300 0
1 1.2 1.4 1.6 1.8 2 5 10 15 20 25 30 35 40
time (sec) t

Fig. 9.20 Inverter of power generation unit 2 (a) Three-phase voltage variables VL , (red line Va ,
blue line Vb , green line Vc ) (b) active and reactive power of the inverter

(a) 1.5 (b) 1

0.8
sync error DG3 −DG1 (p.u.)

0.6
0.4
Gen 3 ω (p.u.)

1
0.2
0
−0.2
0.5
−0.4
−0.6
−0.8
0 −1
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 9.21 a Angular speed of power generation unit 3 (red line setpoint, blue line real value), b
synchronization error between power generation units 3 and 1

due to external perturbations (e.g., grid faults or disturbances due to the connection
or disconnection from the grid of power generation units).
It is noted that a microgrid comprising DC power generation units and the associ-
ated inverters can function in two modes, that is the isolated (stand-alone) mode and
the grid-connected mode. The chapter’s results are not constrained to the case of an
isolated microgrid, but one can consider also the case of multiple microgrids con-
nected simultaneously to the main electricity network [46, 603]. In the latter case, the
control system for the distributed inverters can be designed considering as common
dq reference frame the one of the main grain. Thus the frequencies of the voltages
9.6 Simulation Tests 489

(a) (b)

d/dt d1 (p.u.)
1 1 1 1

d1 (p.u.)
d/dt VL d (p.u.)
VL d (p.u.)

0.8 0.5 0.5 0.5

0.6
0 0 0
0.4 0 10 20 30 40 0 10 20 30 40
0.2 −0.5
time time
0 −1

d/dt d2 (p.u.)
0 2 4 0 2 4 1 1

d2 (p.u.)
time time 0.5 0

1 1 0 −1
0 10 20 30 40 0 10 20 30 40
d/dt VL q (p.u.)
VL q (p.u.)

0.5 0.5 time time

d/dt d3 (p.u.)
0 0 1 1

d3 (p.u.)
−0.5 −0.5
0.5 0.5
−1 −1
0 2 4 0 2 4 0 0
0 10 20 30 40 0 10 20 30 40
time time
time time

Fig. 9.22 Inverter of power generation unit 3 (a) Voltage vector components VL d and VL q and their
derivatives, (red line setpoint, blue line real value, green line estimated value) (b) estimation (green
line) of disturbance inputs (blue line) and of their derivatives with the use of the Derivative-free
nonlinear Kalman Filter

(a)300 (b) 1
pf
0.9 qf

200
0.8
0.7
100
pf − qf (p.u.)

0.6
Vabc

0 0.5
0.4
−100
0.3
0.2
−200
0.1
−300 0
1 1.2 1.4 1.6 1.8 2 5 10 15 20 25 30 35 40
time (sec) t

Fig. 9.23 Inverter of power generation unit 3 (a) Three-phase voltage variables VL , (red line Va ,
blue line Vb , green line Vc ) (b) active and reactive power of the inverter

Table 9.2 RMSE for the RMSE1 RMSE2 RMSE3

distributed inverters
ω 0.0225 0.0427 0.0199
VL d 0.0180 0.0008 0.0003
VL q 0.0246 0.0020 0.0010
490 9 Differential Flatness Theory in Power Electronics

at the output of the individual inverters, in each microgrid, are finally synchronized
to the frequency of the main grid.
Control of distributed inverters finally takes the form of a networked control prob-
lem [382, 471, 483]. The information exchanged between the inverters describes often
the inverter’s power ratings and is used for the management of power flow from the
individual inverters to the main grid. In the present chapter, the exchanged informa-
tion describes the virtual turn angles of the inverters (if the inverters are viewed as
equivalents to synchronous generators). This information is used for synchronizing
the frequency of the inverters’ output voltage to that of the main grid. It is also noted
that the compensation of communication delays is significant for implementing con-
trol of the distributed power generation system over a communication network. It has
been shown that the Derivative-free nonlinear Kalman Filter can be also the basis for
eliminating the effects of data transfer delays and data losses in networked control
schemes [450]. Actually, this requires redesign of the Kalman Filter recursion with
the inclusion of a modified Riccati equation for the computation of the covariance
matrix of the state vector estimation error, as well as the inclusion of a smoothing
term in the update of the state vector estimate.
Finally, it is worth mentioning that in the design of the distributed inverters control
loop, robustness remains a primary objective [295, 585]. For the method presented
in the previous sections, it is noted first that the linearized equivalent model of the
inverter given in Eq. (9.118) exhibits multiple poles at the origin. This particular form
implies an infinite gain margin and a sufficiently large phase margin. Next, it is noted
that the linearized equivalent model of the ith inverter, after application of the pole
placement technique of Eq. (9.119) has poles, exclusively in the left complex semi-
plane. Besides, the inclusion of the additional control input that compensates for the
estimated additive disturbance terms improves the robustness features of this control
loop. Finally, it is noted that the stability and robustness features of the control scheme
which comprises also estimation and compensation of the disturbances are similar
to those of LQG control. The presented simulation experiments have confirmed the
efficiency of the control method in tracking rapidly changing reference setpoints
while also succeeding good transients, even in the case of grid-induced disturbances.
Chapter 10
Differential Flatness Theory for Internal
Combustion Engines

10.1 Overview

In this chapter results will be presented about differential flatness theory-based con-
trol and filtering for combustion engines. The first section of the chapter studies
robust nonlinear control for gas exchange valves in ship diesel engines, with the
use of the Derivative-free nonlinear Kalman Filter. Robust control of gas exchange
valves is important for improving the efficiency of diesel engines. Due to the rela-
tively wide range of typical internal combustion engine operating conditions imposed
on a variable valve timing actuator, the control scheme must compensate both for
model uncertainties, parametric variations, and rapid external disturbances. By apply-
ing differential flatness theory the initial nonlinear model of the valve’s actuation is
transformed into the linear canonical (Brunovsky) form. For the latter model it is pos-
sible to design a state feedback controller that enables accurate tracking of the valve’s
reference setpoints. To estimate the nonmeasurable state variables of the model and
the unknown external disturbances, the Derivative-free nonlinear Kalman Filter is
used as a disturbance observer. As explained in previous chapters, the Derivative-
free nonlinear Kalman Filter consists of the standard Kalman Filter recursion on the
linearized equivalent model of the valve and of computation of state and disturbance
estimates using a diffeomorphism (relations about state variables transformation)
provided by differential flatness theory. Evaluation tests are performed for assessing
the performance of the proposed control scheme.
Next, the chapter continues by showing how differential flatness theory-based
filtering and control methods can be used for developing embedded control systems
for the transportation industry and particularly for turbocharged diesel engines. It is
shown that the dynamic model of the turbocharged diesel engine is differentially flat
and admits dynamic feedback linearization. It is also shown that the dynamic model
can be written in the linear Brunovsky canonical form for which a state feedback
controller can be easily designed. To compensate for modeling errors and external
disturbances the Derivative-free nonlinear Kalman Filter is used and redesigned as

© Springer International Publishing Switzerland 2015 491

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_10
492 10 Differential Flatness Theory for Internal Combustion Engines

a disturbance observer. The filter consists of the Kalman Filter recursion on the
linearized equivalent of the diesel engine model and of an inverse transformation
based on differential flatness theory which enables to obtain estimates for the state
variables of the initial nonlinear model. Once the disturbance variables are identified
it is possible to compensate for them by including an additional control term in
the feedback loop. The efficiency of the proposed control method is tested through
simulation experiments.
In the following section, the problem of nonlinear control for turbocharged diesel
engines is revisited, and this time is solved with observer-based adaptive fuzzy control
which makes use of differential flatness theory. As explained, the dynamic model
can be written in the linear Brunovsky canonical form for which a state feedback
controller can be easily designed. To compensate for modeling errors and external
disturbances an adaptive fuzzy control scheme is implemented making use of the
transformed dynamical system of the diesel engine that is obtained through the
application of differential flatness theory. The control algorithm aims at satisfying the
H∞ tracking performance criterion, which means that the diesel engine is provided
with improved robustness against modeling errors and external disturbances. After
transforming the MIMO diesel engine system into the canonical form, the resulting
control inputs are shown to contain nonlinear elements which depend on the system’s
parameters. The nonlinear terms which appear in the control inputs are approximated
with the use of neuro-fuzzy networks. Moreover, since only the system’s output
is measurable the complete state vector has to be reconstructed with the use of
a state observer. It is shown that a suitable learning law can be defined for the
aforementioned neuro-fuzzy approximators so as to preserve the closed-loop system
stability. With the use of Lyapunov stability analysis it is proven that the proposed
observer-based adaptive fuzzy control scheme for the diesel engine results in H∞
tracking performance.
Continuing with potential applications of differential flatness theory in embedded
control systems for the automotive industry, a nonlinear filtering and control method
is proposed for spark-ignited (SI) engines. The design of the SI engine’s control loop
is primarily based on differential flatness theory and on the use of the previously
analyzed nonlinear filtering approach, known as Derivative-free nonlinear Kalman
Filtering. It is shown that through the proposed method, efficient control of the
engine parameters (such as intake pressure and turn speed) can be succeeded. The
followed methodology solves additional problems that arise in the design of the
control loop, for example, that (i) specific variables of the engine’s state vector
are not directly measurable (e.g., the ones associated with input pressure), (ii) the
dynamic model of the SI engine is not always an accurate one, while it is subjected
to external perturbations and disturbances (such as friction torques). The proposed
control scheme is evaluated through simulation experiments.
Next, an adaptive fuzzy controller is designed for spark-ignited (SI) engines,
under the constraint that the system’s model is unknown. After transforming the
SI engine model into the canonical form, the resulting control inputs are shown to
contain nonlinear elements which depend on the system’s parameters. Once again,
the nonlinear terms which appear in the control inputs are approximated with the use
10.1 Overview 493

of neuro-fuzzy networks. It is shown that a suitable learning law can be defined for the
aforementioned neuro-fuzzy approximators so as to preserve the closed-loop system
stability. With the use of Lyapunov stability analysis it is proven that the proposed
adaptive control scheme results in H∞ tracking performance. The efficiency of the
control loop for the spark-ignited engine is checked through simulation experiments.
Finally, in this chapter differential flatness theory and the Derivative-free nonlinear
Kalman Filter are used for air–fuel ratio control in combustion engines. It is proven
that the air–fuel ratio system is a differentially flat one and admits dynamic feedback
linearization. Using a change of variables that is based on differential flatness theory,
it is shown that the air–fuel ratio system can be transformed to the linear canonical
form, for which the design of a state feedback controller is easier. Moreover, to
compensate for modeling uncertainties and external disturbances, the Derivative-
free nonlinear Kalman Filter is designed as a disturbance observer. The estimation of
the perturbations that affect the air–fuel system enables their compensation through
the inclusion of an additional term in the feedback control law. The efficiency of the
proposed nonlinear feedback controller for the air–fuel ratio system is tested through
simulation experiments.

10.2 Flatness-Based Control of Valves in Marine Diesel

Engines

10.2.1 Overview

In this section, an approach is presented about developing more efficient control

for gas exchange valves in marine internal combustion engines. The section studies
robust nonlinear control for gas exchange solenoid valves in ship diesel engines with
the use of the Derivative-free nonlinear Kalman Filter. Recent results on nonlinear
feedback control of diesel engines can be found in [82, 546], while previous results
on modeling and control of variable valve timing systems can be found in [103, 149,
217, 281, 353, 556]. Robust control of gas exchange solenoid valves is important
for improving the efficiency in the operation of diesel engines. Due to the relatively
wide range of typical internal combustion engine operating conditions imposed on a
variable valve timing actuator, the controller must compensate both for slowly vari-
ations of model parameters and for rapid external disturbances. Changes occurring
over many valve opening and closing cycles are classified as slow changes. Exam-
ples of slow changes are variations in coil resistance and oil viscosity, as a result of
temperature or mechanical wear. On the other hand, rapidly varying disturbances,
such as those due to combustion gas forces (particularly on the exhaust gas valve),
can vary within a range of several hundred Nts from one cycle to the next and are
classified as rapid changes. It is noted that the valve control problem is complicated
because disturbance forces exerted on the valve depend on certain in-cylinder para-
meters, such as pressure which cannot be directly measured and have to be estimated
through an identification (filtering) procedure.
494 10 Differential Flatness Theory for Internal Combustion Engines

The dynamic model of the valve associates the variation of the valve’s position
and coil current to a control input voltage [103]. In this section, it is shown how a
nonlinear controller for the valve’s model can be obtained through the application
of differential flatness theory [152, 465, 516, 535]. The flat output for the valve’s
model is taken to be the valve’s position. By expressing all state variables and the
control input of the valve’s model as functions of the flat output and its derivatives the
system’s dynamic model is transformed into the linear Brunovksy (canonical) form
[250, 344]. For the latter model it is possible to design a state feedback controller
that enables accurate tracking of the valve’s reference setpoints. It is also shown that
transformation of the nonlinear model of the valve to the linear canonical form can be
also succeeded with the use of differential geometric methods and the computation
of Lie derivatives.
By exploiting the valve’s exactly linearized model and its transformation into a
canonical form it is possible to design a linear state estimator for approximating the
system’s state vector through the processing of measurements coming from a small
number of sensors. To this end, the concept of Derivative-free nonlinear Kalman Fil-
tering is employed. Unlike the Extended Kalman Filter, the proposed filtering method
provides estimates of the state vector of the nonlinear system without the need for
derivatives and Jacobians calculation [422, 427]. Throughout the previous chapters
it has been shown that, by avoiding linearization approximations, the proposed fil-
tering method improves the accuracy of estimation of the system’s state variables.
Moreover, it is shown that it is possible to redesign the Kalman Filter in the form of a
disturbance observer and using the estimation of the disturbance to develop an auxil-
iary control input that compensates for perturbations’ effects. In this way the valve’s
control system can become robust to model parametric uncertainty or to external dis-
turbances. It is also noted that in terms of computation speed the proposed Kalman
Filter-based disturbance estimator is faster than perturbation estimators that may be
based on other nonlinear filtering approaches (e.g., Extended Kalman Filter). Thus
an advantage is gained for the real-time estimation of the unknown valve’s dynam-
ics in operating conditions that require fast transient responses. The efficiency of the
proposed nonlinear control and Kalman Filter-based disturbances estimation scheme
is evaluated through numerical simulation tests. It will be shown that by accurately
estimating disturbance terms the control loop succeeds elimination of the tracking
error for all state variables of the valve and results in improved functioning of the
internal combustion engine.

10.2.2 Dynamic Model of the Valve

10.2.2.1 Electromagnetic Force on Valve’s Armature

The motion equation of the solenoid valve (see Figs. 10.1 and 10.2) is described by
a forced oscillator model, that is,
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 495

Fig. 10.1 Cylinder valve

diagram in the ship
diesel engine

Fig. 10.2 Diagram of the

valve actuation
mechanism

lm
ẍ = − m1 (b ẋ + kx − Fv + Fg + lv Fm, j (x, i)) (10.1)

where m is the mass of the valve, b is the damping coefficient, K is an elasticity

coefficient, Fv is the valve spring preload, Fg is the force generated by the cylinder’s
gas, and Fm, j = {−Fm,o , Fm,c } is the electromagnetic force at the valve’s armature
(at opening or closing). Moreover, lv is the radial distance from the armature pivot
496 10 Differential Flatness Theory for Internal Combustion Engines

point to where the longitudinal armature and valve axes intersect and lm is the radial
distance from the armature pivot point to where the resulting magnetic force acts on
the armature.
Denoting as i the coil’s current, it holds that the aggregate energy that is stored in
the coil is [103]
i
Wc (x, i) = 0 λ(x, ξ )dξ
(10.2)

where λ(x, i) is the magnetic flux that is computed as

λ(x, i) = ψ(1 − e−ig(x) ) (10.3)

with
β
g(x) = κ−x −α (10.4)

where α, β, κ, ψ are constants. By substituting Eq. (10.3) into Eq. (10.2) one obtains
i
Wc (x, i) = 0 ψ(1 − e−ξ g(x) )dξ ⇒
−ig(x) (10.5)
Wc (x, i) = ψ(i + e g(x)−1 )

The magnetic force is finally given by the relation

∂ Wc (x,i)
Fm (x, i) = ∂x ⇒
 (10.6)
ψg (x)
Fm (x, i) = g 2 (x)
(1 − (1 + ig(x))e−g(x)i )

10.2.2.2 Force Due to Gas in the Cylinders

Next, to model the gas force in the cylinder one has

Fg = C g f P(x, Po , t)Av (10.7)

where Av is the face area of the valve and C f g is the gas force coefficient which
approximates flow losses behind the valve during opening. Typical values for C g are
C g ∈[0.7, 0.85]. Additionally, the model of the gas force can be written as

Fg = C g f P0 Av f 1 (x) = γ f 1 (x) (10.8)

where f 1 (x) = c1 +c2 x L +c3 x L2 , with x L = x +4·10−3 (m) to denote the valve’s lift,
whereas coefficients ci , ı = 1, 2, 3 are found through a least-squares identification
procedure. The coefficient γ = C g f P0 Av is considered to remain invariant during
the opening cycle.
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 497

About the derivative of Fgs one has

Ḟg = C g f P0 Av γ1 ẋ1 (c2 + 2c3 x L ) (10.9)

10.2.2.3 State-Space Model of the Valve

From Kirchoff’s law in the valve’s coil one has

v = iR + dt ⇒u
dλ
= dλ
dt (10.10)

where u = v − iR and λ(x, i) = ψ(1 − e−ig(x) ). By differentiating the relation about

the magnetic flux with respect to time one has

dλ
dt = −ψ(−g(x))e−ig(x) dt
di
− ψ(−ig (x))e−ig(x) (10.11)

di
Solving Eq. (10.11) with respect to dt and using Eq. (10.10) one finally obtains

eig(x) (x)
di
dt = ψg(x) u − i gg(x) ẋ (10.12)

Therefore the basic relations describing the dynamics of the valve are

lm
ẍ = − m1 (b ẋ + K x − Fv (x) + Fg (x) − Lv Fm (x, i))
eig(x)

g (x)
(10.13)
di
dt = ψg(x) u −i g(x) ẋ

The following state variables are defined: x1 = x, x2 = ẋ and x3 = i and the state
vector is x = [x1 , x2 , x3 ]T . Therefore, it holds

ẋ1 = x2
lm
ẋ2 = − m1 [bx2 + K x1 − Fv (x) + Fg (x) − lg Fm (x)] (10.14)

(x) e x3 g(x)
ẋ3 = −x3 gg(x) x2 + ψg(x) u

which also takes the matrix form

ẋ = f (x) + g(x)u (10.15)

or equivalently

⎛ ⎞ ⎛ x2
⎞ ⎛⎞
ẋ1 0
⎜ − [bx + − (x) + Fg (x) − lm ⎟ ⎜ ⎟
l g Fm (x)]⎠ + ⎝ 0 ⎠ u
1
⎝ẋ2 ⎠ = ⎝ m 2 K x 1 Fv
 x g(x) e 3
ẋ3 −x3 g (x) x2 g(x) ψg(x)
(10.16)
498 10 Differential Flatness Theory for Internal Combustion Engines

10.2.2.4 Variable Valve Timing Actuation Technology

In constant valve timing actuation systems, the timing of opening and closing of
intake and exhaust valves is predetermined and does not change during the engine’s
operation (Fig. 10.3). It is a critical parameter for the stable and reliable operation of
the combustion engine. On the other hand, in variable valve timing actuation systems
the timing of opening and closing of intake and exhaust valves can be varied according
to operating conditions (such as temperature, engine’s speed, and external load).
By opening intake valves earlier (during the exhaust stroke) or by closing exhaust
valves later (during the intake stroke), more burned gas is retained in the cylinder
for combustion, thus resulting in improved combustion efficiency and in reduction
of exhaust gases (e.g., NOx emissions). Less fuel consumption is also succeeded if
the intake valve is closed earlier (during compression stroke). Late closing has as
a result the escape of a large part of the gas initially confined in the cylinder, and
consequently this reduces the volume of air that is left for combustion. On the other
hand, there are constraints in variable valve timing control.

10.2.3 Input–Output Linearization Using Lie Algebra

It will be shown that transformation of the valve’s dynamics into the linear canonical
form can be also succeeded using a differential geometry approach. Taking that the
output of the system is y1 = h 1 (x) = x1 one defines the following variables

Fig. 10.3 Four-stroke cycle of an internal combustion diesel engine

10.2 Flatness-Based Control of Valves in Marine Diesel Engines 499

z 1 = x1 , z 2 = L f h 1 (x), z 3 = L 2f h 1 (x) (10.17)

where L f h 1 (x) = ∇h 1 (x)· f (x) is the Lie derivative.

The state-space representation of the nonlinear system dynamics was shown to be

ẋ = f (x) + g(x)u (10.18)

where x = [x1 , x2 , x3 ]T = [x, ẋ, i] and vector fields f (x) and g(x) are defined as
⎛ ⎞
x2
⎜− 1 [bx2 + K x1 − Fv (x) + Fg (x) − lm
Fm (x)]⎟
f (x) = ⎝ m 
lv ⎠ (10.19)
(x)
−x3 gg(x) x2
⎛ ⎞
0
⎜ ⎟
g(x) = ⎝ 0 ⎠ (10.20)
x g(x) e 3
ψg(x)

It holds that

z 2 = L f h 1 (x) = ∂h ∂h 1
∂ x1 f 1 + ∂ x2 f 2 +
1 ∂h 1
∂ x3 f 3 ⇒ (10.21)
z 2 = L f h 1 (x) = x2

Similarly, it holds

z 3 = L 2f h 1 (x) = ∂∂zx21 f 1 + ∂∂zx22 f 2 + ∂∂zx23 f 3 ⇒

(10.22)
z 3 = f 2 ⇒z 2 = − m1 [bx2 + K x1 − Fv (x) + Fg (x) − llmv Fm (x)]

Additionally, it holds
∂z 3 ∂z 3 ∂z 3 ∂ f3 ∂ f2 ∂ f2
ż 3 = ∂ x1 ẋ 1 + ∂ x2 ẋ 2 + ∂ x3 ẋ 3 ⇒ż 3 = ∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 ⇒ (10.23)

and since L 2f h 1 (x) = f 2 the previous relation can be written as

ż 3 = L 3f h 1 (x) + L g L 2f h 1 (x)u (10.24)

The term L 3f h 1 (x) is equal to the previously computed function f a (x) given in
Eq. (10.47) and the term L g L 2f h 2 (x) is equal to the previously computed function
ga (x). Thus one obtains
500 10 Differential Flatness Theory for Internal Combustion Engines

lm
L 3f h 1 (x) = − mb {− m1 [bx2 + K x1 − Fv (x) + Fg (x) − lv Fm (x)]} − m x2 −
K

− m1 [C f g P0 Av γ1 x2 (c2 + 2c3 x1 + 2c3 q1 )]+

   
ψg (x1 )g 2 (x1 )−ψ g (x1 )2g(x1 )g (x1 )
lm
+ ml v
{ g 4 (x1 )
[1 − (1 + x3 g(x1 ))e−g(x1 )x3 ]x2 +

 
+ ψg (x1 )
g 2 (x )
[−x3 g (x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g (x1 )x3 )e−g(x1 )x3 ]}x2 +
1
 
+{ ψg (x1 )
g 2 (x1 )
(x1 )
[−g(x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g(x1 ))e−g(x1 )x3 ]}[−x3 gg(x 1)
x2 ]
(10.25)
and

lm ψg (x1 )
L g L 2f h 1 (x) = mlv { g 2 (x1 ) ·
x3 g(x1 )
(10.26)
·[−g(x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g(x1 ))e−g(x1 )x3 ] eψg(x1 ) }

Using the aforementioned variables it can be confirmed that it holds

ż 1 = z 2 ż 2 = 3 ż 3 = L 3f h 1 (x) + L g L 2f h 1 (x)u (10.27)

By defining the new control input v = L 3f h 1 (x) + L g L 2f h 1 (x)u one finally obtains
that
(3)
z1 = v (10.28)

which is written in the state-space form

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (10.29)
ż 3 000 z3 1

with measurement equation

⎛ ⎞

z1
z 1meas = 1 0 0 ⎝z 2 ⎠ (10.30)
z3

It is noted that since L g L 2f h 1 (x) = L g L n−1

f h 1 (x) =0 the relative degree of the
system is n = 3. For the linearized equivalent of the system one can proceed to the
design of a state feedback controller, following the stages described in the end of
Sect. 10.2.4.
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 501

10.2.4 Input–Output Linearization Using Differential

Flatness Theory

10.2.4.1 Differential Flatness of the Valve’s Model

Differential flatness theory will be used to obtain a linearized description of the

valve’s dynamics. First it is shown that the system is a differentially flat one, which
means that all its state variables and the control input can be written as functions of
the flat output and its derivatives. The flat output is taken to be y = x1 . It holds

x1 = y, x2 = ẋ1 = ẏ (10.31)

From the second row of the state-space equations one has

Fv (x) Fg (x) lm
ẋ2 = − mb x2 − K
m x1 − m − m + mlv Fm (x) (10.32)

where function Fg has been defined as

Fg (x) = C g f P0 Av [c1 + c2 (x1 + q1 ) + c2 (x1 + q1 )2 ] (10.33)

and function Fm has been defined as


ψ g (x1 )
Fm (x) = g 2 (x1 )
[1 − (1 + ig(x1 ))e−g(x1 )x3 ] (10.34)

Using Eqs. (10.33) and (10.34) into Eq. (10.32) one gets

K x − Fv − C P A [c + c (x + q ) + c (x + q )2 ]+
ẋ2 = − mb x2 − m 1 m gf 0 v 1 2 1 1 3 1 1
 (10.35)
lm ψg (x1 )
+ ml lm
− ml (1 + x3 g(x1 ))e−g(x1 )x3
v g 2 (x 1 ) v

To get a closed form solution for x3 , the Taylor series expansion of the term e−g(x1 )x3
is used. Thus one has

g 2 (x1 ) 2
e−g(x1 )x3 = 1 − g(x1 )x3 + 2 x3 + ··· (10.36)

Keeping the first two terms of the expansion and substituting x1 = y and x2 = ẏ
one gets

Fv
ÿ = − mb ẏ − K
m y− m − C g f P0 Av [c1 + c2 (y + q1 ) + c3 (y + q1 )2 ]+
 (10.37)
lm ψg (y) lm
+ mlv g2 (y) − ml v
(1 + x3 g(y))(1 − x3 g(y))
502 10 Differential Flatness Theory for Internal Combustion Engines

After intermediate operations and solving with respect to x3 one obtains

mlv Fv
x3 = lm {| ÿ + b
m ẏ + K
m y+
m+

lm ψg (y) lm 1/2
+C g f P0 Av [c1 + c2 (y + q1 ) + c3 (y + q1 )2 ] − ml v g (y)
2 + mlv |} (10.38)

⇒x3 = f 1 (y, ẏ, ÿ)

Moreover, from the third row of the state-space equations one has

(x) e x3 g(x)
ẋ3 = −x3 gg(x) x2 + ψg(x) u⇒
 f 1 (y, ẏ, ÿ)g(y)
(10.39)
g (y)
f˙1 (y, ẏ, ÿ) = − f 1 (y, ẏ, ÿ) g(y) ẏ + e ψg(x) u⇒

Thus, solving with respect to the control input u one gets


(y)
[ f˙1 (y, ẏ, ÿ)+ f 1 (y, ẏ, ÿ) gg(y) ẏ]ψg(y)
u= e f 1 (y, ẏ, ÿ)g(y)
⇒
(10.40)
u = f 2 (y, ẏ, ÿ)

According to the above, all state variables and the control input of the valve’s model
can be written as functions of the flat output and its derivatives, therefore the system
is a differentially flat one.

10.2.4.2 Transformation into the Linear Canonical Form

Next, it will be shown that using differential flatness theory the valve’s model can
be transformed into the linear canonical form. By differentiating the second row of
the state equations one has
lm
ẍ2 = − mb ẋ2 − K
m ẋ 1 − 1
m Ḟv (x) − 1
m Ḟg (x) + mlv Ḟm (v) (10.41)

where it holds Ḟv = 0 and defining x L = x + 4 × 10−3 = x + q1 one has

Ḟg = C f g P0 Av γ1 x2 (c2 + 2c3 x + 2c3 q1 ) (10.42)

It also holds that


ψ g (x)
Fm (x, i) = g 2 (x)
[1 − (1 + ig(x))]e−g(x)i (10.43)

and by differentiating with respect to time one gets

Ḟm = dF
d x1 ẋ 1 + dF
d x2 ẋ 2 (10.44)
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 503

which after intermediate operations gives

  
Ḟm = { ψg (x1 )g 2 (x1 )−ψ g (x1 )2g(x1 )g (x1 )
g 4 (x1 )
[1 − (1 + x3 g(x1 ))e−g(x1 )x3 ]+

 
+ ψg (x1 )
g 2 (x )
[−x3 g (x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g (x1 )x3 )e−g(x1 )x3 ]}x2 +
1
(10.45)

+{ ψg (x1 )
g 2 (x1 )
[−g(x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g(x1 ))e−g(x1 )x3 ]}·
 x3 g(x1 )
(x1 )
·[−x3 gg(x 1)
x2 + eψg(x1 ) u]

By substituting Eqs. (10.42) and (10.45) into Eq. (10.41) one gets
(3)
x1 = f a (x) + ga (x)u (10.46)

where

f a (x) = − mb {− m1 [bx2 + K x1 − Fv (x) + Fg (x) − llmv Fm (x)]} − m x2 −

K

− m1 [C f g P0 Av γ1 x2 (c2 + 2c3 x1 + 2c3 q1 )]+

   
ψg (x1 )g 2 (x1 )−ψ g (x1 )2g(x1 )g (x1 )
lm
+ ml v
{ g 4 (x1 )
[1 − (1 + x3 g(x1 ))e−g(x1 )x3 ]x2 +

 
+ ψg (x1 )
g 2 (x )
[−x3 g (x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g (x1 )x3 )e−g(x1 )x3 ]}x2 +
1
 
+{ ψg (x1 )
g 2 (x1 )
(x1 )
[−g(x1 )e−g(x1 )x3 − (1 + x3 g(x1 ))(−g(x1 ))e−g(x1 )x3 ]}[−x3 gg(x 1)
x2 ]
(10.47)
and

lm ψg (x1 ) −g(x1 )x3 −
ga (x) = mlv { g 2 (x1 ) [−g(x 1 )e
x3 g(x1 ) (10.48)
−(1 + x3 g(x1 ))(−g(x1 ))e−g(x1 )x3 ] eψg(x1 ) }

Thus, finally denoting v = f a (x) + ga (x)u one has a relation of the form
(3)
x1 = v (10.49)

By defining the new state variables z 1 = x1 , z 2 = ẋ1 and z 3 = ẍ3 one obtains a
description of the system’s dynamics in the following state-space form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (10.50)
ż 3 000 z3 1

with measurement equation

⎛ ⎞

z1
z 1m = 1 0 0 ⎝z 2 ⎠ (10.51)
z3
504 10 Differential Flatness Theory for Internal Combustion Engines

According to the above a description of the system’s dynamics in the linear canonical
form has been obtained. For this linearized model the feedback controller’s design
proceeds as follows:
(3)
v = z 1,d − k1 (z̈ 1 − z̈ 1,d ) − k2 (ż 1 − ż 1,d ) − k3 (z 1 − z 1,d ) (10.52)

Denoting the tracking error e = z 1 − z 1,d , this leads to a dynamic error equation for
the closed-loop system, of the form

e(3) + k1 ë + k2 ė + k3 e = 0 (10.53)

By selecting the feedback gains ki , i = 1, 2, 3 such that the above characteristic

polynomial is a Hurwitz one, the tracking error converges asymptotically to zero,
that is, limt→∞ e(t) = 0 or limt→∞ z 1 (t) = z 1,d (t).
The control signal that is finally given as input to the system is computed by

v = f a (x) + ga (x)u⇒u = ga−1 (x)[v − f a (x)] (10.54)

10.2.5 Disturbances Compensation with Derivative-Free

Nonlinear Kalman Filter

Next, it is assumed that the dynamic model of the valve is characterized by parametric
uncertainties or changes and external disturbances, and thus takes the form

z (3) = v + d̃ (10.55)

Without loss of generality it is considered that the dynamics of the disturbances is

described by its third-order derivative, thus one has d̃ (3) = f d (y, ẏ, ÿ). The system’s
state vector is extended to comprise disturbance terms. The new state variables are
given by

z 1 = x1 , z 2 = ẋ1 , z 3 = ẍ1
˙ z = d̃¨ (10.56)
z 4 = d̃, z 5 = d̃, 6

The state-space equations of the system now become

ż = Az + Bve (10.57)
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 505

where
⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 00
⎜0 0 1 0 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0⎟ ⎜ ⎟
A=⎜ ⎟ B = ⎜1 0 ⎟ (10.58)
⎜0 0 0 0 1 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 1⎠ ⎝0 0⎠
0 0 0 0 0 0 01

and the control input becomes ve = [v, f d ]T . The associated disturbances estimator
is described by

ẑ˙ = Ao ẑ + Bo ve + K (z meas − Co ẑ)

(10.59)
ẑ m = Co ẑ
⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 00
⎜0 0 1 0 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0⎟ ⎜ ⎟
Ao = ⎜ ⎟ Bo = ⎜1 0⎟ (10.60)
⎜0 0 0 0 1 0⎟⎟ ⎜0 0⎟
⎜ ⎜ ⎟
⎝0 0 0 0 0 1 ⎠ ⎝0 0⎠
0 0 0 0 0 0 00

while matrix Co is given by

Co = 1 0 0 0 0 0 (10.61)

where K ∈R 6×1 is the state estimator’s gain. The above disturbance observer enables
simultaneous estimation of the nonmeasurable state variables z 2 = ẋ, z 3 = ẍ, and
estimation of the disturbance term d̃ and of its derivatives.
Defining as Ad , Bd and Cd , the discrete-time equivalents of matrices Ao , Bo
and Co respectively, the associated Kalman Filter-based estimator is given by
[31, 405, 408]
measurement update:

K (k) = P − (k)CdT [Cd ·P − (k)CdT + R]−1

ẑ(k) = ẑ − (k) + K (k)[z meas (k) − Cd ẑ − (k)] (10.62)
P(k) = P − (k) − K (k)Cd P − (k)

time update:

P − (k + 1) = Ad (k)P(k)AdT (k) + Q(k)

(10.63)
ẑ − (k + 1) = Ad (k)ẑ(k) + Bd (k)v(k)
506 10 Differential Flatness Theory for Internal Combustion Engines

10.2.6 Simulation Tests

The performance of the proposed control scheme for valves in ship diesel engines,
that was based on differential flatness theory and on the use of the Derivative-free
nonlinear Kalman Filter, was assessed through simulation experiments. The tracking
results are shown in Figs. 10.4, 10.5, 10.6, 10.7, 10.8, 10.9, 10.10, 10.11, 10.12 and
10.13. It can be observed that in all test cases the proposed nonlinear control method

(a)4 10
(b) 1

disturbance
0.5
3 5
1

x2
x

0
0 0.5 1 1.5 2 2.5 3 3.5 4
2 0
time (sec)
d /dt disturbance d/dt disturbance 0.5
1 −5
0 1 2 3 4 0 1 2 3 4
0
time (sec) time (sec)
150 100 −0.5
0 0.5 1 1.5 2 2.5 3 3.5 4
time (sec)
100 50 0.2
x3

50 0 0
2

−0.2
2

0 −50 0 0.5 1 1.5 2 2.5 3 3.5 4

0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.4 Diesel engine valve control test case 1: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

(a) (b)
2 10 4
disturbance

1 5 2
x1

x2

0 0
0
0 0.5 1 1.5 2 2.5 3 3.5 4
−1 −5 time (sec)
disturbance d/dt disturbance

1
−2 −10
0 1 2 3 4 0 1 2 3 4
0
time (sec) time (sec)
25 600 −1
0 0.5 1 1.5 2 2.5 3 3.5 4
20 400 time (sec)
0.2
3

15 200
u
x

0
10 0
2
d /dt

−0.2
2

5 −200 0 0.5 1 1.5 2 2.5 3 3.5 4

0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.5 Diesel engine valve control test case 2: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 507

(a) (b)
2 4 2

disturbance
2 1
1
x1

x2
0
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0
−2 time (sec)

disturbance d/dt disturbance

1
–1 −4
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0

80 150
−1
0 0.5 1 1.5 2 2.5 3 3.5 4
100 time (sec)
60
0.2
3

50
u
x

40 0
0

2
d 2 /dt
−0.2
20 −50 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.6 Diesel engine valve control test case 3: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

(a) (b)
2 4 2
disturbance

2 1
1
x1

x2

0
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0
−2 time (sec)
disturbance d/dt disturbance

0.5
–1 −4
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0

80 150 −0.5
0 0.5 1 1.5 2 2.5 3 3.5 4
100 time (sec)
60 0.2
3

50
x

0
40
2

0
d /dt

−0.2
2

20 −50 0 0.5 1 1.5 2 2.5 3 3.5 4

0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.7 Diesel engine valve control test case 4: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

succeeds fast and accurate tracking of the reference setpoints. It can be also noticed
that good transients were succeeded.
The list of the reference setpoints and of the disturbance inputs considered as each
simulation experiment is as follows (Table 10.1):
The section’s results can be applied to all types of valves used in combustion
engines for control of gas and fuel inflow and outflow. Once it is shown that the valve’s
dynamic model satisfies differential flatness properties, it is possible to transform
it into a linearized form for which the design of a feedback controller becomes
508 10 Differential Flatness Theory for Internal Combustion Engines

(a) (b)
1.5 2 2

disturbance
1
1 0
x1

x2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
0.5 −2
time (sec)

disturbance d/dt disturbance

0.4
0 −4
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0.2

70 100 0
0 0.5 1 1.5 2 2.5 3 3.5 4
60 time (sec)
50 0.2
x3

50

0 0
40

2
d 2 /dt
−0.2
30 −50 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.8 Diesel engine valve control test case 5: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

(a) (b)
2.5 4 2
disturbance

1
2 2
x1

2
x

0
0 0.5 1 1.5 2 2.5 3 3.5 4
1.5 0
time (sec)
disturbance d/dt disturbance

0.4
1 −2
0 1 2 3 4 0 1 2 3 4
time (sec) 0.2
time (sec)

100 40 0
0 0.5 1 1.5 2 2.5 3 3.5 4
80 time (sec)
20 0.2
3

60
x

0 0
40
2
d 2 /dt

−0.2
20 −20 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.9 Diesel engine valve control test case 6: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

easy (e.g., with the use of pole placement methods). The robustness features of
the proposed control method are enhanced by the Derivative-free nonlinear Kalman
Filter. The latter enables to estimate disturbance terms that affect the valve’s dynamic
model and subsequently to compensate for them.
Comparing to other control methods, a generic conclusion is that it is not rec-
ommended to use PID control for precision and reliability demanding function of
combustion engine valves. This is because the tuning of PID controller is mostly
based on heuristics. Moreover, the stability of the control loop can be studied in
10.2 Flatness-Based Control of Valves in Marine Diesel Engines 509

(a) (b)
8 10 2

disturbance
6 1
0
x1

2
4

x
0
−10 0 0.5 1 1.5 2 2.5 3 3.5 4
2 time (sec)

disturbance d/dt disturbance

1
0 −20
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0

200 600 −1
0 0.5 1 1.5 2 2.5 3 3.5 4
150 400 time (sec)
0.2
x3

100 200
0
50 0

2
d 2 /dt
−0.2
0 −200 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.10 Diesel engine valve control test case 7: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

(a) 8 20
(b) 2
disturbance

6 10 1
x1

x2

4 0
0
0 0.5 1 1.5 2 2.5 3 3.5 4
2 −10 time (sec)
d/dt disturbance

0.5
0 −20
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0

300 400 −0.5

0 0.5 1 1.5 2 2.5 3 3.5 4
time (sec)
disturbance

200 200 0.2

3

u
x

100 0 0
2

−0.2
d 2 /dt

0 −200 0 0.5 1 1.5 2 2.5 3 3.5 4

0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.11 Diesel engine valve control test case 8: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

limited cases associated with linear system dynamics, while the robustness of the
control loop is questionable in case of external perturbations. PID controllers do not
assure satisfactory performance of the control loop under variable operating condi-
tions, while the controller has to be retuned at each change of operating points.
Unlike PID control, the presented nonlinear control method for combustion engine
valves is of proven stability and robustness. This is confirmed in the linearized equiv-
alent model of the valve after application of the pole placement technique has poles
exclusively in the left complex semi-plane. Besides the inclusion of the additional
510 10 Differential Flatness Theory for Internal Combustion Engines

(a) 2 20
(b) 2

disturbance
1 10 1
1

x2
0 0
x

0
0 0.5 1 1.5 2 2.5 3 3.5 4
−1 −10 time (sec)

disturbance d/dt disturbance

1
−2 −20
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec) 0

25 600 −1
0 0.5 1 1.5 2 2.5 3 3.5 4
20 400 time (sec)
0.2
3

15 200
x

0
10 0

2
d 2 /dt
−0.2
5 −200 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.12 Diesel engine valve control test case 9: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

(a) 3 5
(b) 2
disturbance

2.5 1
1

2 0
x

0
0 0.5 1 1.5 2 2.5 3 3.5 4
1.5 time (sec)
d/dt disturbance

0.5
1 −5
0 1 2 3 4 0 1 2 3 4
time (sec) time (sec)

150 200 0
0 0.5 1 1.5 2 2.5 3 3.5 4
time (sec)
disturbance

100 100 0.2

x3

50 0 0
2
d 2 /dt

−0.2
0 −100 0 0.5 1 1.5 2 2.5 3 3.5 4
0 1 2 3 4 0 1 2 3 4 time (sec)
time (sec) time (sec)

Fig. 10.13 Diesel engine valve control test case 10: a convergence of the valve’s state variables to
setpoints and associated control input, b estimation of valve’s disturbance input and of its derivatives

control input that compensates for the estimated additive disturbance terms improves
the robustness features of this control loop. It is also noted that the linearized equiv-
alent model of the valve, e.g., Eq. (10.29), exhibits multiple poles at the origin. Thus
stabilization can be succeeded by pole placement methods. Finally, it is noted that
the stability and robustness features of the control scheme which comprises also esti-
mation and compensation of the disturbances are similar to those of LQG control.
The above provide sufficient evidence about the stability and disturbance rejection
capability of the proposed feedback control scheme.
10.3 Flatness-Based Control of Diesel Combustion Engines 511

Table 10.1 Setpoints and disturbance inputs

No test Setpoint Disturbance
No 1 xd (t) = 3.3 d(t) = 0.5

No 2 xd (t) = sin 6πt

T d(t) = 1.8 + 0.5sin( 2π
Td1 ) + 0.2cos( Td2 )
t 2π t

T =2 Td1 = 10, Td2 = 20

No 3 xd (t) = sin( 2.5
2π
t) d(t) = 1.5 + 0.2cos( 2π Td )
t

T = 2.5 T d = 20
No 4 xd (t) = sin( 2π T )
t
d(t) = 1 + 0.2sin( 2π
Td )
t

T =2 Td = 10
No 5 xd (t) = 0.15(0.5 + 2 πT )t d(t) = 1 + 0.2sin( 2π
Td1 ) + 0.2sin( Td2 )
t 2π t

T = 2 and t ≤ tsim 2 Td1 = 40, Td2 = 50

xd (t) = 0.15(0.5 − π tsim + 2 πT t)
tsim
T = 2 and t > 2
No 6 xd (t) = 2 + 0.3 · (t − n · tsim 4 ) d(t) = 1.1 + 0.5sin( 2π
Td1 ) + 0.6sin ( Td2 )
t 2 2π t

n · tsim
4 ≤ t < (n + 1) · tsim
4 , n = 0, 1, 2, . . . Td1 = 40, Td2 = 50
No 7 xd (t) = 3 t ≤ tsim 3 d(t) = 1.4 + 0.2cos( 2π
Td )
t
tsim
xd (t) = 6 3 ≤ t ≤ 2 3 tsim Td = 20
xd (t) = 4 2t3sim ≤ t ≤ tsim
No 8 xd (t) = 5 t ≤ tsim 4 d(t) = 1 + 0.3cos( 2π
Td )
t
tsim
xd (t) = 7.1 4 ≤ t ≤ 2 4 tsim Td = 40
xd (t) = 4 2t4sim ≤ t ≤ 3t4sim
xd (t) = 6 3t4sim ≤ t ≤ tsim
No 9 xd (t) = sin(3π t), t ≤ tsim 2 d(t) = 1.3 + 0.2sin( 2π
Td1 ) + 0.2cos( Td2 )
t 2π t

xd (t) = 0.9(1 + cos(3π t)) t > tsim 2 Td1 = 40, Td2 = 60

No 10 xd (t) = 1.7 + sin2 ( 23 π t) d(t) = 1.0 + 0.2t + 0.3sin( 2π
Td )
Td = 40

10.3 Flatness-Based Control of Diesel Combustion Engines

10.3.1 Overview

In this section, the use of differential flatness theory-based methods for control and
filtering of nonlinear diesel engine models is examined. The development of embed-
ded control for transportation systems comes against problems of increased difficulty
arising from the highly nonlinear dynamics of the considered systems and from the
technical difficulties in measuring specific elements of the systems’ state vector [27,
181, 222, 293, 339, 356, 481]. Previous approaches to the control of turbocharged
diesel engines and exhaust gas recirculation systems comprise PID and Lyapunov
methods. One can also find results on neural and fuzzy control both for Compression
Ignition (CI) and Spark Ignition (SI) engines [10, 57, 83, 106, 339, 361, 374, 375]. In
particular, control of turbocharged diesel engines is a complicated problem because
the nonlinear model of the engine cannot be subjected to static feedback linearization
512 10 Differential Flatness Theory for Internal Combustion Engines

[113, 251, 375, 551]. Therefore, one has to apply dynamic feedback linearization by
extending the state vector of the system, so as to include as additional state variables
the initial control inputs and their derivatives. Following, this approach, which is also
known as dynamic extension, it is possible to show that the model of the turbocharged
diesel engine is a differentially flat one. Moreover, by expressing all state variables
and the control inputs of the model one can arrive at an equivalent description of the
system in the linear canonical (Brunovsky) form.
In this section, the nonlinear control problem of the turbocharged diesel engines is
solved with the use of differential flatness theory and of the Derivative-free nonlinear
Kalman Filter. It is shown that the dynamic model of the turbocharged diesel engine
is differentially flat and admits dynamic feedback linearization. It is also shown that
the dynamic model can be written in the linear Brunovsky canonical form for which a
state feedback controller can be easily designed. To compensate for modeling errors
and external disturbances the Derivative-free nonlinear Kalman Filter is used and
redesigned as a disturbance observer. As already explained, the filter consists of
the Kalman Filter recursion on the linearized equivalent of the diesel engine model
and on an inverse transformation based on differential flatness theory which enables
to obtain estimates for the state variables of the initial nonlinear model [435, 436,
437, 441, 447, 448]. Once the disturbances variables are identified it is possible to
compensate for them by including an additional control term in the feedback loop.
The efficiency of the proposed control method for the diesel engine is tested through
simulation experiments.

10.3.2 Dynamic Model of the Turbocharged Diesel Engine

The basic parameters of the diesel engine (Fig. 10.14) are: (i) Gas pressure in the
intake manifold p1 , (ii) Gas pressure in the exhaust manifold p2 , (iii) Turbine power
Pt , (iv) Compressor power Pc . Additional variables of importance are Wc which is
the compressor mass flow rate, T1 the intake manifold temperature, T2 is the exhaust
manifold temperature, Wt is the turbine mass flow rate, and WEGR is the exhaust gas
recirculation flow rate.
The basic relations of the diesel engine’s dynamics are:

Ṫ1
ṗ1 = K 1 (Wc + u 1 − K e p1 ) + T1 p1
Ṫ2
ṗ2 = K 2 (K e p1 − u 1 − u 2 ) + T2 p2
(10.64)
Ṗc = τ1 (ηm Pt − Pc )

The control inputs to this model are the exhaust gas recirculation (EGR) flow rate
u 1 = WEGR and the compressor mass flow rate u 2 = Wt . Moreover, it holds that

Wc = Pc pμK−1
c
(10.65)
1
10.3 Flatness-Based Control of Diesel Combustion Engines 513

Fig. 10.14 Diagram of the turbocharged diesel engine

−μ
Pt = K t (1 − p2 )u 2 (10.66)

The model is simplified by setting Ṫ1 = 0 and Ṫ2 = 0. In such a case the associated
state-space equations are given by

ṗ1 = K 1 (Wc + u 1 − K e p1 )
ṗ2 = K 2 (K e p1 − u 1 − u 2 ) (10.67)
Ṗc = τ (ηm Pt
1
− Pc )

Moreover, it holds that

Wc = Pc pμK−1
c
(10.68)
1

Pt = K t (1 − p2 μ )u 2 (10.69)

The description of the diesel engine in state-space form is given by

ẋ = f (x) + ga (x)u 1 + gb (x)u 2 (10.70)

514 10 Differential Flatness Theory for Internal Combustion Engines

where
⎛ ⎞
K 1 K c pPμc − K 1 K e p1
⎜ 1 ⎟
f (x) = ⎝ K 2 K e p1 ⎠
− Pc
⎛ τ ⎞ ⎛ ⎞ (10.71)
K1 0
ga (x) = ⎝−K 2 ⎠ gb (x) = ⎝ −K 2 ⎠
0 −μ
K o (1 − p2 )

With respect to the control, the variables of the output are: (i) the input manifold
pressure p1 and (ii) the compressor mass flow rate Wc ,
  
p1 p1
y= = Pc pμK−1
c (10.72)
Wc
1

10.3.3 Nonlinear Control of the Diesel Engine Using Lie

Algebra

10.3.3.1 A Low-Order Feedback Control Scheme

The previous definition of the system’s outputs given in Eq. (10.72) is used. The
linearization of the diesel engine dynamics is based on the following relations:

z 11 = h 1 (x)
z 21 = L f h 1 (x)
ż 11 = L f h 1 (x) + L ga h 1 (x)u 1 + L gb h 1 (x)u 2
(10.73)
z 12 = h 2 (x)
z 22 = L f h 2 (x)
ż 12 = L f h 2 (x) + L ga h 2 (x)u 1 + L gb h 2 (x)u 2

It holds that

z 11 (x) = p1
∂h 1 ∂h 1
L f h 1 (x) = + ∂h
∂ x1 f 1 + ∂ x2 f 2∂ x3 f 3 ⇒
1
(10.74)
L f h 1 (x) = 1 f 1 + 0 f 2 + 0 f 3 ⇒L f h 1 (x) = K 1 K c pμP−1
c
− K 1 K c p1
1

Moreover, it holds that

L ga h 1 (x) = ∂h ∂h 1 ∂h 1
∂ x1 ga1 + ∂ x2 ga2 + ∂ x3 ga3 ⇒
1
(10.75)
L ga h 1 (x) = 1ga1 + 0ga2 + 0ga3 ⇒L g h 1 (x) = K 1
10.3 Flatness-Based Control of Diesel Combustion Engines 515

while it also holds that

L gb h 1 (x) = ∂h ∂h 1 ∂h 1
∂ x1 gb1 + ∂ x2 gb2 + ∂ x3 gb3 ⇒
1
(10.76)
L gb h 1 (x) = 1gb1 + 0gb2 + 0gb3 ⇒L g h 1 (x) = 0

Similarly it holds

∂h 2 ∂h 2 ∂h 2
L f h 2 (x) = ∂ x1 f 1 +∂ x2 f 2 + ∂ x3 f 3 ⇒
μ−1   (10.77)
−K c μ p1 Pc
L f h 2 (x) = pc μ 2 K 1 K c pμ −1−K + π μK−1
c
(− Pτc )
π1 −1 1 1 K e p1 1

Moreover, it holds
∂h 2 ∂h 2 ∂h 2
L ga h 2 (x) = ∂ x1 ga1 + ∂ x2 ga2 + ∂ x3 ga3 ⇒
μ−1
−K c μ p1 Kc
L ga h 2 (x) = pc μ 2 ga1 + 0ga2 + μ
p −1 a3
g = 0⇒ (10.78)
π1 −1 1
μ−1
−K c μ p1
L ga h 2 (x) = Pc μ 2 K1
p1 −1

Equivalently, one obtains

∂h 2 ∂h 2 ∂h 2
L gb h 2 (x) = ∂ x1 gb1 + ∂ x2 gb2 + ∂ x3 gb3 ⇒
μ−1
−K μ p
L gb h 2 (x) = Pc cμ 12 gb1 + 0gb2 + pμK−1 c
gb3 ⇒ (10.79)
p1 −1 1
Kc −μ
L gb h 2 (x) = pμ −1 K o (1 − p2 )
1

From the relations

ż 11 = L f h 1 (x) + L ga h 1 (x)u 1 + L gb h 2 (x)u 2

(10.80)
ż 12 = L f h 2 (x) + L ga h 2 (x)u 1 + L gb h 2 (x)u 2

which is also written as

⎛ ⎞
 1 K 1 K c pμP−1
c
− K 1 K c P1
ż 1
= ⎝ −K c μ pμ−1   ⎠+
1
ż 12 pc μ 12 K 1 K c pμ −1−K Pc
+ π μK−1
c
(− Pτc )
⎛ π1 −1 1 K p
1 e 1
⎞ 1 (10.81)
K1 0  
⎝ −K c μ p1μ−1 ⎠ u1
−μ
Pc μ 2 K 1 pμK−1 c
K o (1 − p2 ) u2
p1 −1 1

Therefore, one arrives at a relation of the form

z̃˙ 1 = f˜a + M̃a ũ (10.82)

516 10 Differential Flatness Theory for Internal Combustion Engines

The new control inputs can be defined as

v1 = L f h 1 (x) + L ga h 1 (x)u 1 + L gb h 2 (x)u 2

(10.83)
v2 = L f h 2 (x) + L ga h 2 (x)u 1 + L gb h 2 (x)u 2

one has the dynamics

ż 11 = v1
(10.84)
ż 12 = v2

In such a case the feedback control law results into

v1 = ż 1,d
1 − K 1 (z 1 − z 1 )
p 1 1,d
(10.85)
v2 = ż 1,d
2 − K 2 (z 2 − z 2 )
p 1 1,d

results in asymptotic elimination of the tracking error. Therefore, in that case the
tracking error dynamics becomes

ė1 + K 1p e1 = 0⇒limt→∞ e1 (t) = 0, K 1p > 0

(10.86)
ė2 + K 2p e2 = 0⇒limt→∞ e2 (t) = 0, K 2p > 0

The previous results are confirmed through the computation of time derivatives and
the use of differential flatness theory:
pc
y1 = p1 ⇒ ẏ1 = ṗ1 ⇒ ẏ1 = k1 kc pμ −1−k + k1 u 1 (10.87)
1 1 k e p1

Similarly, it holds

∂ y2 ∂ y2 ∂ y2
y2 = pc pμk−1
c
⇒ ẏ2 = ∂ x1 ẋ 1 + ∂ x2 ẋ 2 + ∂ x3 ẋ 3 ⇒
1

μ−1
−K c μ p1 Kc
ẏ2 = pc μ 2 ẋ1 + 0 ẋ2 + μ ẋ ⇒
p1 −1 3
( p1 )−1

μ−1
−K c μ p1
ẏ2 = pc μ 2 [(K 1 K c pμP−1
c
− K 1 K e p1 ) + K 1 u 1 ]+
( p1 )−1 1 (10.88)
Kc −μ
+ μ
p1 −1
[− Pτc + K o (1 − p2 )u 2 ]⇒

μ−1
−K c μ p1
ẏ2 = pc μ 2 (K 1 K c pμP−1
c
− K 1 K e p1 ) + Kc
μ
p1 −1
(− Pτc )+
( p1 )−1 1
μ−1
−K c μ p1 Kc −μ
+ pc μ 2 K1u1 + μ K (1 −
p1 −1 o
p2 )u 2
( p1 )−1
10.3 Flatness-Based Control of Diesel Combustion Engines 517

Therefore, one arrives again at a dynamics of the following form:

⎛ ⎞
  K 1 K c pμP−1c
− K 1 K c p1
ẏ1
= ⎝ −K c μ pμ−1 ⎠+
1
ẏ2 pc μ 1 2 (K 1 K c pμP−1 c
− K 1 K c p1 ) + Kμ−1
c
(− Pτc )
⎛ ( p1 −1) 1 ⎞p1 (10.89)
K1 0  
⎝ μ−1   ⎠ u1
+ −K μ p −μ
pc cμ 1 2 K 1 pμK−1 c
K o (1 − p2 ) u2
( p1 −1) 1

which finally gives

ỹ˙ = f˜a + M̃a u (10.90)

10.3.4 A Dynamic Extension-Based Feedback Control Scheme

In the following, dynamic feedback linearization is performed, which means that

the control inputs that appear in the linearized form of the system are not only
functions of the initial control variables u 1 , u 2 but they also contain terms which are
based on the derivatives u̇ 1 , u̇ 2 . Thus, one has v1 = f 1 (u 1 , u̇ 1 ) and v2 = f 2 (u 2 , u̇ 2 ).
Equivalently, this means that the control inputs u 1 , u 2 which are finally applied to the
real system depend on v1 , v2 through an integration relation, that is, v1 = f 1 (u 1 , u̇ 1 )
and v2 = f 2 (u 2 , u̇ 2 ).
The dynamical system of the diesel engine is written in an extended form using
the variables v1 = u̇ 1 = ż, v2 = u 2 which means u 1 = v1 dt, u 2 = v2 . Thus, using
Eqs. (10.70) and (10.71) and by substituting u 1 = z 1 and ż = v1 as intermediate
state variable it holds

ṗ1 = K 1 K c pμP−1
c
− K 1 K e p1 + K 1 z
1
ṗ2 = K 2 K e p1 − K 2 z − K 2 v2 (10.91)
Ṗc = − Pτc + K o (1 − p2 −μ )
ż = v1

therefore, by defining the state vector x = [x1 , x2 , x3 , x4 ]T = [ p1 , p2 , Pc , z]T the

state-space description of the diesel engine model becomes

ẋ1 = K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4
1
ẋ2 = K 2 K e x1 − K 2 x4 − K 2 v2 (10.92)
ẋ3 = − xτ3 + K o (1 − x2 −μ )
ẋ4 = v1
518 10 Differential Flatness Theory for Internal Combustion Engines

Consequently, in matrix form one has

⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ1 K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 0 0
⎜ẋ2 ⎟ ⎜ 1
K 2 K e x1 − K 2 x4 ⎟ ⎜0⎟ ⎜ ⎟
⎜ ⎟=⎜ ⎟ + ⎜ ⎟ v1 + ⎜1⎟ v2 (10.93)
⎝ẋ3 ⎠ ⎝ − xτ3 + K o (1 − x2 )
μ ⎠ ⎝0⎠ ⎝0 ⎠
ẋ4 0 1 0

The system’s outputs are chosen to be

y1 = x1 = p1
y2 = Pc pμK−1
c
⇒y2 = x3 x μK−1
c (10.94)
1 1

Linearization of the system is performed using the following relations-state variables

z 11 = h 1 (x)
z 2 = L f h 1 (x)
1
(10.95)
ż 21 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

and equivalently

z 12 = h 2 (x)
z 2 = L f h 2 (x)
2
(10.96)
ż 22 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

It holds that
z 11 = h 1 (x)⇒z 11 = x1 (10.97)

z 21 = L f h 1 (x)⇒z 21 = ∂h ∂h 1 ∂h 1 ∂h 1
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 + ∂ x4 f 4 ⇒
1

z 21 = 1· f 1 + 0· f 2 + 0· f 3 + 0· f 4 ⇒z 21 = f 1 ⇒ (10.98)
z 21 = K 1 K 3 x px−1
3
− K 1 K e x1 + K 1 x4
1

Moreover, it holds that

∂z 21 ∂z 21 ∂z 21
L 2f h 1 (x) = L f z 21 ⇒L 2f h 1 (x) = ∂ x1 f 1 + ∂ x1 f 2 + ∂ x3 f 3 ⇒

μ−1
−x3 μx1
L 2f h 1 (x) = (K 1 K 3 μ
x1 −1
− K 1 K e ) f 1 + 0 f 2 + ( xKμ1−1
K3
) f3 + K 1 f4 ⇒
1

μ−1
−x3 μx1
L 2f h 1 (x) = (K 1 K 3 μ
x1 −1
− K 1 K e )(K 1 K 3 x μx−1
3
− K 1 K e x1 + K 1 x4 )+
1
K1 K3 x3 −μ
+( x μ −1 )(− τ + K o (1 − x2 ))
1
(10.99)
10.3 Flatness-Based Control of Diesel Combustion Engines 519

Equivalently, one computes

∂z 21 ∂z 21
∂z 21 ∂z 21
L ga L f h 1 (x) = L ga z 21 ⇒L ga L f h 1 (x) = ∂ x1 ga1 + ∂ x2 ga2
∂ x3 ga3 + + ∂ x4 ga4 ⇒
μ−1
−x3 μx1
L ga L f h 1 (x) = (K 1 K 3 μ
x1 −1
− K 1 K e )ga1 + 0ga2 + ( xKμ1−1
K3
)ga3 + K 1 ga4 ⇒
1
L ga L f h 1 (x) = K 1
(10.100)
and in a similar manner one obtains
∂z 21 ∂z 21
∂z 21 ∂z 21
L gb L f h 1 (x) = L gb z 21 ⇒L gb L f h 1 (x) = ∂ x1 gb1 + ∂ x2 gb2
∂ x3 gb3 + + ∂ x4 gb4 ⇒
μ−1
−x3 μx1
L gb L f h 1 (x) = (K 1 K 3 μ
x1 −1
− K 1 K e )gb1 + 0gb2 + ( xKμ1−1
K3
)gb3 + K 1 gb4 ⇒
1
L gb L f h 1 (x) = 0
(10.101)
Following an equivalent procedure one computes
μ−1
−K c μx1
z 22 = L f h 2 (x) = x3 μ 2 (K 1 K c x μx−1
3
− K 1 K c x1 + K 1 x4 )+
x1 −1 1 (10.102)
+ Kc
μ
x1 −1
x3
(−
τ + K o (1 − x2−mu ))

In a similar manner one gets

∂z 22 ∂z 22 ∂z 22 ∂z 22
L 2f h 2 (x) =
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 + ∂ x4 f 4 ⇒
μ−2 μ μ−1 μ μ−1
(−K c μμ−1x1 )(x1 −1)2 −(−K c μx1 )2(x1 −1)μx1 K 1 K c x3
L 2f h 2 (x) = {x3 μ ( x μ −1−K )+
1 K c +K 1 x 4
4
x −1 1 1
μ−1 μ−1
−K c μx1 K 1 K c −x3 μx1
x3 μ
x1 −1
( μ 2 − K 1 K c )+
(x1 −1)
μ−1
−K cμx1 −μ
+ μ 2 (− xτ3 + K o (1 − x2 ))}(K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 )+
x1 −1 1
−μ−1
+ x μK−1
c
K o μx2 (K 2 K e x1 − K 2 x4 )+
1
μ−1
−K c μx1 K1 Kc Kc −μ
+{x3 μ 2 μ
x1 −1
+ μ
x1 −1
(− τ1 )}(− xτ3 + K o (1 − x2 ))+
x1 −1
μ−1
−K c μx1
{x3 μ 2 K 1 }0
x1 −1
(10.103)

Equivalently, one computes

∂z 22 ∂z 22 ∂z 22 ∂z 22
L ga L f h 2 (x) = ∂ x1 ga1 + ∂ x2 ga2 + ∂ x3 ga3 + ∂ x4 ga4 ⇒
∂z 22 −K c μx1
μ−1 (10.104)
L ga L f h 2 (x) = ∂ x4 1⇒L ga L f h 2 (x) = x 3 (x μ −12 ) K 1
1
520 10 Differential Flatness Theory for Internal Combustion Engines

and similarly

∂z 22 ∂z 22 ∂z 22 ∂z 22
L gb L f h 2 (x) = ∂ x1 gb1 + ∂ x2 gb2 + ∂ x3 gb3 + ∂ x4 gb4 ⇒
∂z 22
(10.105)
Kc −μ−1
L gb L f h 2 (x) = ∂ x2 1⇒L gb L f h 2 (x) = μ
x1 −1
K o μx2

Consequently, after the change of coordinates one has the following description

ż 11 = z 21
ż 21= L 2f h 1 (x) +
L ga L f h 1 (x)v1 + L gb L f h 1 (x)v2
(10.106)
ż 12 = z 22
ż 2 = L f h 2 (x) + L ga L f h 2 (x)v1 + L gb L f h 2 (x)v2
2 2

The system’s description takes also the following matrix form

  2    
z̈ 1 L f h 1 (x) L ga L f h 1 (x) L gb L f h 1 (x) v1
= + (10.107)
z̈ 2 L h 2 (x)
2 L L
ga f 2 h (x) L L
gb f 2 h (x) v2
f

that is

z̃¨ = f˜a + M̃a v (10.108)

Moreover, by defining the new control inputs one has

1 = L 2 h (x) + L L h (x)v + L L h (x)v

vin f 1 ga f 1 1 gb f 1 2
2 = L 2 h (x) + L L h (x)v + L L h (x)v
(10.109)
vin f 2 gb f 2 1 gb f 2 2

one has ż 11 = z 21 , ż 21 = vin

1 , ż 2 = z 2 and ż 2 = v2 , the system’s dynamics is written
1 2 2 in
in the canonical Brunovsky form. Moreover, the state-space description of the diesel
engine becomes
⎛ 1⎞ ⎛ ⎞ ⎛ 1⎞ ⎛ ⎞
ż 1 0 1 0 0 z1 0 0  
⎜ 1⎟ ⎜ ⎜ 2⎟ ⎜
⎜ż 2 ⎟ ⎜0 0 0 0⎟ ⎜ z ⎟
⎟ ⎜ 1 ⎟ + ⎜1 0⎟ 1
⎟ vin
⎜ 2⎟ = ⎝ (10.110)
⎝ż 1 ⎠ 0 0 0 1⎠ ⎝z 2 ⎠ ⎝0 0⎠ vin
2
1
2
ż 2 0 0 0 0 z 22 0 1

The selection of the state feedback control law, which assures zeroing of the tracking
error is
1 = z̈ 1 − K 1 (ż 1 − ż 1 ) − K 1 (z 1 − z 1 )
vin 1,d d 1 1,d p 1 1,d
2 = z̈ 2 − K 2 (ż 2 − ż 2 ) − K 2 (z 2 − z 2 )
(10.111)
vin 1,d d 1 1,d p 1 1,d
10.3 Flatness-Based Control of Diesel Combustion Engines 521

Since

ṽin = f˜a + M̃a ṽ⇒ṽ = M̃a−1 (ṽin − f˜a ) (10.112)

Additionally, using
   
v u̇ 1
ṽ = 1 = (10.113)
v2 u2

it can be seen that to compute the control input u 1 which is actually applied to the
diesel motor, an integration with respect to time has to be carried out for the auxiliary
control input.

10.3.5 Nonlinear Control of the Diesel Engine Using

Differential Flatness Theory

The results about dynamic state feedback system linearization can be confirmed with
the computation of time derivatives and differential flatness theory. The following
differentially flat system outputs are considered

y1 = p1 = x1
y2 = Pc pμK−1
c
⇒y2 = x3 x μK−1
c (10.114)
1 1

The dynamics of the extended system is

⎛ ⎞ ⎛ ⎞ ⎛ ⎞ ⎛ ⎞
ẋ1 K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 0 0
⎜ẋ2 ⎟ ⎜ 1
− ⎟ ⎜0 ⎟ ⎜1 ⎟
⎜ ⎟=⎜ K K x
2 e 1 K x
2 4 ⎟ + ⎜ ⎟ v1 + ⎜ ⎟ v2 (10.115)
⎝ẋ3 ⎠ ⎝ −x3 μ ⎠ ⎝0 ⎠ ⎝0⎠
τ + K o (1 − x 2 )
ẋ4 0 1 0

It holds that y1 = x1 therefore

x1 = q1 (y, ẏ)
μ μ
y2 (x1 −1) y2 (y1 −1) (10.116)
y2 = x3 x μK−1
c
⇒x3 = Kc ⇒x3 = Kc ⇒x3
1

therefore x3 = q3 (y, ẏ) and thus variable x3 is a function of the flat output and its
derivatives. From the first row of the state-space equations one has

ẋ1 = K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 ⇒
1 x3
ẋ1 −K 1 K e μ +K 1 K e x1
x1 −1 (10.117)
x4 = K1 ⇒
x4 = q4 (y, ẏ)
522 10 Differential Flatness Theory for Internal Combustion Engines

From the 4th row of the state-space equations one has

ẋ4 = v1 ⇒v1 = q5 (y, ẏ) (10.118)

where v1 is a function of the flat output and its derivatives. From the 3rd row of the
state-space equations one has
−μ
ẋ3 = − xτ3 + K o (1 −x2 )⇒ 
x x μ
−μ ẋ + 3 −ẋ3 + τ3 (10.119)
x2 = − 3K o τ ⇒x2 = Ko

From the second row of the state-space equations one obtains

ẋ2 = K 2 K e x1 − K 2 x4 + v2 ⇒
v2 = ẋ2 − K 2 K e x1 + K 2 x4 ⇒ (10.120)
v2 = q6 (y, ẏ)

Therefore, all state variables of the system and the control inputs can be written as
functions of the flat output and its derivatives. Therefore, the system of the diesel
engine is differentially flat and can be subjected to dynamic feedback linearization.
By considering the flat output

y1 = x1
y2 = x3 x μK−1
c (10.121)
1

and by differentiating with respect to time one obtains the linearized model of the
system

y1 = x1
ẏ1 = ẋ1 ⇒ ẏ1 = K 1 K c x μx−1
3
− K 1 K c x1 + K 1 x4 (10.122)
1

and equivalently
μ−1
−x3 μx1
ÿ1 = (K 1 K c μ 2 − K 1 K c )ẋ1 + 0 ẋ2 + (K 1 K c x μ1−1 )ẋ3 + K 1 ẋ4 (10.123)
x1 −1 1

or
μ−1
−x3 μx1
ÿ1 = (K 1 K c μ 2 − K 1 K c )(K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 )+
x1 −1 1 (10.124)
−μ
+(K 1 K c x μ1−1 )(− xτ3 + K o (1 − x2 )) + K 1 v1
1
10.3 Flatness-Based Control of Diesel Combustion Engines 523

In a similar manner, one has

y2 = x3 x μK−1
c
1
μ−1
x3 K c μx1 Kc
ẏ2 = μ 2 ẋ1 + μ ẋ ⇒
x1 −1 3
x1 −1
μ−1
x3 K c μx1 Kc
ẏ2 = μ 2 ẋ1 + μ ẋ ⇒
x1 −1 3
x1 −1
μ−1
x3 K c μx1 Kc −μ
ẏ2 = μ 2 (K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 ) + μ
x1 −1
(− xτ3 + K o (1 − x2 ))
x1 −1 1
(10.125)
and equivalently it holds
∂ ẏ2 ∂ ẏ2 ∂ ẏ2 ∂ ẏ2
ÿ2 = ∂ x1 ẋ 1 + ∂ x2 ẋ 2 + ∂ x3 ẋ 3 + ∂ x4 ẋ 4 ⇒ (10.126)

or
μ−2 μ μ−1 μ μ−1
x3 K c μ(μ−1)x1 (x1 −1)2 −x3 K c μx1 2(x1 −1)μx1
ÿ2 = { μ
(x1 −1)4
(K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 )+
1
μ−1
−K c μx1 μ Kc μ−1
+ μ 2 (− xτ3 + K o (1 − x2 ))}ẋ1 + μ
x1 −1
(K o μx2 )ẋ2 +
(x1 −1)
μ−1
K c μx1 Kc
+{ μ 2 (K 1 K c (x μx−1)
3
− K 1 K e x1 + K 1 x4 + μ
x1 −1
(− τ1 ))}ẋ3 +
(x1 −1) 1
μ−1
x3 K c μx1
+ μ 2 K 1 ẋ4
x1 −1
(10.127)

Therefore, it holds
μ−2 μ μ−1 μ μ−1
x3 K c μ(μ−1)x1 (x1 −1)2 −x3 K c μx1 2(x1 −1)μx1
ÿ2 = { μ
(x1 −1)4
(K 1 K c x μx−1
3
− K 1 K e x1 +
1
μ−1
−K c μx1 μ Kc
K 1 x4 ) + μ 2 (− xτ3 + K o (1 − x2 ))}(K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 ) + μ
x1 −1
(x1 −1) 1
μ−1
μ−1 K c μx1
(K o μx2 )(K 2 K e x1 − K 2 x4 + v2 ) + { μ 2 (K 1 K c (x μx−1)
3
− K 1 K e x1 + K 1 x4 +
(x1 −1) 1
μ−1
Kc −μ x3 K c μx1
μ
x1 −1
(− τ1 ))}(− xτ3 + K o (1 − x2 )) + μ 2 K 1 }v1
x1 −1

Therefore, in complete analogy to the relations computed with the use of Lie algebra
it holds

ÿ2 = L 2f h 2 (x) + L ga L f h 2 v1 + L gb L f h 2 v2 (10.128)

where μ−2 μ μ−1 μ μ−1

x K μ(μ−1)x1 (x1 −1)2 −x3 K c μx1 2(x1 −1)μx1
L 2f h 2 (x) = { 3 c μ
(x −1)4
(K 1 K c x μx−1
3
− K1 Ke
1 1
μ−1
−K c μx1 μ
x1 + K 1 x4 ) + μ 2 (− xτ3 + K o (1 − x2 ))}(K 1 K c x μx−1
3
− K 1 K e x1 + K 1 x4 )+
(x1 −1) 1
524 10 Differential Flatness Theory for Internal Combustion Engines

μ−1
Kc μ−1 K μx
μ
x1 −1
(K o μx2 )(K 2 K e x1 − K 2 x4 ) + { cμ 1 2 (K 1 K c x μx−1
3
) − K 1 K e x1 + K 1 x4 +
(x1 −1) 1
Kc x3 −μ
μ
x1 −1
(− 1
τ )}(− τ + K o (1 − x 2 ))
and also
μ−1
x3 K c μx1
L ga L f h 2 (x) = μ
(x1 −1)2
K1 (10.129)

Kc μ−1
L gb L f h 2 (x) = μ
x1 −1
(K o μx2 ) (10.130)

Therefore, one is led again in a description of the system in the following form

ÿ1 = L 2f h 1 (x) + L ga L f h 1 (x)u 1 + L gb L f h 1 (x)u 2

(10.131)
ÿ2 = L 2f h 2 (x) + L ga L f h 2 (x)u 1 + L gb L f h 2 (x)u 2

Again, by defining the new control inputs one has

1 = L 2 h (x) + L L h (x)v + L L h (x)v

vin f 1 ga f 1 1 gb f 1 2
2 = L 2 h (x) + L L h (x)v + L L h (x)v (10.132)
vin f 2 gb f 2 1 gb f 2 2

one has ż 11 = z 21 , ż 21 = vin

1 , ż 2 = z 2 and ż 2 = v2 , the system’s dynamics is written
1 2 2 in
in the canonical Brunovsky form. Moreover, the state-space description of the diesel
engine becomes
⎛ 1⎞ ⎛ ⎞ ⎛ 1⎞ ⎛ ⎞
ż 1 0 1 0 0 z1 0 0  
⎜ż 1 ⎟ ⎜0 0 0 0⎟ ⎜ z 2 ⎟ ⎜1 0⎟ 1
⎜ 2⎟ = ⎜ ⎟ ⎜ 1⎟ + ⎜ ⎟ vin (10.133)
⎝ż 2 ⎠ ⎝0 0 0 1⎠ ⎝z 12 ⎠ ⎝0 0 ⎠ 2
vin
1
ż 22 0 0 0 0 z 22 0 1

The design of a feedback controller for the above state-space model is as analyzed in
the case of linearization with the use of Differential Geometry (computation of Lie
derivatives).
The selection of the state feedback control law, which assures zeroing of the
tracking error is

1 = z̈ 1 − K 1 (ż 1 − ż 1 ) − K 1 (z 1 − z 1 )
vin 1,d d 1 1,d p 1 1,d
2 = z̈ 2 − K 2 (ż 2 − ż 2 ) − K 2 (z 2 − z 2 )
(10.134)
vin 1,d d 1 1,d p 1 1,d

Since

ṽNN = f˜a + M̃a ṽ⇒ṽ = M̃a−1 (ṽin − f˜a ) (10.135)

10.3 Flatness-Based Control of Diesel Combustion Engines 525

Moreover, with it holds

   
v1 u̇ 1
ṽ = = (10.136)
v2 u2

to compute the control input u 1 which is actually applied to the diesel motor an
integration with respect to time has to be carried out for the auxiliary control input.

10.3.6 Disturbances Compensation Using the Derivative-Free

Nonlinear Kalman Filter

It is assumed that model uncertainty effects and external perturbation terms are
described in the model of the diesel engine as additive disturbance inputs which
appear in the linearized equivalent. Thus, one has the dynamics

ÿ1 = L 2f h 1 (x) + L ga h 1 (x)v1 + L gb L f h 2 (x)v2 + d̃1

(10.137)
ÿ2 = L 2f h 2 (x) + L ga h 2 (x)v1 + L gb L f h 2 (x)v2 + d̃2

and after defining v1 = L 2f h 1 (x)+L ga h 1 (x)v1 +L gb L f h 2 (x)v2 and v2 = L 2f h 2 (x)+

L ga h 2 (x)v1 + L gb L f h 2 (x)v2 one gets

ÿ1 = vin
1 + d̃
1
(10.138)
ÿ2 = vin
2 + d̃
2

Without loss of generality, it is assumed that the disturbance terms are now described
by the associated second-order derivatives, that is,

d̃¨1 = f d1
(10.139)
d̃¨2 = f d2

In the latter case, the system’s dynamics can be extended by considering as additional
state variables the disturbance terms and their derivatives. Thus ones obtains

z 1 = p1 z 2 = ż 1 z 3 = Pc pμK−1c
z 4 = ż 3
1 (10.140)
˙
z 5 = d̃1 z 6 = d̃1 z 7 = d̃2 z 8 = d̃˙2

which gives

ż = Az + B ṽin
(10.141)
zm = C z
526 10 Differential Flatness Theory for Internal Combustion Engines

where z = [z 1 , z 2 , . . . , z 8 ]T , ṽin = [v1 , v2 , f d1 , f d2 ]T and matrices A, B and C are

defined as follows
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
01000000 0000 10
⎜0 0 0 0 1 0 0 0⎟ ⎜1 0 0 0 ⎟ ⎜0 1⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0⎟ ⎜0 0 0 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 1 0⎟ ⎜0 1 0 0⎟ T ⎜0 0⎟
A=⎜ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟
0 0 0 0 0 1 0 0 ⎟ B = ⎜0 0 0 0⎟ C = ⎜0 0⎟ (10.142)
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0⎟ ⎜0 0 1 0⎟ ⎜0 0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 1⎠ ⎝0 0 0 0⎠ ⎝0 0⎠
00000000 0001 00

In the design of the Kalman Filter-based disturbances estimator it is assumed that

the disturbances’ dynamics is completely unknown. Thus, the considered dynamics
now is

ẑ = Ao ẑ + Bo vin + K [z m − ẑ m ]⇒
(10.143)
ẑ = Ao ẑ + Bo vin + K [C z − C ẑ]

where vin = [v1 , v2 ]T , Ao = A, Co = C and

 
01000000
BoT = (10.144)
00010000

For the above definition of dynamics of the disturbances estimator, the selection of
the observer’s gain K can be performed using the standard Kalman Filter recursion.
Prior to this, matrices Ao , Bo , and Co are brought to the discrete-time form Ad ,
Bd , and Cd using common discretization methods. The discrete-time Kalman Filter
recursion is [31, 408, 414]
measurement update:

K (k) = P − (k)CdT [Cd P − (k)CdT + R]−1

ẑ(k) = x̂ − (k) + K (k)[z m (k) − ẑ m (k)] (10.145)
P(k) = P − (k) − K (k)Cd P − (k)

time update:

P − (k + 1) = Ad P(k)AdT + Q
(10.146)
ẑ(k + 1) = Ad ẑ(k) + Bd vin (k)

From the previous estimation procedure one can reconstruct the state vector of the
initial nonlinear model of the diesel engine. It holds that y1 = z 1 , ẏ1 = z 2 , y2 = z 3
and ẏ2 = z 4 , where [y1 , y2 ] is the flat output of the system. Thus, one obtains
10.3 Flatness-Based Control of Diesel Combustion Engines 527

μ
( ŷ1 −1)
x̂1 = ŷ1 x̂3 = ŷ2 Kc

 μ (10.147)
ŷ˙1 −K 1 K e
x̂3
+K 1 K e x̂1
x̂˙3 + τ3 μ
x̂
x̂1 −1
x̂2 = − Ko x̂4 = K1

Thus, by measuring only the gas pressure in the intake manifold p1 and the com-
pressor power Pc one can also obtain estimates of the gas pressure in the exhaust
manifold p2 .
Once the disturbance terms have been estimated, it is possible to compensate them
by including their estimated values in the control signal. In such a case the control
signal becomes
1,∗
vin = vin
1 − ẑ
5
2,∗ (10.148)
vin = vin2 − ẑ
7

10.3.7 Simulation Tests

Through simulation experiments it has been confirmed that the proposed control and
Kalman Filter-based estimation scheme can (i) succeed convergence of the elements
of the state vector of the turbocharged diesel engine to the desirable setpoints, (ii)
estimate nonmeasurable elements of the state vector as well as disturbance terms that
affect the engine’s dynamics. Indicative results are presented in Fig. 10.15a, b where
the convergence of the state vector elements to the associated setpoints are depicted.

(a) 6 40
(b) 3 1.5
d/dt d − d/dt d − est
d1 − d1 − est

2 1
1
2
1

4 20
xe1 − xed

xe − xed

1 0.5
2

2 0
1

0 0

0 −20 −1 −0.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time time time time

0.6 300 3 1.5

d/dt d − d/dt d − est
d − d − est

0.4 2 1
2
xe 3 − xed 3

xe4 − xed4

200
2

0.2 1 0.5

100
2
2

0 0 0

−0.2 0 −1 −0.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time time time time

Fig. 10.15 a Convergence of the state variables to the associated setpoints 1 (red line setpoint, blue
line real value, green line estimated value), b estimation (blue line) of perturbation terms (red line)
affecting the diesel engine
528 10 Differential Flatness Theory for Internal Combustion Engines

(a) 6 100 (b) 3 1.5

d/dt d1 − d/dt d1 − est

2 1

d1 − d1 − est
2
xe − xed1

4 50

xe 2 − xed
1 0.5
1

2 0
0 0

0 −50 −1 −0.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time time time time

1 300 3 1.5

d/dt d2 − d/dt d − est

d − d − est
2 1

2
xe 3 − xed 3

xe 4 − xed4

0.5 200

1 0.5

2
0 100

2
0 0

−0.5 0 −1 −0.5
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
time time time time

Fig. 10.16 a Convergence of the state variables to the associated setpoints 2 (red line setpoint, blue
line real value, green line estimated value), b estimation (blue line) of perturbation terms (red line)
affecting the diesel engine

It can be confirmed that the proposed differential flatness theory-based filtering

and control method for turbocharged diesel engines succeeds both good tracking of
the reference setpoints and good transients (Fig. 10.16).

10.4 Adaptive Control for Diesel Combustion Engines

10.4.1 Overview

In this section the nonlinear control problem of the turbocharged diesel engines is
solved with the use of differential flatness theory and of adaptive fuzzy control.
It has been shown that the dynamic model of the turbocharged diesel engine is
differentially flat and admits dynamic feedback linearization. It has been also shown
that the dynamic model can be written in the linear Brunovsky canonical form for
which a state feedback controller can be easily designed. To cope with modeling
uncertainties and external disturbances this section proposes to implement adaptive
fuzzy control, with the use of the transformed diesel engine’s dynamical model that
was obtained through the application of differential flatness theory.
Actually, the section develops a solution for the problem of observer-based adap-
tive fuzzy control of diesel engines and in general for MIMO nonlinear dynamical
systems that admit dynamic feedback linearization. The design of the adaptive fuzzy
controller considers that only the system’s output is measured and that the system’s
model is unknown. The control algorithm aims at satisfying the H∞ tracking per-
formance criterion, which also means that improved robustness is succeeded against
modeling errors and the external disturbances. After transforming the MIMO system
10.4 Adaptive Control for Diesel Combustion Engines 529

of the diesel engine into the canonical form, the resulting control inputs are shown
to contain nonlinear elements which depend on the system’s parameters. Since the
parameters of the system are unknown, then the nonlinear terms which appear in
the control inputs have to be approximated with the use of neuro-fuzzy networks.
Moreover, since only the system’s output is measurable the complete state vector
of the diesel engine has to be reconstructed with the use of a state observer. As in
the examples on adaptive fuzzy control which were presented in the previous chap-
ters, in this section it is shown that a suitable learning law can be defined for the
aforementioned neuro-fuzzy approximators so as to preserve the closed-loop system
stability. Lyapunov stability analysis proves also that the proposed observer-based
adaptive fuzzy control scheme results in H∞ tracking performance, in accordance
to the results of [407, 410, 413, 429, 433, 454].
For the design of the observer-based adaptive fuzzy controller one has to solve
two Riccati equations, where the first one is associated with the controller and the
second one is associated with the observer. Parameters that affect the closed-loop
robustness are: (i) the feedback gain vector K , (ii) the observer’s gain vector K o , (iii)
the positive definite matrices P1 and P2 which are computed from the solution of the
two algebraic Riccati equations and which weigh the above-mentioned observer and
controller terms. The proposed control architecture guarantees that the output of the
diesel engine control loop will asymptotically track the desired trajectory and that
H∞ performance will be achieved. The efficiency of the proposed control method is
confirmed through simulation experiments.

10.4.2 Observer-Based Adaptive Fuzzy Control for the Diesel

Combustion Engine

10.4.2.1 Control Law

According to the preceding sections, the MIMO nonlinear model of the diesel engine
can be transformed into the Brunovsky form as described by Eq. (3.49) or equivalently
to the trivial form of Eq. (5.78). This contains the unknown nonlinear functions f (x)
and g(x). Thus, the transformed diesel engine model is written in the form

(ri ) p
yi = f i (x) + j=1 gi j (x)u j + dj (10.149)

which can be also written in the vector form

y (r ) = f (x) + g(x)u + d (10.150)

In case that the complete state vector x is measurable, these unknown functions can
be approximated by
530 10 Differential Flatness Theory for Internal Combustion Engines

fˆ(x|θ f ) = Φ f (x)θ f
(10.151)
ĝ(x|θg ) = Φg (x)θg

where

Φ f (x) = (ξ 1f (x), ξ 2f (x), . . . , ξ nf (x))T (10.152)

with ξ if (x), ı = 1, . . . , n being the vector of kernel functions (e.g. normalized fuzzy
Gaussian membership functions), where

ξ if (x) = (φ i,1 i,2 i,N

f (x), φ f (x), . . . , φ f (x)) (10.153)

thus giving
⎛ ⎞
φ 1,1 1,2
f (x) φ f (x) · · · φ 1,N
f (x)
⎜ 2,1 2,2 ⎟
⎜φ (x) φ f (x) · · · φ 2,N
f (x)⎟
Φ f (x) = ⎜ f ⎟ (10.154)
⎝ ··· ··· ··· ··· ⎠
φ n,1
f (x) φ n,2
f (x) · · · φ n,N
f (x)

while the weights vector is defined as

 
θ f T = θ 1f , θ 2f , . . . , θ Nf (10.155)

j = 1, . . . , N is the number of basis functions that is used to approximate the

components of function f which are denoted as i = 1, . . . , n. Thus, one obtains the
relation of Eq. (10.151), i.e., fˆ(x|θ f ) = Φ f (x)θ f .
In a similar manner, for the approximation of function g one has

T
Φg (x) = ξg1 (x), ξg2 (x), . . . , ξgN (x) (10.156)

with ξgi (x), ı = 1, . . . , N being the vector of kernel functions (e.g., normalized fuzzy
Gaussian membership functions), where
 
ξgi (x) = φgi,1 (x), φgi,2 (x), . . . , φgi,N (x) (10.157)

thus giving
⎛ ⎞
φg1,1 (x) φg1,2 (x) · · · φg1,N (x)
⎜φg2,1 (x) φg2,2 (x) · · · φg2,N (x)⎟
Φg (x) = ⎜
⎝ ···
⎟ (10.158)
··· ··· ··· ⎠
φgn,1 (x) φgn,2 (x) · · · φgn,N (x)
10.4 Adaptive Control for Diesel Combustion Engines 531

while the weights vector is defined as

p

θg = θg1 , θg2 , . . . , θg (10.159)

where the components of matrix θg are defined as

 T
j
θg = θgj1 , θgj2 , . . . , θgjN (10.160)

j = 1, . . . , p is the number of basis functions that is used to approximate the

components of function g which are denoted as i = 1, . . . , n. Thus one obtains
about matrix θg ∈R N × p
⎛ 1 2 p ⎞
θg1 θg1 · · · θg1
⎜ θg1 θg2 · · · θg ⎟p
θg = ⎜ 2 2
⎝··· ··· ··· ···⎠
2⎟
(10.161)
p
θg1N θg2N · · · θg N

It holds that
⎛ ⎞ ⎛ 1 p⎞
g1 g1 g12 · · · g1
⎜ g2 ⎟ ⎜ g 1 g22 · · · g2 ⎟
p
g=⎜ ⎟ ⎜ 2
⎝· · ·⎠ = ⎝· · ·
⎟ (10.162)
··· ··· · · ·⎠
p
gn gn1 gn2 · · · gn

Using the above, one finally has the relation of Eq. (10.151), i.e., ĝ(x|θg ) = Φg (x)θg .
If the state variables of the system are available for measurement, then a state feedback
control law can be formulated as
(r )
u = ĝ −1 (x|θg )[− fˆ(x|θ f ) + ym − K T e + u c ] (10.163)

where fˆ(x|θ f ) and ĝ(x|θg ) are neuro-fuzzy models to approximate f (x) and g(x),
respectively. u c is a supervisory control term, e.g., H∞ control term that is used to
compensate for the effects of modeling inaccuracies and external disturbances. More-
over, K T is the feedback gain matrix that assures that the characteristic polynomial
of matrix A − B K T will be a Hurwitz one.

10.4.2.2 Estimation of the Diesel Engine’s State Vector

The control of the system described by Eq. (10.149) or Eq. (10.150), becomes more
complicated when the state vector x is not directly measurable and has to be recon-
structed through a state observer. The following definitions are used
532 10 Differential Flatness Theory for Internal Combustion Engines

• error of the state vector e = x − xm

• error of the estimated state vector ê = x̂ − xm
• observation error ẽ = e − ê = (x − xm ) − (x̂ − xm )

When an observer is used to reconstruct the state vector, the control law of
Eq. (10.163) is written as

u = ĝ −1 (x̂|θg )[− fˆ(x̂|θ f ) + ym(r ) − K T ê + u c ] (10.164)

Applying Eq. (10.164) to the nonlinear system described in the form of Eq. (10.150),
results into
(r )
y (r ) = f (x) + g(x)ĝ −1 (x̂)[− fˆ(x̂) + ym − K T ê + u c ] + d⇒
(r )
y (r ) = f (x) + [g(x) − ĝ(x̂) + ĝ(x̂)]ĝ (x̂)[− fˆ(x̂) + ym − K T ê + u c ] + d⇒
−1
(r )
y (r ) = [ f (x) − fˆ(x̂)] + [g(x) − ĝ(x̂)]u + ym − K T ê + u c + d
(10.165)
(r )
It holds e = x − xm ⇒ y (r ) = e(r ) + ym . Substituting y (r ) in the above equation
gives
(r ) (r )
e(r ) + ym = ym − K T ê + u c + [ f (x) − fˆ(x̂)]+
(10.166)
+[g(x) − ĝ(x̂)]u + d

and equivalently

ė = Ae − B K T ê + Bu c + B{[ f (x) − fˆ(x̂)]+

(10.167)
+[g(x) − ĝ(x̂)]u + d̃}

e1 = C T e (10.168)

where e = [e1 , e2 , . . . , e p ]T with ei = [ei , ėi , ëi , . . . , eiri −1 ]T , i = 1, 2, . . . , p

and equivalently ê = [ê1 , ê2 , . . . , ê p ]T with êi = [êi , ėˆi , ëˆi , . . . , êiri −1 ]T , i =
1, 2, . . . , p. Matrices A, B and C have been defined in Eq. (3.48).
A state observer is designed according to Eqs. (10.167) and (10.168) and is given
by [452]:

ê˙ = Aê − B K T ê + K o [e1 − C T ê] (10.169)

ê1 = C T ê (10.170)

The feedback gain matrix is denoted as K ∈R n× p . The observation gain matrix is

denoted as K o ∈R n× p and its elements are selected so as to assure the asymptotic
elimination of the observation error.
10.4 Adaptive Control for Diesel Combustion Engines 533

10.4.3 Application of Flatness-Based Adaptive Fuzzy Control

to the MIMO Diesel Engine Model

10.4.3.1 Differential Flatness of the Diesel Engine

After the application of differential flatness theory and the associated state variables
transformation, the initial nonlinear model of the diesel engine given in Eq. (10.64)
can be written in an equivalent state-space form

ẍ1 = f 1 (x) + g1 (x)u

(10.171)
ẍ3 = f 2 (x) + g2 (x)u

It holds that

ẋ1 = x2
ẋ2 = f 1 (x) + g1 (x)u
(10.172)
ẋ3 = x4
ẋ4 = f 2 (x) + g2 (x)u

The design of a feedback linearizing controller for the MIMO model of the diesel
engine and the associated stability proof, follows the stages which were analyzed in
Sect. 5.3. Actually, from Eq. (10.172) it holds
     
ẍ1 f 1 (x) g1 (x)
= + u i.e.
ẍ3 f 2 (x) g2 (x)
(10.173)
 −1    
g (x) ẍ f (x)
u= 1 { 1 − 1 }
g2 (x) ẍ3 f 2 (x)

Therefore, the considered robotic system is a differentially flat one. Next, taking into
account also the effects of additive disturbances the dynamic model becomes

ẍ1 = f 1 (x, t) + g1 (x, t)u + d1

(10.174)
ẍ3 = f 2 (x, t) + g2 (x, t)u + d2
       
ẍ1 f (x, t) g (x, t) d
= 1 + 1 u+ 1 (10.175)
ẍ3 f 2 (x, t) g2 (x, t) d2

The following control input is defined

 −1  d     T  
ĝ1 (x, t) ẍ fˆ (x, t) K1 u c1
u= { 1d − ˆ1 − e + } (10.176)
ĝ2 (x, t) ẍ3 f 2 (x, t) K 2T u c2
534 10 Differential Flatness Theory for Internal Combustion Engines

where [u c1 u c2 ]T is a robust control term that is used for the compensation of the
model’s uncertainties as well as of disturbances and K iT = [k1i , k2i , . . . , kn−1
i , kni ].
Substituting Eq. (10.176) into Eq. (10.175) the closed-loop tracking error dynamics
is obtained
      −1
ẍ1 f 1 (x, t) g1 (x, t) ĝ1 (x, t)
= + ·
ẍ3 f 2 (x, t) g2 (x, t) ĝ2 (x, t)
(10.177)
     T    
ẍ d fˆ1 (x, t) K1 u c1 d
·{ 1d − ˆ − T e+ u }+ 1
ẍ3 f 2 (x, t) K 2 c2 d2

Equation (10.177) can now be written as

     
ẍ1 f 1 (x, t) g1 (x, t) − ĝ1 (x, t)
= +{ +
ẍ3 f 2 (x, t) g2 (x, t) − ĝ2 (x, t)
   −1  d     T    
ĝ1 (x, t) ĝ (x, t) ẍ fˆ (x, t) K1 u c1 d1
+ } 1 ·{ 1d − ˆ1 − e + } +
ĝ2 (x, t) ĝ2 (x, t) ẍ3 f 2 (x, t) K 2T u c2 d2
(10.178)
and using Eq. (10.176) this results into
     
ë1 f (x, t) − fˆ1 (x, t) g1 (x, t) − ĝ1 (x, t)
= 1 + u−
ë3 f 2 (x, t)− fˆ2(x, t) g2 (x, 
t) − ĝ2 (x, t)
 (10.179)
K 1T u d
− e + c1 + 1
K 2T u c2 d2

The following description for the approximation error is defined

   
f 1 (x, t) − fˆ1 (x, t) g1 (x, t) − ĝ1 (x, t)
w= + u (10.180)
f 2 (x, t) − fˆ2 (x, t) g2 (x, t) − ĝ2 (x, t)

Moreover, the following matrices are defined

⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0
⎜0 0 0 0⎟ ⎜1 0⎟
A=⎜
⎝0
⎟, B = ⎜ ⎟
0 0 1⎠ ⎝0 0⎠
0 0 0 0 0 1 (10.181)
 
K 11 K 21 K 31 K 41
KT =
K 12 K 22 K 32 K 42
10.4 Adaptive Control for Diesel Combustion Engines 535

Using matrices A, B, K T , Eq. (10.179) is written in the following form

 
f 1 (x, t) − fˆ1 (x, t)
ė = (A − B K T )e + Bu c + B{ +
f 2 (x, t) − fˆ2 (x, t)
(10.182)
 
g1 (x, t) − ĝ1 (x, t)
+ u + d̃}
g2 (x, t) − ĝ2 (x, t)

When the estimated state vector x̂ is used in the feedback control loop, equivalently
to Eq. (10.167) one has
 
f (x, t) − fˆ1 (x̂, t)
ė = Ae − B K T ê + Bu c + B{ 1 +
  f 2 (x, t) − fˆ2 (x̂, t) (10.183)
g (x, t) − ĝ1 (x̂, t)
+ 1 u + d̃}
g2 (x, t) − ĝ2 (x̂, t)

and considering that the approximation error w is now denoted as

   
f 1 (x, t) − fˆ1 (x̂, t) g1 (x, t) − ĝ1 (x̂, t)
w= + u (10.184)
f 2 (x, t) − fˆ2 (x̂, t) g2 (x, t) − ĝ2 (x̂, t)

Equation (10.183) can be also written as

ė = Ae − B K T ê + Bu c + Bw + B d̃ (10.185)

The associated state observer will be described again by Eqs. (10.169) and (10.170).

10.4.3.2 Dynamics of the Observation Error

The observation error for the MIMO model of the diesel engine is defined as ẽ =
e − ê = x − x̂. Subtracting Eq. (10.169) from Eq. (10.167) as well as Eq. (10.170)
from Eq. (10.168) one gets

ė − ê˙ = A(e − ê) + Bu c + B{[ f (x, t) − fˆ(x̂, t)]+

+[g(x, t) − ĝ(x̂, t)]u + d̃} − K o C T (e − ê)
e1 − ê1 = C T (e − ê)

or equivalently

ẽ˙ = Aẽ + Bu c + B{[ f (x, t) − fˆ(x̂, t)] + [g(x, t) − ĝ(x̂, t)]u + d̃} − K o C T ẽ
ẽ1 = C T ẽ
536 10 Differential Flatness Theory for Internal Combustion Engines

which can be written as

ẽ˙ = (A − K o C T )ẽ + Bu c + B{[ f (x, t) − fˆ(x̂, t)] + [g(x, t) − ĝ(x̂, t)]u + d̃}
(10.186)

ẽ1 = C T ẽ (10.187)

or equivalently, it can be written as

ẽ˙ = (A − K o C T )ẽ + Bu c + B{w + d̃} (10.188)

ẽ1 = C T ẽ (10.189)

10.4.3.3 Approximation of the Unknown Dynamics of the Diesel Engine

Next, the following approximators of the unknown system dynamics are defined
 
fˆ1 (x̂|θ f ) x̂∈R 4×1 fˆ1 (x̂|θ f ) ∈ R 1×1
fˆ(x̂) = (10.190)
fˆ2 (x̂|θ f ) x̂∈R 4×1 fˆ2 (x̂|θ f ) ∈ R 1×1

with kernel functions n

j=1 μ A j ( x̂ j )
i
i, j
φ f (x̂) =  N n (10.191)
i=1 j=1 μ A ( x̂ j )
i
j

where l = 1, 2, x̂ is the estimate of the state vector and μ Ai (x̂) is the ith membership
j
function of the antecedent (IF) part of the lth fuzzy rule. Similarly, the following
approximators of the unknown system dynamics are defined (Fig. 10.17)

Fig. 10.17 Neuro-fuzzy approximator used for estimating the unknown system dynamics
10.4 Adaptive Control for Diesel Combustion Engines 537
 
ĝ1 (x̂|θg ) x̂∈R 4×1 ĝ1 (x̂|θg ) ∈ R 1×2
ĝ(x̂) = (10.192)
ĝ2 (x̂|θg ) x̂∈R 4×1 ĝ2 (x̂|θg ) ∈ R 1×2

The values of the weights that result in optimal approximation are

θ ∗f = arg minθ f ∈Mθ f [supx̂∈Ux̂ ( f (x) − fˆ(x̂|θ f ))]

(10.193)
θg∗ = arg minθg ∈Mθg [supx̂∈Ux̂ (g(x) − ĝ(x̂|θg ))]

where the variation ranges for the weights are defined as

Mθ f = {θ f ∈R h : ||θ f ||≤m θ f }
(10.194)
Mθg = {θg ∈R h : ||θg ||≤m θg }

The value of the approximation error defined in Eq. (10.180) that corresponds to the
optimal values of the weights vectors θ ∗f and θg∗ one has
 

w= f (x, t) − fˆ(x̂|θ ∗f ) + g(x, t) − ĝ(x̂|θg∗ ) u (10.195)

which is next written as

 
w = f (x, t) − fˆ(x̂|θ f ) + fˆ(x̂|θ f ) − fˆ(x̂|θ ∗f ) +

(10.196)
+ g(x, t) − ĝ(x̂|θg ) + ĝ(x̂|θg ) − ĝ(x̂|θg∗ ) u

which can be also written in the following form

w = wa + wb (10.197)

where

wa = {[ f (x, t) − fˆ(x̂|θ f )] + [g(x, t) − ĝ(x̂|θg )]}u (10.198)

wb = {[ fˆ(x̂|θ f ) − fˆ(x̂|θ ∗f )] + [ĝ(x̂, θg ) − ĝ(x̂|θg∗ )]}u (10.199)

Moreover, the following weights error vectors are defined

θ̃ f = θ f − θ ∗f
(10.200)
θ̃g = θg − θg∗
538 10 Differential Flatness Theory for Internal Combustion Engines

10.4.4 Lyapunov Stability Analysis

10.4.4.1 Design of the Lyapunov Function

The stages of Lyapunov stability analysis for observer-based adaptive neuro-fuzzy

control of MIMO nonlinear dynamical systems, based exclusively on output feed-
back, have been analyzed in Chap. 5. The Lyapunov stability proof which is followed
in the case of the diesel engine is similar. The adaptation law of the neuro-fuzzy
approximators weights θ f and θg as well as the equation of the supervisory control
term u c are derived from the requirement for negative definiteness of the Lyapunov
function

V = 21 ê T P1 ê + 21 ẽ T P2 ẽ + 2γ1 θ̃ f θ̃ f
1 T
+ 2γ2 tr [θ̃g θ̃g ]
1 T
(10.201)

The selection of the Lyapunov function is based on the following principle of indirect
adaptive control ê : limt→∞ x̂(t) = xd (t) and ẽ : limt→∞ x̂(t) = x(t). This yields
limt→∞ x(t) = xd (t). Substituting Eqs. (10.169), (10.170), (10.186), and (10.187)
into Eq. (10.201) and differentiating results into

V̇ = 21 ê˙ T P1 ê + 21 ê T P1 ê˙ + 21 ẽ˙ T P2 ẽ + 21 ẽ T P2 ẽ+

˙
1 ˙T ˙ T (10.202)
+ θ̃ θ̃ + tr [θ̃ θ̃ ] ⇒
γ1 f f
1
γ2 g g

V̇ = 21 {(A − BK T )ê + K o C T ẽ}T P1 ê + 21 ê T P1 {(A − BK T )ê + K o C T ẽ}+

+ 21 {(A − K o C T )ẽ + Bu c + B d̃ + Bw}T P2 ẽ+
(10.203)
+ 21 ẽ T P2 {(A − K o C T )ẽ + Bu c + B d̃ + Bw}+
+ 1 θ̃˙ T θ̃ + 1 tr [θ̃˙ θ̃ ] ⇒
T
γ1 f f γ2 g g

V̇ = 21 {ê T (A − BK T )T + ẽ T C K oT }P1 ê + 21 ê T P1 {(A − BK T )ê + K o C T ẽ}+

+ 21 {ẽ T (A − K o C T )T + u cT B T + wT B T + d̃ T B T }P2 ẽ+
ẽ P {(A − K C T )ẽ + Bu + Bw + B d̃} + 1 θ̃˙ T θ̃ + 1 tr [θ̃˙ θ̃ ] ⇒
T
1 T
2 2 o c γ1 f f γ2 g g
(10.204)
V̇ = 21 ê T (A − BK T )T P1 ê + 21 ẽ T C K oT P1 ê+
+ 21 ê T P1 (A − BK T )ê + 21 ê T P1 K o C T ẽ+
+ 2 ẽ (A − K o C T )T P2 ẽ + 21 (u cT + wT + d̃ T )B T P2 ẽ+
1 T

+ 21 ẽ T P2 (A − K o C T )ẽ + 21 ẽ T P2 B(u c + w + d̃)+ (10.205)

+ 1 θ̃˙ T θ̃ + 1 tr [θ̃˙ θ̃ ]
T
γ1 f f γ2 g g

Assumption 1: For given positive definite matrices Q 1 and Q 2 there exist positive
definite matrices P1 and P2 , which are the solution of the following Riccati equations
[452]
10.4 Adaptive Control for Diesel Combustion Engines 539

(A − BK T )T P1 + P1 (A − BK T ) + Q 1 = 0 (10.206)
T
 ) P2 + P2 (A − K o C )−
(A − K o C T T
(10.207)
−P2 B r − ρ 2 B P2 + Q 2 = 0
2 1 T

The conditions given in Eqs. (10.206) and (10.207) are related to the requirement
that the systems described by Eqs. (10.169), (10.170), (10.186), and (10.187) are
asymptotically stable. Substituting Eqs. (10.206) and (10.207) into V̇ yields

V̇ = 21 ê T {(A − BK T )T P1 + P1 (A − BK T )}ê + ẽ T C K oT P1 ê+

+ 21 ẽ T {(A − K o C T )T P2 + P2 (A − K o C T )}ẽ+ (10.208)
+ẽ T P B(u + w + d̃) + 1 θ̃˙ T θ̃ + 1 tr [θ̃˙ θ̃ ]
T
2 c γ1 f f γ2 g g

that is
 
V̇ = − 21 ê T Q 1 ê + ẽ T C K oT P1 ê − 21 ẽ T {Q 2 − P2 B 2
r − 1
ρ2
B T P2 }ẽ+
(10.209)
1 ˙T ˙T
+ẽ T P2 B(u c + w + d̃) + γ1 θ̃ f θ̃ f + γ2 tr [θ̃ g θ̃g ]
1

The supervisory control u c is decomposed in two terms, u a and u b .

• The control term u a is given by

1
u a = − ẽ T P2 B + Δu a (10.210)
r

where assuming that the measurable elements of vector ẽ are {ẽ1 , e˜3 , . . . , e˜k }, the
term Δu a is such that
⎛ ⎞
p11 ẽ1 + p13 ẽ3 + · · · + p1k ẽk
⎜
1 ⎜ p13 ẽ1 + p33 ẽ3 + · · · + p3k ẽk ⎟
− r ẽ P2 B + Δu a = − r ⎝
1 T ⎟ (10.211)
··· ······ ⎠
p1k ẽ1 + p3k ẽ3 + · · · + pkk ẽk

• The control term u b is given by

u b = −[(P2 B)T (P2 B)]−1 (P2 B)T C K oT P1 ê (10.212)

• u a is an H∞ control used for the compensation of the approximation error w

and the additive disturbance d̃. Its first component − r1 ẽ T P2 B has been chosen so
as to compensate for the term r1 ẽ T P2 BBT P2 ẽ, which appears in Eq. (10.208). By
including also the second component Δu a , one has that u a is computed based on the
feedback only the measurable variables {ẽ1 , e˜3 , . . . , e˜k }, out of the complete vector
ẽ = [ẽ1 , ẽ2 , . . . , ẽn ]. Equation (10.210) is finally rewritten as u a = − r1 ẽ T P2 B +
Δu a .
540 10 Differential Flatness Theory for Internal Combustion Engines

• u b is a control used for the compensation of the observation error (the control term
u b has been chosen so as to satisfy the condition ẽ T P2 Bu b = −ẽ T C K oT P1 ê.
The adaptive control scheme for the diesel engine is depicted in Fig. 10.18.

Substituting Eqs. (10.210) and (10.212) in V̇ and assuming that Eqs. (10.206) and
(10.207) hold, one gets

V̇ = − 21 ê T Q 1 ê + ẽ T C K oT P1 ê − 21 ẽ T Q 2 ẽ+
+ r1 ẽ T P2 BBT P2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ+
+ẽ T P2 Bu a + ẽ T P2 Bu b + ẽ T P2 B(w + d̃)+ (10.213)

+ 1 θ̃˙ T θ̃ + 1 tr [θ̃˙ θ̃ ]
T
γ1 f f γ2 g g

or equivalently,

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ+
(10.214)
1 ˙T ˙T
+ẽ T P2 B(w + d̃ + Δu a ) + γ1 θ̃ f θ̃ f + γ2 tr [θ̃ g θ̃g ]
1

Fig. 10.18 The proposed H∞ control scheme

10.4 Adaptive Control for Diesel Combustion Engines 541

It holds that θ̃˙ f = θ̇ f − θ˙∗f = θ˙f and θ̃˙g = θ̇g − θ˙g∗ = θ˙g . The following weight
adaptation laws are considered:

θ̇ f = −γ1 Φ(x̂)T B T P2 ẽ
(10.215)
θ̇g = −γ2 Φ(x̂)T B T P2 ẽu T

where assuming N fuzzy rules and associated kernel functions the matrices dimen-
sions are θ f ∈R N ×1 , θg ∈R N ×2 , Φ(x)∈R 2×N , B∈R 4×2 , P∈R 4×4 , and ẽ∈R 4×1 .
The update of θ f is a gradient-type algorithm. The update of θg is also a gradient-
type algorithm, where u c implicitly tunes the adaptation gain γ2 [31, 414]. Substi-
tuting Eq. (10.215) in V̇ gives

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ + B T P2 ẽ(w + d + Δu a )+

+ γ11 (−γ1 )ẽ T P2 BΦ(x̂)(θ f − θ ∗f ) + γ12 (−γ2 )tr [u ẽ T P2 BΦ(x̂)(θg − θg∗ )]
(10.216)
or

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ + B T P2 ẽ(w + d̃ + Δu a )+

+ γ11 (−γ1 )ẽ T P2 BΦ(x̂)(θ f − θ ∗f ) + γ12 (−γ2 )tr [u ẽ T P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))]
(10.217)

Taking into account that u ∈ R 2×1 and ẽ T PB(ĝ(x|θg ) − ĝ(x|θg∗ )) ∈ R 1×2 it holds

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ + B T P2 ẽ(w + d̃ + Δu a )+

+ γ11 (−γ1 )ẽ T P2 BΦ(x̂)(θ f − θ ∗f ) + γ12 (−γ2 )tr [ẽ T P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))u]
(10.218)

Since ẽ T P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))u∈R 1×1 it holds

tr (ẽ T P2 B(ĝ(x|θg ) − ĝ(x|θg∗ )u)) =

(10.219)
= ẽ T P2 B(ĝ(x|θg ) − ĝ(x|θg∗ ))u

Therefore, one finally obtains

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ + B T P2 ẽ(w + d̃ + Δu a )+

+ γ11 (−γ1 )ẽ T P2 BΦ(x̂)(θ f − θ ∗f ) + γ12 (−γ2 )ẽ T P2 B(ĝ(x̂|θg ) − ĝ(x̂|θg∗ ))u
(10.220)

Next, the following approximation error is defined

wα = [ fˆ(x̂|θ ∗f ) − fˆ(x̂|θ f )] + [ĝ(x̂|θg∗ ) − ĝ(x̂|θg )]u (10.221)

542 10 Differential Flatness Theory for Internal Combustion Engines

Thus, one obtains

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ+
(10.222)
+B T P2 ẽ(w + d̃) + ẽ T P2 Bwα

Denoting the aggregate approximation error and disturbances vector as

w1 = w + d̃ + wα + Δu a (10.223)

the derivative of the Lyapunov function becomes

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 1 T
2ρ 2
ẽ P2 BBT P2 ẽ + ẽ T P2 Bw1 (10.224)

which in turn is written as

V̇ = − 21 ê T Q 1 ê − 21 ẽ T Q 2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ+

(10.225)
+ 21 ẽ T PBw1 + 21 w1T B T P2 ẽ

Lemma: The following inequality holds

1 T
2 ẽ P2 Bw1 + 21 w1T B T P2 ẽ − 2ρ1 2 ẽ T P2 BBT P2 ẽ
(10.226)
≤ 21 ρ 2 w1T w1

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ 2 a 2 + ρ12 b2 − 2ab ≥ 0 ⇒
2 ρ a + 2ρ 2 b − ab ≥ 0 ⇒
1 2 2 1 2
(10.227)
ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2 ⇒
2 ab + 2 ab − 2ρ 2 b ≤ 2 ρ a
1 1 1 2 1 2 2

The following substitutions are carried out: a = w1 and b = ẽ T P2 B and the previous
relation becomes
1 T
2 w1 B T P2 ẽ + 21 ẽ T P2 Bw1 − 2ρ1 2 ẽ T P2 BBT P2 ẽ
(10.228)
≤ 21 ρ 2 w1T w1

The above relation is used in V̇ , and the right part of the associated inequality is
enforced
1 1 1
V̇ ≤ − ê T Q 1 ê − ẽ T Q 2 ẽ + ρ 2 w1T w1 (10.229)
2 2 2
10.4 Adaptive Control for Diesel Combustion Engines 543

Thus, Eq. (10.229) can be written as

1 1
V̇ ≤ − E T QE + ρ 2 w1T w1 (10.230)
2 2
where
   
ê Q1 0
E= , Q= = diag[Q 1 , Q 2 ] (10.231)
ẽ 0 Q2

Hence, the H∞ performance criterion is derived. For ρ sufficiently small Eq. (10.229)
will be true and the H∞ tracking criterion will be satisfied. In that case, the integration
of V̇ from 0 to T gives
T T T
V̇ (t)dt ≤ − 21 0 ||E||2 dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
0 T T
2V (T ) − 2V (0) ≤ − 0 ||E||2Q dt + ρ 2 0 ||w1 ||2 dt ⇒ (10.232)
T T
2V (T ) + 0 ||E||2Q dt ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt
∞
It is assumed that there exists a positive constant Mw > 0 such that ||w1 ||2 dt ≤
T 0
Mw . Therefore for the integral 0 ||E||2Q dt one gets
 ∞
||E||2Q dt ≤ 2V (0) + ρ 2 Mw (10.233)
0
∞
Thus, the integral 0 ||E||2Q dt is bounded and according to Barbalat’s Lemma

limt→∞ E(t) = 0 ⇒
limt→∞ ê(t) = 0 (10.234)
limt→∞ ẽ(t) = 0

Therefore limt→∞ e(t) = 0.

10.4.5 Simulation Tests

The performance of the proposed observer-based adaptive fuzzy MIMO controller

was tested in the MIMO nonlinear model of the turbocharged diesel engine
(Fig. 10.14). The differentially flat model of the diesel engine and its transforma-
tion to the Brunovksy form has been analyzed in Sect. 10.3.5.
The state feedback gain was K ∈R 2×4 . The basis functions used in the estimation
x̂−c j 2
of f i (x̂, t), i = 1, 2 and gi j (x̂, t), i = 1, 2, j = 1, 2 were μ A j (x̂) = e( σ ) , j =
1, . . . , 3. Since there are four inputs x̂1 , x̂˙1 and x̂3 , x̂˙3 and each one of them consists
544 10 Differential Flatness Theory for Internal Combustion Engines

of 3 fuzzy sets, for the approximation of functions f i (x̂, t) i = 1, 2, there will be 81

fuzzy rules of the form:

R l : IF x̂1 is Al1 AND x̂˙1 is Al2

(10.235)
AND x̂3 is Al3 AND x̂˙3 is Al4 THEN fˆil is bl

The aggregate output of the neuro-fuzzy approximator (rule base) is

81 
ˆl 4 μl (x̂i )
l=1 f i
fˆi (x̂, t) =
i=1 Ai
81 4 (10.236)
l=1 i=1 μ A ( x̂i )
l
i

The estimation of the control input gain functions ĝi j (x̂, t) i = 1, 2 was derived
in a similar way. The sampling period was taken to be 0.01 s. In the beginning of
the training of the neuro-fuzzy approximators their weights were initialized to zero.
Moreover, the elements of the diesel engine’s state vector were also initialized to
zero. The positive definite matrices P1 ∈R 4×4 and P2 ∈R 4×4 stem from the solution
of the algebraic Riccati equations (10.206) and (10.207) for Q 1 and Q 2 are also
positive definite.
The approximations fˆ and ĝ were used in the derivation of the control law, given
by Eq. (10.164). To show the disturbance rejection capability of the proposed adaptive
fuzzy controller at the beginning of the second half of the simulation time additive
sinusoidal disturbances of amplitude A = 0.5 and period T = 10 s were applied to
the diesel engine model.
The performance of the differential flatness theory-based adaptive fuzzy control
loop was tested in the case of tracking of several reference setpoints. The obtained
results are depicted in Figs. 10.19, 10.20, 10.21, 10.22, and 10.23. It can be observed

(a) (b)
state variable p1

8 100 10
state variable z1

state variable z2

6 5
50

4
0
0 0 5 10 15 20 25 30 35 40
2 time
state variable p2

100
0 −50
0 10 20 30 40 0 10 20 30 40
0
t (sec) t (sec)
8 150 −100
0 5 10 15 20 25 30 35 40
state variable z4
state variable z3

6 100 time
c
state variable P

40
4 50
20
2 0

0
0 −50 0 5 10 15 20 25 30 35 40
0 10 20 30 40 0 10 20 30 40 time
t (sec) t (sec)

Fig. 10.19 a Tracking of a reference setpoint 1 by the state variables z i , i = 1, . . . , 4 of the

transformed diesel engine model. b Tracking of a reference setpoint 1 by the state variables xi , i =
1, . . . , 3 of the initial nonlinear diesel engine model
10.4 Adaptive Control for Diesel Combustion Engines 545

(a) (b)

state variable p1
8 100 10
state variable z1

state variable z2
6 5
50

4
0
0 5 10 15 20 25 30 35 40
0
2 time

state variable p2
500
0 −50
0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) 0

6 150 −500
0 5 10 15 20 25 30 35 40
state variable z3

state variable z4

100 time

state variable Pc
4 40
50

2 20
0
0
0 −50 0 5 10 15 20 25 30 35 40
0 10 20 30 40 0 10 20 30 40 time
t (sec) t (sec)

Fig. 10.20 a Tracking of a reference setpoint 2 by the state variables z i , i = 1, . . . , 4 of the

transformed diesel engine model. b Tracking of a reference setpoint 2 by the state variables xi , i =
1, . . . , 3 of the initial nonlinear diesel engine model

(a) (b)
1
state variable p

8 100 10
state variable z1

state variable z2

6 50 5

4 0
0
0 5 10 15 20 25 30 35 40
2 −50 time
2
state variable p

10
0 −100
0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) 0

8 100 −10
0 5 10 15 20 25 30 35 40
state variable z3

state variable z4

6 50 time
c
state variable P

4
4 0
2
2 −50

0
0 −100 0 5 10 15 20 25 30 35 40
0 10 20 30 40 0 10 20 30 40 time
t (sec) t (sec)

Fig. 10.21 a Tracking of a reference setpoint 3 by the state variables z i , i = 1, . . . , 4 of the

transformed diesel engine model. b Tracking of a reference setpoint 3 by the state variables xi , i =
1, . . . , 3 of the initial nonlinear diesel engine model

that the proposed adaptive fuzzy control scheme succeeded fast and accurate tracking
of all these setpoints.
The RMSE (root mean square error) of the examined control loop is also calcu-
lated (assuming the same parameters of the controller) in the case of tracking of the
previous setpoints 1–5. The results are summarized in Table 10.2. It can be seen that
the transient characteristics of the control scheme are also quite satisfactory.
546 10 Differential Flatness Theory for Internal Combustion Engines

(a) (b)

state variable p1
10 60 10
state variable z1

state variable z2
8
40 5
6
20
0
4 0 5 10 15 20 25 30 35 40
0 time
2

state variable p2
10
0 −20
0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) 0

8 100 −10
0 5 10 15 20 25 30 35 40
state variable z3

state variable z4

6 time

state variable Pc
50 4
4

0 2
2
0
0 −50 0 5 10 15 20 25 30 35 40
0 10 20 30 40 0 10 20 30 40 time
t (sec) t (sec)

Fig. 10.22 a Tracking of a reference setpoint 4 by the state variables z i , i = 1, . . . , 4 of the

transformed diesel engine model. b Tracking of a reference setpoint 4 by the state variables xi , i =
1, . . . , 3 of the initial nonlinear diesel engine model

(a) (b)
state variable p 1

10 60 10
state variable z1

state variable z2

8
40 5
6
20
0
4 0 5 10 15 20 25 30 35 40
0 time
2
state variable p 2

10
0 −20
0 10 20 30 40 0 10 20 30 40
t (sec) t (sec) 0

8 150 −10
0 5 10 15 20 25 30 35 40
state variable z3

state variable z4

6 100 time
state variable Pc

5
4 50

2 0
0
0 −50 0 5 10 15 20 25 30 35 40
0 10 20 30 40 0 10 20 30 40 time
t (sec) t (sec)

Fig. 10.23 a Tracking of a reference setpoint 5 by the state variables z i , i = 1, . . . , 4 of the

transformed diesel engine model. b Tracking of a reference setpoint 5 by the state variables xi , i =
1, . . . , 3 of the initial nonlinear diesel engine model
10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 547

Table 10.2 RMSE of diesel engine’s state variables

parameter p1 p2 Pc
RMSEa 0.0001 0.0004 0.0002
RMSEb 0.0202 0.0204 0.0055
RMSEc 0.0079 0.0411 0.0087
RMSEd 0.0001 0.0009 0.0005
RMSEe 0.0001 0.0215 0.0128

10.5 Flatness-Based Control and Kalman Filtering

for the Spark-Ignited Engine

10.5.1 Overview

In this section it is shown that the developed methods of differential flatness theory-
based control and filtering can be also applied to the model of the spark-ignited
engine. The problem of control of the rotation speed of SI engines, as well as the
problem of control of the engine’s manifolds pressure, has been approached with
different methods [505, 598]. In [8] a nonlinear state-space controller for turn speed
(and consequently for torque) of a spark-ignited engine is proposed. The controller’s
design is based on feedback linearization in combination with pole placement. In
[599] time-varying internal model-based design is applied to compensate for the time-
varying (but angle-dependent) pressure pulsations in the fuel injection system of SI
engines. In [281] a control method for the air-path system of SI engines is presented.
The first part considers generation of the motion planning trajectory of the intake
manifold pressure from a torque setpoint. Then, feedforward and feedback control
laws are presented. In [357] a feedback linearization approach based on differential
flatness theory is proposed for the control of the air system of a turbocharged gasoline
engine. Finally, in [149] a model-based approach is pursued to maximize an SI
engine’s torque through optimal control of the variable valve timing (VVT) and
the variable gas turbine (VGT) model. Several other results on nonlinear control of
spark-ignited engines have been presented in [27, 57, 106, 374, 481].
In this section, a new nonlinear control and filtering scheme is proposed for SI
engines making use of differential flatness theory. By showing that the SI engine
model is a differentially flat one it becomes possible to transform it to the linear
canonical form. For the latter description of the system’s dynamics the design of a
state feedback controller becomes easier. Considering also that specific elements of
the engine’s state vector cannot be directly measured, it is proposed to estimate them
with the use of the previously analyzed Derivative-free nonlinear Kalman Filter. As
shown in the previous sections, this filter consists of the Kalman Filter recursion on
the linearized equivalent of the SI engine and on the use of an inverse transformation
(diffeomorphism) that is based on differential flatness theory which provides the
548 10 Differential Flatness Theory for Internal Combustion Engines

estimates of the state vector elements of the initial nonlinear system. By designing
the nonlinear Kalman Filter as a disturbance observer, it is also possible to estimate
in real-time disturbance terms affecting the SI engine’s model and subsequently to
compensate for them.

10.5.2 Dynamic Model of the SI Engine

10.5.2.1 State-Space Description of the SI Engine

It is possible to control the intake pressure pm and the rotational speed of the engine’s
shaft ω by adjusting the angle of the air throttle. It is considered that the associated
control loop is independent from the loops of the fuel injection control and spark
timing control (Fig. 10.24).
The basic equations of the system are:

ω̇ = kω1 pm (t − τd ) + kω2 + kω3 T fm

ṗm = k p1 ω pm + k p2 ω + k p3 u (10.237)
y1 = ω

Fig. 10.24 Diagram of the spark-ignited engine

10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 549

The variable of the intake pressure appears with time delay in the equation of the turn
speed in the second row of the model of the SI engine. Using that pmd = pm (t − τd )
and

pm (t − τd ) = 1
τ s+1 pm (10.238)

while τ = ad /ω and ad is a parameter that is measured in radians. Denoting kd =

−1/ad one has about the dynamics of the delayed intake pressure variable

pmd = kd ω( pm d − pm ) (10.239)

Using the previous formulation, and defining the state variables x1 = ω, x2 = pm d

and x3 = pm , the dynamics of the SI engine is written as [505]

ẋ1 = kω1 x2 + kω2 + kω3 T fm

ẋ2 = K d x1 (x2 − x3 ) (10.240)
ẋ3 = k p1 x1 x3 + k p2 x1 + k p3 u

where T fm are friction torques, which can be also perceived as disturbances. In the
above equations coefficients k pi , i = 1, 2, 3, kωi , i = 1, 2, 3, and K d are associated
with the combustion cycle of the SI engine and are defined in [505, 598]. The model
also takes the matrix form ẋ = f (x) + g(x)u with
⎛ ⎞ ⎛ ⎞
kω1 x2 + kω2 + kω3 T fm 0
f (x) = ⎝ kd x1 (x2 − x3 ) ⎠ g(x) = ⎝ 0 ⎠ (10.241)
k p1 x 1 x 3 + k p2 x 1 k p3

10.5.3 Feedback Linearizing Control of the SI Engine Using

Lie Algebra

First, it will be shown that the dynamic model of the SI engine can be linearized using
Lie algebra. Using Lie derivatives, the following state variables are defined for the SI
engine model of Eq. (10.240): z 1 = h 1 (x) = x1 , z 2 = L f h 1 (x) and z 3 = L 2f h 1 (x).
It holds that

z 2 = L f h 1 (x) = ∂h ∂h 1 ∂h 1
∂ x1 f 1 + ∂ x2 f 2 + ∂ x3 f 3 ⇒
1
(10.242)
z 2 = f 1 ⇒z 2 = kω1 x2 + kω2 + kω3 T fm

In a similar manner one obtains

z 3 = L 2f h 1 (x) = ∂∂zx21 f 1 + ∂∂zx22 f 2 + ∂∂zx23 f 3 ⇒

z 3 = f 1 ⇒z 2 = kω1 x2 + kω2 + kω3 T fm ⇒ (10.243)
z 3 = kω1 f 2 ⇒z 3 = kω1 kd x1 (x2 − x3 )
550 10 Differential Flatness Theory for Internal Combustion Engines

Moreover, it holds that

L 3f h 1 (x) = ∂∂zx31 f 1 + ∂∂zx32 f 2 + ∂∂zx33 f 3 ⇒

L 3f h 1 (x) = kω1 kd (x2 − x3 ) f 1 + kω1 kd x1 f 2 − kω1 kd x1 f 3 ⇒
(10.244)
L 3f h 1 (x) = kω1 kd (x2 − x3 )[kω1 x2 + kω2 + kω3 T fm ]+
+kω1 kd x1 [kd x1 (x2 − x3 )] − kω1 kd x1 [k p1 x1 x3 + k p2 x1 ]

Additionally, it holds that

L g L f h 1 (x) = L g z 2 ⇒L g L f h 1 (x) = ∂∂zx21 g1 + ∂∂zx22 g2 + ∂z 2

∂ x3 g3 ⇒
L g L f h 1 (x) = ∂∂zx23 k p3 ⇒L g L f h 1 (x) = ∂∂ xf32 k p3 (10.245)
⇒L g L f h 1 (x) = 0

while one also obtains

L g L 2f h 1 (x) = L g z 3 ⇒L g L 2f h 1 (x) = ∂∂zx31 g1 + ∂z 3

∂ x2 g2 + ∂z 3
∂ x3 g3 ⇒ (10.246)
L g L 2f h 1 (x) = −kω1 kd k p1 x1

which also shows that the relative degree of the SI engine model is n = 3. It can be
also confirmed that it holds

ż 1 = z 2
ż 2 = z 3 (10.247)
ż 3 = L 3f h 1 (x) + L g L 2f h 1 (x)u

which after defining the new control input v = L 3f h 1 (x) + L g L 2f h 1 (x)u can be
written in the linear canonical (Brunovsky) form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (10.248)
ż 3 000 z3 1

(3)
From the relation z 1 = v the state feedback control law for the SI engine that assures
asymptotic convergence of the state vector z to the desirable setpoint z d is given by
(3)
v = z 1d − k1 (z̈ 1 − z̈ 1,d ) − k2 (ż 1 − ż 1,d ) − k3 (z 1 − z 1,d ) (10.249)

Using that the control input for the linearized model is v = L 3f h 1 (x) + L g L 2f h 1 (x)u
the control input that is finally applied to the SI engine is

u= 1
L g L 2f h 1 (x)
[v − L 3f h 1 (x)] (10.250)
10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 551

10.5.4 Feedback Linearizing Control of the SI Engine Using

Differential Flatness Theory

Next, it will be shown that the dynamic model of the SI engine can be linearized using
differential flatness theory. The state-space description of the SI engine dynamics
given in Eq. (10.240) is considered again

ẋ1 = kω1 x2 + kω2 + kω3 T fm (10.251)

ẋ2 = K d x1 (x2 − x3 ) (10.252)

ẋ3 = k p1 x1 x3 + k p2 x1 + k p3 u (10.253)

The flat output of the SI engine model is taken to be y = x1 , which is the engine’s
turn speed. It will be shown that all state variables of the system and the control
input can be written as functions of the flat output and its derivatives, and thus the
SI engine model is a differentially flat one.
Equation (10.251) is solved with respect to x2 . Considering that the friction torque
T fm is constant or piecewise constant this gives

ẋ1 −kω2 −kω3 T f m ẏ−kω2 −kω3 T f m

x2 = kω1 ⇒x2 = kω1 ⇒
(10.254)
x2 = f 2 (y, ẏ)

It is also noted that the friction torque can be considered as a disturbance term to the
engine’s model and this approach will be followed in Sect. 10.5.5. Next, Eq. (10.252)
is solved with respect to x3 . This gives

˙
x3 = kd x1 x2 −ẋ2
kd x 1 ⇒x3 = kd y f2 (y,kẏ)−
dy
f 2 (y, ẏ)
⇒
(10.255)
x3 = f 3 (y, ẏ)

Moreover, from Eq. (10.253) it holds

ẋ3 −k p1 x1 x3 −k p2 x1 f˙3 (y, ẏ)−k p1 y f 3 (y, ẏ)−k p2 y

u= k p3 ⇒u = k p3
(10.256)

Therefore, all state variables and the control input in the model of the SI engine are
described as functions of the flat output and its derivatives. Consequently, the SI
engine model is a differentially flat one. It holds that

y = x1
(10.257)
ẏ = ẋ1 ⇒ ẏ = kω1 x2 + kω2 + kω3 T fm
552 10 Differential Flatness Theory for Internal Combustion Engines

Differentiating once more with respect to time gives

ÿ = kω1 ẋ2 + kω3 Ṫ fm ⇒

(10.258)
ÿ = kω1 kd x1 (x2 − x3 ) + kω3 Ṫ fm

By deriving once more with respect to time one gets

y (3) = kω1 kd ẋ1 (x2 − x3 ) + kω1 kd x1 ẋ2 − kω1 kd x1 ẋ3 + kω3 T̈ fm ⇒

y (3) = kω1 kd (x2 − x3 )[kω1 x2 + kω2 + kω3 T fm ]+
(10.259)
+kω1 kd x1 [kd x1 (x2 − x3 )] − kω1 kd x1 [k p1 x1 x3 + k p2 x1 ]+
−kω1 kd k p3 x1 u + kω3 T̈ fm

Therefore, one arrives at a description of the SI engine dynamics which is equivalent

to the one obtained from linearization with the use of Lie algebra

y (3) = L 3f h 1 (x) + L g L 2f h 1 (x)u (10.260)

where

L 3f h 1 (x) = kω1 kd (x2 − x3 )[kω1 x2 + kω2 + kω3 T fm ]+

(10.261)
+kω1 kd x1 [kd x1 (x2 − x3 )] − kω1 kd x1 [k p1 x1 x3 + k p2 x1 ] + kω3 T̈ fm

and also

L g L 2f h 1 (x) = −kω1 kd k p3 x1 (10.262)

For the previous description of the SI engine dynamics, the following state variables
are defined z 1 = y, z 2 = ẏ and z 3 = ÿ, the following state-space model is obtained

ż 1 = z 2
ż 2 = z 3 (10.263)
ż 3 = L 2f h 1 (x) + L g L 2f h 1 (x)u

The linearized model of the SI engine is finally written in the Brunovsky canonical
form
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (10.264)
ż 3 000 z3 1

From the relation z 1(3) = v the state feedback control law for the SI engine that assures
asymptotic convergence of the state vector z to the desirable setpoint z d is given by
(3)
v = z 1d − k1 (z̈ 1 − z̈ 1,d ) − k2 (ż 1 − ż 1,d ) − k3 (z 1 − z 1,d ) (10.265)
10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 553

Using that the control input for the linearized model is v = L 3f h 1 (x) + L g L 2f h 1 (x)u
the control input that is finally applied to the SI engine is

u= 1
L g L 2f h 1 (x)
[v − L 3f h 1 (x)] (10.266)

10.5.5 Compensation of Disturbances Using

the Derivative-Free Nonlinear Kalman Filter

It was shown that following the linearization approach based on differential flatness
theory, the model of the SI engine is written in the canonical form

y (3) = L 3f h 1 (x) + L g L 2f h 1 (x)u + d̃ (10.267)

The additive disturbance term is given by d̃ = kω1 kd (x2 − x3 )kω3 T fm + kω3 T̈ fm and
describes the effects of friction torque. This term can also incorporate the effects
of model parametric uncertainty. It can be considered that the dynamics of the dis-
turbance term is described by its nth order derivative and the associated initial con-
ditions. Without loss of generality, by setting n = 3 one has d̃ (3) = f d . Next, as
new state variables of the SI engine model, the disturbance d̃ and its derivatives are
˙ and z = d̃.
introduced. That is, z 1 = y, z 2 = ẏ, z 3 = ÿ, z 4 = d̃, z 5 = d̃, ¨ Thus one
6
arrives at the extended state-space model

ż = Az + B ṽ
(10.268)
zm = C z

where the control input ṽ = [v, f d ]T and matrices A, B and C are defined as
⎛ ⎞ ⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 00 1
⎜0 0 1 0 0 0⎟ ⎜0 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0⎟ ⎜ ⎟ ⎜ ⎟
A=⎜ ⎟ B = ⎜1 0⎟ C T = ⎜0⎟ (10.269)
⎜0 0 0 0 1 0⎟⎟ ⎜ ⎟ ⎜0⎟
⎜ ⎜0 0 ⎟ ⎜ ⎟
⎝0 0 0 0 0 1 ⎠ ⎝ 00 ⎠ ⎝0⎠
0 0 0 0 0 0 01 0

The associated state estimator is

ẑ˙ = Ao ẑ + Bo ṽ + K f (z meas − Co ẑ) (10.270)

where Ao = A, Co = C and matrix Bo is given by

 
001000
BoT = (10.271)
000000
554 10 Differential Flatness Theory for Internal Combustion Engines

The estimation of the extended state vector is performed using the discrete-time
Kalman Filter. To this end, matrices Ao , Bo , Co are discretized into Ad , Bd , C D
using common discretization methods. Then, the recursion of the Kalman Filter
consists of the measurement update and time update stage [414, 434, 445]:
measurement update:

K (k) = P − (k)CdT [Cd P − (k)CdT + R]−1

ẑ(k) = x̂ − (k) + K (k)[z m (k) − ẑ m (k)] (10.272)
P(k) = P − (k) − K (k)Cd P − (k)

time update:

P − (k + 1) = Ad P(k)AdT + Q
(10.273)
ẑ(k + 1) = Ad ẑ(k) + Bd Ṽ (k)

After the estimation of the state vector ẑ the state feedback control law for the SI
engine given in Eq. (10.265) is modified with the inclusion of the estimated state
variable ẑ 4 . The term ẑ 4 will compensate for the disturbance input d̃
(3)
v = z 1,d − k1 (z̈ 1 − z̈ 1,d ) − k2 (ż 1 − ż 1,d ) − k3 (z 1 − z 1,d ) − ẑ 4 (10.274)

Thus, it is assured that the tracking error for the state vector z will converge asymp-
totically to 0. That is,

limt→∞ e(t) = 0⇒limt→∞ z = z d (10.275)

which also means that limt→∞ xi = xi,d , i = 1, 2, 3. It is also noted that through the
previously analyzed filtering procedure one can obtain estimates of the input pressure
of the SI engine x̂3 = p̂m using Eqs. (10.255) and (10.256) and by substituting in
there the estimates of the flat output ŷ and of its derivatives.

˙ ω +kω T̂ f
ŷ−k
x̂2 = 2
kω1
3 m (10.276)

kd ŷ x̂2 −x̂˙2
x̂3 = kd ŷ
(10.277)

10.5.6 Simulation Tests

Simulation tests have been carried out for the assessment of the performance of
the proposed nonlinear control scheme based on differential flatness theory and on
the Derivative-free nonlinear Kalman Filter. Indicative results were obtained for the
tracking of two different setpoints. It can be observed that the SI engine’s state vector
10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 555

elements (that is engine’s turn speed ω, input pressure subjected to time delay Pm d ,
and input pressure Pm ) converged fast and smoothly to the associated setpoints.
Moreover, the Derivative-free nonlinear Kalman Filter provided accurate estimates
of the aggregate disturbance term Tm which is due to the friction torque T fm and to
its derivatives. This enabled to compensate for the disturbances’ effects by including
in the control input a new term that was based on the disturbances’ estimates. The
associated results in the case of tracking of setpoint 1 are depicted in Figs. 10.25,

(a) 2 (b)10
1.8 9.5

1.6 9

1.4 8.5

1.2 8
ω × 10 3

d
Pm
1 7.5

0.8 7

0.6 6.5

0.4 6

0.2 5.5

0 5
0 5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
time time

Fig. 10.25 Tracking of setpoint 1. a Convergence of the SI engine’s state variable x1 = ω (blue
line) to reference setpoint (red line) and associated state estimate x̂1 (green line). b Convergence of
the SI engine’s state variable x2 = Pm d (blue line) to reference setpoint (red line) and associated
state estimate x̂2 (green line)

(a) 10
(b) 15

9.5
10
9
Estimation of input torque T m

8.5
5
8
Pm

7.5 0

7
−5
6.5

6
−10
5.5

5 −15
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 10.26 Tracking of setpoint 1. a Convergence of the SI engine’s state variable x3 = Pm (blue
line) to reference setpoint (red line) and associate state estimate x̂3 (green line). b Kalman Filter-
based estimation (green line) of the aggregate disturbance torque Tm (blue line)
556 10 Differential Flatness Theory for Internal Combustion Engines

(a) (b)

Tm1 − Tm1−est
3 20
ω × 10 3

2
0
1

0 −20
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Tm2 − Tm2−est
10 5

8
d
Pm

0
6
−5
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Tm3 − Tm3−est
10 1
Pm

8
0
6
−1
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 10.27 Tracking of setpoint 1. a Tracking of reference setpoints by the SI engine’s state variables
x1 = ω, x2 = Pm d , and x3 = Pm . b Kalman Filter-based estimation of disturbance input torque Tm
and of its derivatives

(a) 3 (b) 10
9.5
2.5
9

8.5
2
8
3
ω × 10

Pmd

1.5 7.5

7
1
6.5

6
0.5
5.5

0 5
0 5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
time time

Fig. 10.28 Tracking of setpoint 2. a Convergence of the SI engine’s state variable x1 = ω (blue
line) to reference setpoint (red line) and associated state estimate x̂1 (green line). b Convergence of
the SI engine’s state variable x2 = Pm d (blue line) to reference setpoint (red line) and associated
state estimate x̂2 (green line)

10.26, and 10.27. Additional results for the case of tracking of setpoint 2 are shown
in Figs. 10.28, 10.29, and 10.30.
Finally, the variations of the control input to the SI engine (air throttle angle) for
the cases of the two aforementioned setpoints are given in Fig. 10.31. It has been
confirmed that with the application of the proposed nonlinear control scheme the
control input exhibited smooth variations.
10.5 Flatness-Based Control and Kalman Filtering for the Spark-Ignited Engine 557

(a) 10 (b) 20

9.5
15
9

Estimation of input torque Tm

8.5 10

8
5
Pm

7.5
0
7

6.5 −5

6
−10
5.5

5 −15
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 10.29 Tracking of setpoint 2. a Convergence of the SI engine’s state variable x3 = Pm (blue
line) to reference setpoint (red line) and associated state estimate x̂3 (green line). b Kalman Filter-
based estimation (green line) of the aggregate disturbance torque Tm (blue line)

(a) (b)
Tm1 − Tm1−est

3 20
ω × 10 3

2
0
1

0 −20
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time
Tm2 − Tm2−est

10 10

8
d
Pm

0
6
−10
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time
Tm3 − Tm3−est

10 2
Pm

8
0
6
−2
5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
time time

Fig. 10.30 Tracking of setpoint 2. a Tracking of reference setpoints by the SI engine’s state variables
x1 = ω, x2 = Pm d , and x3 = Pm . b Kalman Filter-based estimation of disturbance input torque Tm
and of its derivatives
558 10 Differential Flatness Theory for Internal Combustion Engines

(a) 20 (b) 20

15 15

10 10

5 5
u

u
0 0

−5 −5

−10 −10

−15 −15

−20 −20
5 10 15 20 25 30 35 40 5 10 15 20 25 30 35 40
time time

Fig. 10.31 Variations of the control input of the SI engine (air throttle angle) in case of a tracking
of reference setpoint 1, b tracking of reference setpoint 2

10.6 Flatness-Based Adaptive Fuzzy Control

of the Spark-Ignited Engine

10.6.1 Overview

In this section a new nonlinear adaptive fuzzy control scheme based on differential
flatness theory is proposed for SI engines. The new results come to extend the method
presented in [407, 410, 413, 433, 454]. By showing that the SI engine model is a
differentially flat one it becomes possible to transform it to the linear canonical form.
For the latter description of the system’s dynamics the design of a state feedback
controller becomes easier. After transformation to the linear canonical form, the
resulting control input for the engine is shown to contain nonlinear elements which
depend on the system’s parameters. If the parameters of the system are unknown,
then the nonlinear terms which appear in the control signal can be approximated with
the use of neuro-fuzzy networks [407, 414]. In this section it is shown that a suitable
learning law can be defined for the aforementioned neuro-fuzzy approximators so
as to preserve the closed-loop system stability. Lyapunov stability analysis proves
also that the proposed flatness-based adaptive fuzzy control scheme results in H∞
tracking performance.
Indirect adaptive fuzzy control based on differential flatness theory can be applied
to various types of combustion engines. This is important for the design of controllers,
capable of efficiently compensating for modeling uncertainties and external distur-
bances in such a type of nonlinear dynamical systems. As explained in previous
sections, unlike other adaptive control schemes which are based on several assump-
tions about the structure of the nonlinear system as well as about the uncertainty
10.6 Flatness-Based Adaptive Fuzzy Control of the Spark-Ignited Engine 559

characterizing the system’s model, the proposed adaptive fuzzy control scheme based
on differential flatness theory offers an exact solution to the design of the controller
for unknown dynamical systems. The only assumption needed for the design of the
controller and for succeeding H∞ tracking performance for the control loop is that
there exists a solution for a Riccati equation associated to the linearized error dynam-
ics of the differentially flat model. This assumption is quite reasonable for various
combustion engines models, thus providing a systematic approach to the design of
reliable controllers for such systems [426, 433].

10.6.2 Flatness-Based Adaptive Fuzzy Control for SI Motors

10.6.2.1 Nonlinear System Transformation into the Brunovsky Form

The adaptive fuzzy control approach to be developed for the spark-ignited engine is
similar to the one explained in Chap. 3. The model of the spark-ignited engine is a
single-input differentially flat dynamical system of the form:

ẋ = f s (x, t) + gs (x, t)(u + d̃), x∈R n , u∈R, d̃∈R (10.278)

where f s (x, t), gs (x, t) are nonlinear vector fields defining the system’s dynamics,
u denotes the control input and d̃ denotes additive input disturbances. Knowing that
the system of Eq. (10.278) is differentially flat, the next step is to try to write it
into a Brunovsky form. It has been shown that, in general, transformation into the
Brunovsky (canonical) form can be succeeded for systems that admit static feed-
back linearization [340]. Single-input differentially flat systems admit static feed-
back linearization, therefore they can be transformed into the Brunovksy form. For
multi-input differentially flat systems there may also exist a transformation into the
Brunovsky form [426].
The selected flat output is denoted by y. For the state variables xi of the system
of Eq. (10.278) it holds

xi = φi (y, ẏ, . . . , y (r −1) ), i = 1, . . . , n (10.279)

while for the control input it holds

u = ψ(y, ẏ, . . . , y (r −1) , y (r ) ) (10.280)

Introducing the new state variables y1 = y and yi = y (i−1) , i = 2, . . . , n, the initial

system of Eq. (10.278) can be written in the Brunovsky form:
560 10 Differential Flatness Theory for Internal Combustion Engines
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 0 1 0 ··· 0 y1 0
⎜ ẏ2 ⎟ ⎜ 0 0 1 ··· 0⎟ ⎜ y2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ · · · ⎟ ⎜· · · ··· ··· ··· · · ·⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ · · · ⎟ + ⎜· · ·⎟ v
⎜ · · · ⎟ = ⎜· · · ··· ··· ··· ⎟ ⎜
· · ·⎟ ⎜ · · · ⎟ ⎜ ⎟ (10.281)
⎜ ⎟ ⎜ ⎟ ⎜· · ·⎟
⎝ ẏn−1 ⎠ ⎝ 0 0 0 ··· 1 ⎠ ⎝ yn−1 ⎠ ⎝0⎠
ẏn 0 0 0 ··· 0 yn 1

where v = f (x, t) + g(x, t)(u + d̃) is the control input for the linearized model,
and d̃ denotes additive input disturbances. Thus one can use that the spark-ignited
engine model is finally written in the form:

y (n) = f (x, t) + g(x, t)(u + d̃) (10.282)

where f (x, t), g(x, t) are unknown nonlinear functions, while as mentioned above d̃
is an unknown additive disturbance. It is possible to make the system’s state vector x
follow a given bounded reference trajectory xd . In the presence of model uncertain-
ties and external disturbances, denoted by wd , successful tracking of the reference
trajectory is provided by the H∞ criterion [413, 524]:
T T
0 e T Qedt ≤ ρ 2 0 wd T wd dt (10.283)

where ρ is the attenuation level and corresponds to the maximum singular value of
the transfer function G(s) of the linearized model associated to Eqs. (10.281) and
(10.282).

10.6.2.2 Control Law for the SI Engine

For the measurable state vector x of the SI engine model in the form of Eqs. (10.281)
and (10.282), and for uncertain functions f (x, t) and g(x, t) an appropriate control
law is
1 (n)
u= [y − fˆ(x, t) − K T e + u c ] (10.284)
ĝ(x, t) d

with e = [e, ė, ë, . . . , e(n−1) ]T and e = y − yd , K T = [kn , kn−1 , . . . , k1 ], such that
the polynomial e(n) + k1 e(n−1) + k2 e(n−2) + · · · + kn e is Hurwitz. The control law
of Eq. (10.284) results into

e(n) = −K T e + u c + [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + g(x, t)d̃

(10.285)

where the supervisory control term u c aims at the compensation of the approximation
error

w = [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u (10.286)

10.6 Flatness-Based Adaptive Fuzzy Control of the Spark-Ignited Engine 561

as well as of the additive disturbance term d1 = g(x, t)d̃. The above relation
can be written in a state-equation form. The state vector is taken to be e T =
[e, ė, . . . , e(n−1) ], which after some operations yields

ė = (A − BK T )e + Bu c + B{[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d1 }

(10.287)
where
⎛ ⎞ ⎛ ⎞
0 1 0 ··· ··· 0 0
⎜ 0 0 1 ··· ··· 0 ⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟
⎜· · · · · · · · · · · · · · · · · ·⎟ ⎜· · ·⎟
A=⎜ ⎜ ⎟ ⎜ ⎟
· · · · · · · · · · · · · · · · · · ⎟ , B = ⎜· · ·⎟ (10.288)
⎜ ⎟ ⎜ ⎟
⎝ 0 0 0 ··· ··· 1 ⎠ ⎝0⎠
0 0 0 ··· ··· 0 1

and K = [kn , kn−1 , . . . , k2 , k1 ]T . As explained above, the control signal u c is an

auxiliary control term, used for the compensation of d̃ and w, which can be selected
according to H∞ control theory:

u c = − r1 B T Pe (10.289)

10.6.2.3 Approximators of Unknown System Dynamics

The approximation of functions f (x, t) and g(x, t) of Eq. (10.282) can be carried
out with neuro-fuzzy networks (Fig. 3.1). The estimation of f (x, t) and g(x, t) can
be written as [544]:

fˆ(x|θ f ) = θ Tf φ(x), ĝ(x|θg ) = θgT φ(x), (10.290)

where φ(x) are kernel functions with elements

n
i=1 μ A (xi )
l
φ l (x) =  N n i l l = 1, 2, . . . , N (10.291)
l=1 i=1 μ A (xi )
i

It is assumed that the weights θ f and θg vary in the bounded areas Mθ f and Mθg
which are defined as

Mθ f = {θ f ∈ R h : ||θ f || ≤ m θ f },
(10.292)
Mθg = {θg ∈ R h : ||θg || ≤ m θg }

with m θ f and m θg positive constants.

562 10 Differential Flatness Theory for Internal Combustion Engines

The values of θ f and θg that give optimal approximation are:

θ ∗f = arg minθ f ∈Mθ f [supx∈Ux | f (x) − fˆ(x|θ f )|]

(10.293)
θg∗ = arg minθg ∈Mθg [supx∈Ux |g(x) − ĝ(x|θg )|]

The approximation error of f (x, t) and g(x, t) is given by

w = [ fˆ(x, |θ ∗f ) − f (x, t)] + [ĝ(x|θ ∗f ) − g(x, t)]u ⇒

(10.294)
w = {[ fˆ(x|θ ∗f ) − fˆ(x|θ f )] + [ fˆ(x|θ f ) − f (x, t)]}+
+{[ĝ(x|θg∗ ) − ĝ(x|θg )] + [ĝ(x|θg ) − g(x, t)]u}

where: (i) fˆ(x|θ ∗f ) is the approximation of f for the best estimation θ ∗f of the weights’
vector θ f , (ii) ĝ(x|θg∗ ) is the approximation of g for the best estimation θg∗ of the
weights’ vector θg .
The approximation error w can be decomposed into wa and wb , where

wa = [ fˆ(x|θ f ) − fˆ(x|θ ∗f )] + [ĝ(x|θg ) − ĝ(x|θg∗ )]u

(10.295)
wb = [ fˆ(x|θ ∗f ) − f (x, t)] + [ĝ(x|θg∗ ) − g(x, t)]u

Finally, the following two parameters are defined:

θ̃ f = θ f − θ ∗f
(10.296)
θ̃g = θg − θg∗ .

10.6.3 Lyapunov Stability Analysis

Lyapunov stability analysis for adaptive fuzzy control of the spark-ignited engine
follows the same stages as described in Chap. 3. The adaptation law of the weights
θ f and θg as well as of the supervisory control term u c is derived by the requirement
for negative definite derivative of the quadratic Lyapunov function

1 T 1 T 1 T
V = e Pe + θ̃ θ̃ f + θ̃ θ̃g (10.297)
2 2γ1 f 2γ2 g

Substituting Eq. (10.287) into Eq. (10.297) and differentiating gives

1 T ˙ 1 T ˙
V̇ = 21 ė T Pe + 21 e T P ė + γ1 θ̃ f θ̃ f + γ2 θ̃g θ̃g ⇒
(10.298)
V̇ = 21 e T {(A − BK T )T P + P(A − BK T )}e+
+B T Pe(u c + w + d1 ) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙ g
10.6 Flatness-Based Adaptive Fuzzy Control of the Spark-Ignited Engine 563

Assumption 1: For given positive definite matrix Q and coefficients r and ρ there
exists a positive definite matrix P, which is the solution of the following matrix
equation
T 2 1
(A − BK T ) P + P(A − BK T ) − PB( − 2 )B T P + Q = 0 (10.299)
r ρ

Substituting Eq. (10.299) into V̇ yields after some operations

V̇ = − 21 e T Qe + 21 e T PB( r2 − 1
ρ2
)B T Pe + B T Pe − r1 e T PB +
(10.300)
+B T Pe(w + d1 ) + γ11 θ̃ Tf θ̃˙ f + 1 θ̃ T θ̃˙
γ2 g g

It holds that
θ̃˙ f = θ̇ f − θ˙∗f = θ̇ f
(10.301)
θ̃˙g = θ̇g − θ˙g∗ = θ̇g

The following weight adaptation laws are considered [410]


−γ1 e T PBφ(x) i f ||θ f || < m θ f
θ̇ f = (10.302)
0 ||θ f || ≥ m θ f

−γ2 e T PBφ(x)u c i f ||θg || < m θg
θ̇g = (10.303)
0 ||θg || ≥ m θg

θ̇ f and θ̇g are set to 0, when

||θ f || ≥ m θ f , ||θg || ≥ m θg . (10.304)

The update of θ f follows a gradient algorithm on the cost function 21 ( f − fˆ)2 . The
update of θg is also of the gradient type, while u c implicitly tunes the adaptation gain
γ2 . Substituting Eqs. (10.302) and (10.303) in V̇ finally gives

V̇ = − 21 e T Qe − 2ρ1 2 e T PBBT Pe + e T PB(w + d1 )−

−e T PB(θ f − θ ∗f )T φ(x) − e T PB(θg − θg∗ )T φ(x)u c ⇒
(10.305)
V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBBT Pe + e T PB(w + d1 ) + e T PBwα .

Denoting w1 = w + d1 + wα one gets

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBBT Pe + e T PBw1 or equivalently,
(10.306)
V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBBT Pe + 21 e T PBw1 + 21 w1T B T Pe
564 10 Differential Flatness Theory for Internal Combustion Engines

Lemma: The following inequality holds:

1 T
2 e PBw1 + 21 w1T B T Pe − 1 T
2ρ 2
e PBBT Pe ≤ 21 ρ 2 w1T w1 (10.307)

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality one gets

ρ 2 a 2 + ρ12 b2 − 2ab ≥ 0 ⇒ 21 ρ 2 a 2 + 2ρ1 2 b2 − ab ≥ 0 ⇒

(10.308)
ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2 ⇒ 21 ab + 21 ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2

The following substitutions are carried out: a = w1 and b = e T PB and the previous
relation becomes
1 T
2 w1 B T Pe + 21 e T PBw1 − 1 T
2ρ 2
e PBBT Pe ≤ 21 ρ 2 w1T w1 (10.309)

The previous inequality is used in V̇ , and the right part of the associated inequality
is enforced
1 1
V̇ ≤ − e T Qe + ρ 2 w1T w1 (10.310)
2 2
Equation (10.310) can be used to show that the H∞ performance criterion is satisfied.
The integration of V̇ from 0 to T gives
T T T
0 V̇ (t)dt ≤ − 21 0 ||e||2Q dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
(10.311)
T T
2V (T ) + 0 ||e|| Q dt
2 ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt

Moreover, if there exists a positive constant Mw > 0 such that

∞
0 ||w1 ||2 dt ≤ Mw (10.312)

then one gets

∞
0 ||e||2Q dt ≤ 2V (0) + ρ 2 Mw (10.313)
∞
Thus, the integral 0 ||e||2Q dt is bounded. Moreover, V (T ) is bounded and from the
definition of the Lyapunov function V in Eq. (10.297) it becomes clear that e(t) will
be also bounded since e(t) ∈ Ωe = {e|e T Pe≤2V (0) + ρ 2 Mw }.
According to the above and with the application of Barbalat’s Lemma one finally
obtains that limt→∞ e(t) = 0.
10.6 Flatness-Based Adaptive Fuzzy Control of the Spark-Ignited Engine 565

10.6.4 Simulation Tests

The efficiency of the proposed control scheme was tested in the case of tracking of
different setpoints. The associated results are depicted in Figs. 10.32 and 10.33. It has
been confirmed that the closed control loop succeeded fast and accurate tracking to
all these setpoints, while the variations of the control signal applied to the SI engine
remained smooth.

(a) (b)
5 4
ω × 10 3

ω × 10 3
2

0 0
0 5 10 15 20 0 5 10 15 20
time time
2 2
d
d

1 1
Pm
Pm

0 0
0 5 10 15 20 0 5 10 15 20
time time
2 2
Pm
Pm

1 1

0 0
0 5 10 15 20 0 5 10 15 20
time time

Fig. 10.32 Adaptive fuzzy control of the SI engine: a tracking of setpoint 1 from state variables
xi , i = 1, . . . , 3, b tracking of setpoint 2 from state variables xi , i = 1, . . . , 3

(a) (b)
10 10
3
3
ω × 10

ω × 10

5 5

0 0
0 5 10 15 20 0 5 10 15 20
time time
2 5
Pm d

d
Pm

1 0

0 −5
0 5 10 15 20 0 5 10 15 20
time time
5 5
Pm

Pm

0 0

−5 −5
0 5 10 15 20 0 5 10 15 20
time time

Fig. 10.33 Adaptive fuzzy control of the SI engine: a tracking of setpoint 3 from state variables
xi , i = 1, . . . , 3, b tracking of setpoint 4 from state variables xi , i = 1, . . . , 3
566 10 Differential Flatness Theory for Internal Combustion Engines

10.7 Flatness-Based Control and Kalman Filtering

of the Air–Fuel Ratio

10.7.1 Overview

The problem of embedded nonlinear control for the air–fuel (AF) ratio in combustion
engines is critical for the automotive industry and in general for the development of
more efficient transportation means (e.g., trains, ships, and aircrafts) that will have
increased power while also emit minimal pollution. To this end, several research
results have been produced. In [105] an observer-based fuel injection control algo-
rithm is suggested for fast response and small amplitude chattering of the air-to-fuel
ratio. In [391] the application of linear observer theory to engine control is discussed
with a specific focus on observers based on exhaust measurements. In [174] an inverse
engine model is developed using a nonlinear dynamics identification procedure and
is used for the design of an air–fuel ratio controller. In [148] one can note the use of
exhaust gas oxygen sensors that permit more accurate estimates of A/F ratio and the
use of this variable in PI controller of the combustion engine. In [18, 19] recursive
neural networks models are trained and tested to simulate both forward and inverse
nonlinear dynamics of the intake manifold air–fuel flow. In [20, 313] Kalman fil-
tering methods are used for the identification of the fuel injection dynamics on an
SI engine. In [227] an adaptive control structure is developed for the air–fuel ratio
control problem which consists of an adaptive PI controller and an adaptive Smith
predictor for time-delay systems. In [237] a nonlinear model of the exhaust manifold
is used to describe the dynamics of the different gas mass flows. Then, a Takagi–
Sugeno’s fuzzy model is derived and using this an observer is developed. In [188]
a dynamic feedback linearization approach is proposed for the air–fuel ratio system
considering that some of the parameters of the nonlinear dynamical model are time
invariant or exhibit slow dynamics. In [345] an air-to-fuel ratio control algorithm is
designed based on a switching frequency regulator that has robust stability properties
in the presence of both input and model errors. In [311] a neural network model of the
engine and a neural network controller are developed based on the idea of approx-
imate dynamic programming to achieve optimal control. The method is applied to
both engine torque and exhaust air–fuel ratio (AFR) control. In [61] a robust to delay
errors backstepping control approach is followed and applied for the control of the
Air–Fuel Ratio in Spark Ignition engines. In [390] a switching controller is proposed
based on multiple local models representation of the combustion engine’s air-fuel
ratio dynamics and on the solution of Linear Matrix Inequalities (LMIs). In [195] an
observer-based estimation approach is developed for the air–fuel ratio in combustion
engines, while the provided estimates are used by a decoupled PI compensator that is
designed for each cylinder so as to compensate the cylinder-by-cylinder variations. In
[346] air–fuel ratio control for watercrafts is developed based on feedback lineariza-
tion approach, while load estimation is performed with the use of Kalman Filtering.
In [517] a global linearizing feedback control method is provided for combustion
10.7 Flatness-Based Control and Kalman Filtering of the Air–Fuel Ratio 567

engine models. To cope with model uncertainties the control algorithm is supplied
with adaptive features.
In this section the problem of nonlinear feedback control of the air–fuel ratio in
combustion engines is solved with the use of differential flatness theory and also with
the use of the previously analyzed nonlinear filtering method, known as Derivative-
free nonlinear Kalman Filter. It is shown that the dynamic model of the air–fuel
system is a differentially flat one, which means that all its state vector variables
and its control inputs can be expressed as functions of a unique algebraic variable,
the so-called flat output, and its derivatives. Moreover, it is shown that the air–fuel
ratio system admits dynamic feedback linearization. This means that its state vector
is extended by considering as additional state variable one of its control inputs.
Next, by applying the change of variables that is designated by differential flatness
theory (diffeomorphism), it is shown that the dynamical system can be written in
the canonical Brunovsky form. For the latter linearized description of the air–fuel
system dynamics the design of a state feedback controller becomes easier.
Another problem that has to be solved in the design of the air–fuel ratio control
system is that the model of the dynamical system is subjected to parametric uncertain-
ties and external disturbances [391, 452, 456, 588]. Moreover, there are time delays
in the transmission of the air–fuel sensor (λ sensor) measurements to the electronic
control unit (ECU). To compensate for these perturbations, a disturbances estimator
based on the Derivative-free nonlinear Kalman Filter is used. The estimator applies
the standard Kalman Filter recursion on the linearized equivalent of the system and
makes also use of the inverse of the differential flatness theory-based transformation
to provide estimates of the state variables of the initial nonlinear filter. By estimating
the perturbation terms it becomes easy to compensate for their effects through the
inclusion of an additional term in the feedback control scheme. The efficiency of the
proposed control scheme is tested through simulation experiments.

10.8 Dynamic Model of the Air–Fuel Ratio System

10.8.1 The Air and Fuel Flow Models

The control system of the combustion (piston) engine comprising the control of the
airflow and the control of the fuel flow is depicted in Fig. 10.34. One simple model
of the airflow in an intake manifold is the filling and emptying model [188, 391]. The
airflow enters the manifold through the throttle and is pumped out of the manifold
into the cylinder. Assuming no leaks, the airflow ṁ at into the manifold and the flow
ṁ a out of the manifold are identical only in steady state. During throttle transients,
the difference between the two flows equals the rate of change of air mass in the man-
ifold plenum. Assuming that the manifold pressure is uniform and that the manifold
temperature (Tm ) is constant, then the continuity equation and ideal gas law can be
applied to the manifold plenum. The air-filling dynamics are expressed in terms of
568 10 Differential Flatness Theory for Internal Combustion Engines

Fig. 10.34 The control system of the combustion engine: control of the airflow and control of the
fuel flow

the manifold pressure pm

ṗm = − η120V
v N Vd
m
pm + RTm
Vm ṁ at (10.314)

where Vd is the engine’s volume and Vm is the intake manifold volume. This is
a nonlinear differential equation since the airflow rate through the valve ṁ at is a
nonlinear function of the manifold pressure pm [391]. The volumetric efficiency ηv
is a variable which measures the performance of the cylinder’s inlet port and valve,
and this depends also on pm .
The airflow rate into the cylinder is given by [391]
ηv Vd ω
ṁ a = 120RTm pm (10.315)

The above equations represent a first-order dynamic system that describes the mass
flow in and out of the manifold ignoring inertial effects. A different formulation of
the airflow dynamics is
10.8 Dynamic Model of the Air–Fuel Ratio System 569

ṁ a = τman [m a
1
− m ss (a, ω)] (10.316)

where m a is the mass of air inducted per intake stroke and m ss is the steady state value
of m a . The time constant τman and m ss are determined experimentally and depend
on the throttle angle α and the engine speed ω. Finally, the air charge in the cylinder
(m ac ) during one cycle, is computed from the relation ṁ ac = K (a, ω, Δt)m a , where
K describes the airflow into each cylinder as a function of the time through the
engine’s cycle Δt.
Next, the fuel flow of the engine is described. Previous research has indicated that
the fuel jet from the injector mixes with the airstream and also deposits components on
the wall of the intake system. The fuel vapors and the airborne droplets are transported
with the fast-moving airstream. The liquid fuel flows a1ong the walls (wall film mass
flow) and also enters the airstream through evaporation with a time constant of the
order of a few engine cycles, although it takes much longer in a cold start. At the
steady state conditions of the engine’s functioning, the fuel droplet deposition on the
wall surface balances the fuel evaporation into the airstream. Then, the amount of the
fuel entering the cylinder is equal to the quantity of the fuel injected. The dynamics
of the injected fuel is then modeled by [188, 391]

ṁ WF (t) = − τ1f m WF (t) + (1 − X )ṁ f i (t) (10.317)

ṁ φ (t) = 1
τf m WF (t) + X ṁ f i (t) (10.318)

where ṁ fi is the injected fuel flow rate, m WF is the wall film mass flow, ṁ φ is the
fuel flow rate entering the cylinder, X is the fraction of the injected fuel that enters
the cylinder, τ f is a time constant of the fuel evaporation process.

10.8.2 Dynamics of the Air–Fuel Ratio System

Continuing the previous analysis on air–fuel flow in the combustion engine a compact
model for the dynamics of the air–fuel ratio will be developed [188]. As shown in
Fig. 10.35, the air–fuel ratio control system in combustion engines comprises three
subsystems: (i) the air path Pa , (ii) the fuel path Pφ , and (iii) the combustion part Pλ .
The primary variable in the combustion subsystem is the air–fuel ratio. The air–fuel
ratio is given by [188]
ṁ a (t)
λ(t) = σ0 ṁ φ (t)
(10.319)

Variable ṁ a (t) denotes the air mass flow, while variable ṁ φ (t) denotes the fuel mass
flow.
The primary variable of the air path is the manifold pressure pm (t), with dynamics
given by Eqs. (10.314) and (10.315). Denoting as new control input u a (t) = ṁ at ,
570 10 Differential Flatness Theory for Internal Combustion Engines

Fig. 10.35 Air–fuel ratio control system for combustion engines

and the functions aa (t) = − η120V

v N Vd
m
, ba (t) = RT
Vm and ca (t) =
m ηv Vd ω
120RTm one has the
following description of the airflow dynamics [188]:

ṗm (t) = aa pm (t) + ba u a (t)

(10.320)
ṁ a (t) = ca pm (t)

About the dynamics of the fuel flow this has been described by Eqs. (10.317) and
(10.318). Next, by defining the new control input u φ (t) = ṁ fi (t), and the functions
f φ ( pm , ω) = τ1f and gφ ( pm , ω) = X , the dynamics that is described by the following
set of differential equations [188]

ṁ W F (t) = − f φ ( pm , ω)m W F (t) + [1 − gφ ( pm , ω)]u φ (t)

(10.321)
ṁ φ (t) = f φ ( pm , ω)m W F (t) + gφ ( pm , ω)u φ (t)

The primary variable of the fuel path is the wall film mass flow m W F (t). Variable ω
denotes the engine’s turn speed.

10.8.3 Differential Flatness of the Air–Fuel Ratio System

The air–fuel ratio has been defined as

ṁ a (t)
λ(t) = σ0 ṁ φ (t)
(10.322)
10.8 Dynamic Model of the Air–Fuel Ratio System 571

The dynamics of the air mass flow is given by

ṗm (t) = a0 pm (t) + ba u a (t)

(10.323)
ṁ a (t) = ca pm (t)

The dynamics of the fuel mass flow is given by

ṁ WF (t) = − f φ ( pm , ωe )m WF (t) + [1 − gφ ( pm , ωe )]u φ (t) (10.324)

ṁ φ (t) = f φ ( pm , ωe )m WF (t) + gφ ( pm , ωe )u φ (t) (10.325)

Following the concept of dynamic feedback linearization, the control input u φ (t) is
considered to be an additional state variable, and the associated dynamics is given by

u̇ φ (t) = v1 (10.326)

The flat output for the extended state-space description of the system is taken to be

ỹ = [y1 , y2 , y3 ] = [m a , m φ , u φ ] (10.327)

From Eq. (10.322) it holds that

λ = σ0 ẏẏ21 (10.328)

From Eq. (10.323) it holds that

pm (t) = ca ṁ a (t)⇒ pm (t)

1
= 1
ca ẏ1 (10.329)

From Eq. (10.323) one has

u a (t) = b1a [ ṗm (t) − a0 pm (t)]⇒

u a (t) = b10 [ c10 ÿ1 − ac00 ẏ1 ]⇒ (10.330)
u a (t) = f 1 ( ỹ)

It is considered that the rotational speed of the engine is a function of y1 = m a and

y2 = m φ , that is,

ωe = f 2 (y, ẏ) (10.331)

Equation (10.325) is solved with respect to m WF (t). It holds that

ṁ φ (t)−gφ ( pm ,ωe )vφ (t)

m WF (t) = f φ ( pm ,ωe ) ⇒
ẏ2 (t)−gφ ( pm ,ωe )y3
m WF (t) = f φ ( pm ,ωe ) ⇒ (10.332)
m WF (t) = f 3 (y, ẏ)
572 10 Differential Flatness Theory for Internal Combustion Engines

Moreover, from Eq. (10.326) one has

u̇ φ (t) = v1 ⇒v1 = ẏ3 (10.333)

Therefore, for the extended state-space description of the system having as state
vector

x̃ = [λ, pm , m a , m WF , m φ , u φ ] (10.334)

and as control inputs vector ũ = [u a , v1 , φ( pm , ωe )], where v1 = u̇ φ (t), one has

that all state variables and the control inputs are functions of the flat output ỹ =
[m a , m φ , u φ ] and of its derivatives (by introducing the virtual control input φ( pm , ωe )
one has the dimension of the flat output vector is equal to the dimension of the control
inputs vector). Therefore, the air–fuel ratio control system is a differentially flat one.

10.8.4 Flatness-Based Control of the Air–Fuel Ratio System

It will be shown that the air–fuel ratio control system can be transformed to the linear
canonical form. From Eq. (10.322) one has

ṁ a (t)
λ(t) = σ0 ṁ φ (t)
⇒
d d
ṁ a (t)·ṁ φ (t)−ṁ a (t)· dt ṁ φ (t)
λ̇(t) = σ0 dt ṁ φ (t)2
⇒ (10.335)
d d
[c ṗ (t)]· ṁ φ (t)−ṁ (t)· dt φ ( pm ,ωe )m WF (t)+gφ ( pm ,ωe )u φ (t)]
[ f
λ̇(t) = σ0 dt a m a
ṁ φ (t)2

This is also written as

σ0
λ̇(t) = {c [a p (t) + ba u a (t)]ṁ φ (t)}−
ṁ φ (t)2 a 0 m
− ṁ σ(t)
0 ˙
2 ṁ a (t){ f φ ( pm , ωe )m WF (t) + f φ ( pm , ωe )ṁ WF (t)}− (10.336)
φ
− ṁ σ(t)
0
2 ṁ a (t){ ġφ ( pm , ωe )u φ (t) + gφ ( pm , ωe )u̇ φ (t)}
φ

The previous equation can also take the form

σ0
λ̇(t) = {c a p (t)ṁ φ (t)}−
ṁ φ (t)2 a 0 m
− ṁ σ(t)
0 ˙
2 ṁ a (t){ f φ ( pm , ωe )m WF (t) + f φ ( pm , ωe )ṁ WF (t)}−
φ
− ṁ σ(t)
0
2 ṁ a (t){ ġφ ( pm , ωe )u φ (t)}+ (10.337)
φ
+ ṁ σ(t)
0
2 ca ba ṁ φ (t)u a (t)+
φ
− ṁ σ(t)
0
2 ṁ a (t)gφ ( pm , ωe )u̇ φ (t)
φ
10.8 Dynamic Model of the Air–Fuel Ratio System 573

By defining the following functions and variables

f ( ỹ) =
σ0
{c [a p (t) + ba u a (t)]ṁ φ (t)}−
ṁ φ (t)2 a 0 m
− σ0 ˙
2 ṁ a (t){ f φ ( pm , ωe )m WF (t) + f φ ( pm , ωe )ṁ WF (t)}−
(10.338)
ṁ φ (t)
− ṁ σ(t)
0
2 ṁ a (t){ ġφ ( pm , ωe )u φ (t) + gφ ( pm , ωe )u̇ φ (t)}+
φ

σ0
g( ỹ) = c b ṁ (t)
ṁ φ (t)2 a a φ (10.339)

g̃( ỹ) = − ṁ σ(t)

0
2 ṁ a (t)gφ ( pm , ωe ) (10.340)
φ

and the control input v1 = u̇ φ (t), the system is written in the following form

λ̇(t) = f ( ỹ) + g( ỹ)u a (t) + g̃( ỹ)v1 (t) (10.341)

Next, it is considered that the input u a (t) is kept constant, while the following function
is defined
f˜( ỹ) = f ( ỹ) + g( ỹ)u a (t) (10.342)

Thus, one obtains the following linearized description of the system

λ̇(t) = f˜( ỹ) + g̃( ỹ)v1 (t) (10.343)

If all system parameters are considered to be known, and measurable, then a suitable
feedback control law is

v1 (t) = g̃( ỹ) [λ̇d (t) −

1
f˜( ỹ) − k1 (λ(t) − λd (t))] (10.344)
t
while the control input that is exerted on the system is u φ (t) = 0 v1 (t)dt. Consid-
ering the tracking error variable e = λ(t) − λd (t) and using the feedback control of
Eq. (10.344) the error dynamics of the closed loop, for k1 > 0 is given by

ė(t) + k1 e(t) = 0⇒limt→∞ e(t) = 0 (10.345)

10.8.5 Compensation of Uncertainties

with the Derivative-Free Nonlinear Kalman Filter

Next, it is considered that uncertainties and external disturbances in the air–fuel

control system are modeled as an additive term d̃

λ̇(t) = f˜( ỹ) + g̃( ỹ)v1 + d̃ (10.346)

574 10 Differential Flatness Theory for Internal Combustion Engines

It is also assumed that the disturbance’s dynamics is described by the associated

n-th order derivative (e.g., n = 2) which gives d̃ (n) = f d (t). A new extended state
˙ = [z , z , z ] = z̃. This allows to write
vector for the system is defined as [λ, d̃, d̃] 1 2 3
Eq. (10.346) in the form

ż 1 = f˜(z̃) + g̃(z̃)v1 + z 2
ż 2 = z 3 (10.347)
ż 3 = f d

By defining the new control input v1∗ = f˜(z̃) + g̃(z̃)v1 the previous state-space
description is written as

ż 1 = z 2 + v1
ż 2 = z 3 (10.348)
ż 3 = f d

which also takes the matrix form

z̃˙ = Az̃ + B ṽ
(10.349)
z m = C z̃

or
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 1 0  ∗
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0 v
0⎠ 1
fd
ż 3 000 z3 0 1 (10.350)

z m = 1 0 0 z̃

For the aforementioned system, one can perform estimation of the perturbation term
d̃ using a Kalman Filter-based disturbances estimator [31, 408, 414, 437]

ẑ = Ao ẑ + Bo v∗ + K (z m − ẑ m )
(10.351)
ẑ m = Co ẑ

where Ao = A, Co = C and
⎛ ⎞
1 0
Bo = ⎝0 0⎠ (10.352)
0 0

One can also consider compensation of time delays in the readings of the λ sensor.
Such a time delay is described by

λs (s) = e−as λ(s)⇒λ̇s (t) = − a1 λs (t) + a1 λ(t) (10.353)

10.8 Dynamic Model of the Air–Fuel Ratio System 575

Therefore, the aggregate dynamics of the air–fuel ratio control system becomes

λ̇(t) = f˜( ỹ) + g̃( ỹ)v1 + d̃

(10.354)
λ̇s (t) = − a1 λs (t) + a1 λ(t)

˙ = [z , z , z , z ] = z̃, and
Now, the extended state vector is taken to be [λ, λs , d̃, d̃] 1 2 3 4
it holds that

ż 1 = f˜(z̃) + g̃(z̃)v1 + z 3
ż 2 = − a1 z 2 + a1 z 1
(10.355)
ż 3 = z 4
ż 4 = f d

By defining the new control input v1∗ = f˜(z̃) + g̃(z̃)v1 the previous state-space
description is written as

ż 1 = z 3 + v1∗
ż 2 = − a1 z 2 + a1 z 1
(10.356)
ż 3 = z 4
ż 4 = f d

The state-space description of the system becomes

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 0 0 1 0 z1 1 0  
⎜ż 2 ⎟ ⎜ 1 − a1 0⎟ ⎜ z 2 ⎟ ⎜0 0⎟ ∗
⎜ ⎟ = ⎜a 0 ⎟⎜ ⎟ + ⎜ ⎟ v1
⎝ż 3 ⎠ ⎝ 0 0 0 1 ⎠ ⎝ z 3 ⎠ ⎝0 0 ⎠ fd
(10.357)
ż 4 0 0 0 0 z4 0 1

z m = 0 1 0 0 z̃

Consequently, the estimation of the disturbance term d̃ can also take place with the
use of a Kalman Filter-based disturbances estimator

ẑ = Ao ẑ + Bo v∗ + K (z m − ẑ m )
(10.358)
ẑ m = Co ẑ

where Ao = A, Co = C and
⎛ ⎞
1 0
⎜0 0⎟
Bo = ⎜
⎝0
⎟ (10.359)
0⎠
0 0

The application of the observability condition confirms that the previous description
of the system is again observable.
576 10 Differential Flatness Theory for Internal Combustion Engines

(a) 2 (b) 2
1.5 1.5
AFR

AFR
1 1

0.5 0.5

0 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
Estimation of disturbance

Estimation of disturbance
2 2

1 1
input Tm

input Tm
0 0

−1 −1

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 10.36 Air–fuel ratio control loop performance for setpoint 1. a Control of the AF factor
(air–fuel ratio of the combustion engine) and estimation of disturbance terms Tm with an ideal λ
sensor. b Control of the AF factor (air–fuel ratio of the combustion engine) and Kalman Filter-based
estimation of disturbance terms Tm with a λ sensor exhibiting time delay

(a) 2 (b) 2

1.5 1.5
AFR

AFR

1 1

0.5 0.5

0 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
Estimation of disturbance

Estimation of disturbance

2 2

1 1
input Tm

input Tm

0 0

−1 −1

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 10.37 Air–fuel ratio control loop performance for setpoint 2. a Control of the AF factor
(air–fuel ratio of the combustion engine) and estimation of disturbance terms Tm with an ideal λ
sensor. b Control of the AF factor (air–fuel ratio of the combustion engine) and Kalman Filter-based
estimation of disturbance terms Tm with a λ sensor exhibiting time delay

For the above definition of dynamics of the disturbances estimator, the selection of
the observer’s gain K can be performed using the standard Kalman Filter recursion.
Prior to this, matrices Ao , Bo , and Co are brought to the discrete-time form Ad ,
Bd , and Cd using common discretization methods. The discrete-time Kalman Filter
recursion is [31, 408, 414]
10.8 Dynamic Model of the Air–Fuel Ratio System 577

(a) 6 (b) 6

4 4
AFR

AFR
2 2

0 0
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
Estimation of disturbance

Estimation of disturbance
6 6

4 4
input Tm

input Tm
2 2

0 0

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 10.38 Air–fuel ratio control loop performance for setpoint 3. a Control of the AF factor
(air–fuel ratio of the combustion engine) and estimation of disturbance terms Tm with an ideal λ
sensor. b Control of the AF factor (air–fuel ratio of the combustion engine) and Kalman Filter-based
estimation of disturbance terms Tm with a λ sensor exhibiting time delay

measurement update:

K (k) = P − (k)CdT [Cd P − (k)CdT + R]−1

ẑ(k) = x̂ − (k) + K (k)[z m (k) − ẑ m (k)] (10.360)
P(k) = P − (k) − K (k)Cd P − (k)

time update:

P − (k + 1) = Ad P(k)AdT + Q
(10.361)
ẑ(k + 1) = Ad ẑ(k) + Bd vin (k)

10.8.6 Simulation Tests

The performance of the air–fuel ratio embedded control loop, that was based on lin-
earization with the use of differential flatness theory and on estimation of disturbance
terms with the use of the Derivative-free nonlinear Kalman Filter, was evaluated for
the case of tracking of different setpoints. The associated results are depicted in
Figs. 10.36, 10.37, and 10.38. It has been confirmed that the proposed nonlinear
control scheme succeeded fast and accurate tracking for all reference setpoints.
Chapter 11
Differential Flatness Theory for Chaotic
Dynamical Systems

11.1 Introduction

The present chapter analyzes the use of differential flatness theory-based control
and filtering methods for dynamical systems that exhibit chaotic dynamics. First, the
chapter proposes a solution to the problem of control of nonlinear chaotic dynamical
systems, which is based on differential flatness theory and on adaptive fuzzy control.
An adaptive fuzzy controller is designed for chaotic dynamical systems such as the
Lorenz oscillator, under the constraint that the system’s model is unknown. The
control algorithm aims at satisfying the H∞ tracking performance criterion, which
means that the influence of the modeling errors and the external disturbances on the
tracking error is attenuated to an arbitrary desirable level. After transforming the
chaotic system’s model into a linear form, the resulting control inputs are shown to
contain nonlinear elements which depend on the system’s parameters. The nonlinear
terms which appear in the control inputs are approximated with the use of neuro-
fuzzy networks. As in previous chapters, it is shown again that a suitable learning law
can be defined for the aforementioned neuro-fuzzy approximators so as to preserve
the closed-loop system stability. With the use of Lyapunov stability analysis, it is
proven that the proposed adaptive fuzzy control scheme results in H∞ tracking
performance. The efficiency of the proposed adaptive control scheme is confirmed
through simulation experiments, using as case study the Lorenz chaotic oscillator.
Next, differential flatness theory is applied to communication systems. In this case,
the Derivative-free nonlinear Kalman Filter is used for developing a communication
system that is based on a chaotic modulator such as the Duffing oscillator. In the trans-
mitter’s side, the source of information undergoes modulation (encryption) in which
a chaotic signal generated by the Duffing system is the carrier. The modulated signal
is sent through a communication channel and at the receiver’s side demodulation
of the transmitted signal takes place, by exploiting the estimation provided about
the state vector of the chaotic oscillator by the Derivative-free nonlinear Kalman
Filter. Evaluation tests confirm that the proposed filtering method has improved

© Springer International Publishing Switzerland 2015 579

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_11
580 11 Differential Flatness Theory for Chaotic Dynamical Systems

performance over the Extended and the Unscented Kalman Filter and reduces sig-
nificantly the rate of transmission errors. Moreover, it is shown that the proposed
Derivative-free nonlinear Kalman Filter can work within a dual Kalman Filter-
ing scheme, for performing simultaneously transmitter-receiver synchronization and
estimation of unknown coefficients of the communication channel.

11.2 Flatness-Based Control of Chaotic Dynamical Systems

11.2.1 Overview

Chaotic dynamical systems and chaos-based control has been a topic of research
interest in the last years. As shown in the previous chapters, chaotic dynamics is
met in several applications, for example, in electric power systems (e.g., converters),
in communication systems (chaotic spread spectrum communication), in electronic
circuits, and in mechanical vibrating structures, as well as in biological systems.
The response of chaotic systems is dependant on initial conditions. Actually, by
changing initial conditions, the behavior of chaotic systems varies in a manner that
approximates stochasticity. In [79], the design of a feedback controller for Chua’s
chaotic circuit is analyzed, based on a combination of H∞ control theory, adaptive
fuzzy control, and Lyapunov’s stability method. In [247, 568] Takagi-Sugeno fuzzy
modeling is proposed for chaotic systems. A feedback controller is designed based
on Lyapunov’s stability analysis and the solution of LMIs. In [493], linearization
of chaotic system’s dynamics is proposed with the use of differential flatness the-
ory while a disturbance observer is designed for estimating and compensating for
perturbation effects. Observer-based control of chaotic systems is also proposed in
[158]. In [392], a neural network is used for identifying the unknown dynamics of the
chaotic system, while based on the estimated model a sliding mode feedback con-
troller is designed. In [239], the chaotic system is modeled with the Takagi-Sugeno
fuzzy approach. For this latter representation, an observer-based feedback control
scheme is implemented. Stability analysis is provided with the Lyapunov method. In
[96], an adaptive fuzzy system is used to approximate the unknown system dynamics
and next a state feedback controller is designed. Stability analysis is shown with the
Lyapunov method. In [240], a Takagi-Sugeno fuzzy modeling approach is followed
for the chaotic system of unknown dynamics. Using Lyapunov stability method,
adaptation of the parameters of the Takagi-Sugeno fuzzy model is succeeded and a
feedback control law is designed. In [97], Takagi-Sugeno fuzzy modeling is proposed
for synchronizing the control of chaotic systems. Stability proof is based on previous
results for fractional order systems. In [314], a proportional gain feedback controller
is designed for chaotic systems and stability is shown with the use of the Lyapunov
method. In [601], Takagi-Sugeno fuzzy modeling and Lyapunov stability analysis
are implemented on a chaotic system. The control input has impulsive form. Finally,
in [111] the use of differential flatness theory and of Lie algebra for linearization and
control of chaotic oscillators is analyzed.
11.2 Flatness-Based Control of Chaotic Dynamical Systems 581

This section proposes adaptive fuzzy control for chaotic systems of unknown
dynamical model with the use of differential flatness theory. On the one side, adaptive
fuzzy control has been proven to be an efficient nonlinear control method [111, 302,
377, 396, 524, 537, 544, 553, 581]. On the other side, differential flatness theory
stands for a major direction in the design of nonlinear control systems [152, 263,
286, 340, 465, 495, 535]. It is known that neuro-fuzzy approximators can be used in
indirect adaptive control schemes where their role is to identify online the unknown
system dynamics and to provide the controller with this information that is used for
generating the control inputs [104, 426, 593]. Adaptive fuzzy control system based
on differential flatness theory extends the class of systems to which indirect adaptive
control can be applied [407, 410, 413, 433, 454, 591]. This is important for the design
of controllers, capable of efficiently compensating for modeling uncertainties, and
external disturbances in nonlinear dynamical systems, and particularly in chaotic
dynamical systems. Unlike other adaptive fuzzy control schemes which are based on
several assumptions about the structure of the nonlinear system as well as about the
uncertainty characterizing the system’s model, the proposed adaptive fuzzy control
scheme based on differential flatness theory offers an exact solution to the design
of adaptive controllers for chaotic systems of unknown dynamics [427, 453, 455].
The only assumption needed for the design of the controller and for succeeding H∞
tracking performance for the control loop is that there exists a solution for a Riccati
equation associated with the linearized error dynamics of the differentially flat model.
This assumption is quite reasonable for several nonlinear systems, thus providing a
systematic approach to the design of reliable controllers for such systems [426, 433].
The section’s results come to extend the method presented in [407, 410, 413,
433, 454, 591], as well as in Chap. 3. By showing that chaotic dynamical systems
are differentially flat, it becomes possible to transform them to the linear canonical
form. For the latter description of the system’s dynamics, the design of a state feed-
back controller becomes easier. After transformation to the linear canonical form,
the resulting control input for the chaotic dynamical system is shown to contain non-
linear elements which depend on the system’s parameters. In the case of unknown
parameter values, the nonlinear terms which appear in the control signal can be
approximated with the use of neuro-fuzzy networks [31, 414]. In this section, it is
shown that a suitable learning law can be defined for the aforementioned neuro-fuzzy
approximators so as to preserve the closed-loop system stability. Lyapunov stability
analysis proves also that the proposed flatness-based adaptive fuzzy control scheme
results in H∞ tracking performance.

11.2.2 Differential Flatness of Chaotic Dynamical Systems

Next, it will be shown that benchmark chaotic systems satisfy differential flatness
properties. Among systems one can distinguish the Lorenz chaotic oscillator, the
Duffing chaotic system, and the Genesio-Tessi chaotic system.
582 11 Differential Flatness Theory for Chaotic Dynamical Systems

1. The Lorenz chaotic system

The Lorenz system is obtained from studying a fluid layer heated from below and
cooled from above, such that a temperature difference is established across it. The
convection motion is described by the Navier-Stokes equations. Taking Fourier
expansions of these equations along two spatial directions and truncating the remain-
ing expressions after the third mode, the following simplified model is obtained:

ẋ = σ (y − z)
ẏ = ρx − y − x z (11.1)
ż = −βz + x y

where σ , ρ and β are real parameters denoting the Prandtl number, the Rayleigh
number, and a geometric factor, respectively. The state variables x, y, and z represent
measures of fluid velocities and the spatial temperature distribution in the fluid layer
under gravity. The Rayleigh number can be manipulated by changing the heat transfer
to the fluid from below. This parameter is considered to be a control input u = ρ.
Common values for parameters σ and β are σ = 10 and β = 8/3. Indicative phase
diagrams of the Lorenz chaotic oscillator are shown in Fig. 11.1.
The flat output of the Lorenz system given in Eq. (11.1) is defined as [111]:

x2
λ= 2 − σz + k (11.2)

The transformed state vector is

z = [z 1 , z 2 , z 3 ]T = Φ(x)⇒
(11.3)
z= [x 2 /2 − σ z + k, σ (βz − x 2 ), σ γ x y + 2σ 2 x 2 − σ β 2 z]T

phase diagram x −x phase diagram x −x

1 3 2 3
(a) 4.5 (b) 4.5
4 4

3.5 3.5

3 3

2.5 2.5
x3
3
x

2 2

1.5 1.5

1 1

0.5 0.5

0 0
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
x1 x2

Fig. 11.1 a Indicative phase diagram for the Lorenz system for states x1 and x3 . b Indicative phase
diagram for the Lorenz system for states x2 and x3
11.2 Flatness-Based Control of Chaotic Dynamical Systems 583

where γ = β − 2σ . The inverse transformation is given by

x = [x, y, z]T = Φ −1 (z)


2)
x = ± 2(βε+z
γ (11.4)
y = ± 2σβz 1 +(β+2σ
√ )z 2 +z 3 −2σβk
σ 2γ (βε+z ) 2
2σ ε+z 2
z= σγ

where ε = z 1 − k. The transformed dynamical system can be written in the linear

canonical (Brunovsky) form according to

ż 3 = λ(3) = v (11.5)

where
v = a(x) + β(x)u (11.6)

where u = ρ and

a(x) = −σ γ x 2 z − 4σ 3 x 2 − σ (γ + β 2 + σβ − 6σ 2 )x y + σ 2 γ y 2 + σ β 3 x
(11.7)

β(x) = σ γ x 2 (11.8)

It holds that
z 1 = λ = x 2 /2 − σ z + k
z 2 = λ̇ = σ (βz − x 2 ) (11.9)
z 3 = λ̈ = σ γ x y + 2σ 2 x 2 − β 2 σ x

Finally, it is shown that all state variable of the Lorenz system can be written as
functions of the flat outputs and of its derivatives. Indeed, it holds
√
x = ± 2κ/γ
√
y = ±(λ̈ + (β + 2σ )λ̇ + 2σβλ − 2σ k)/(σ 2γ κ) (11.10)
z = (β λ̇ + 2σβλ − 2σ κ)/(σβγ )

u = {−2βγ κv − 2βγ (βκ + κ + 2cσ − σ k)λ̈−

−2β λ̇(4kc − 2c2 − 2k 2 + 2σ γ κ + βγ κ + 2σβγ c − γ σβκ)−}
−4σβ(βγ κ + 4κc − 2c2 − 2k 2 )λ − 8σ kc2 + 4σβγ kκ− (11.11)
−8σ k 3 + 16σ k 2 c + β 2 γ (β + 2σ )λ˙2 + 2βγ (β + σ )λ̈λ̇+
+βγ λ̈2 + 2σ β 2 γ λ̈λ + 2σ β 3 γ λ̇λ}/(4σβγ κ 2 )
584 11 Differential Flatness Theory for Chaotic Dynamical Systems

where c = λ̇ + βλ, κ = c − k and v = λ(3) . After expressing the dynamics of the

Lorenz system with the new state variables z 1 , z 2 , z 3 one has the following linear
canonical form ⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ż 1 010 z1 0
⎝ż 2 ⎠ = ⎝0 0 1⎠ ⎝z 2 ⎠ + ⎝0⎠ v (11.12)
ż 3 000 z3 1

2. Duffing’s chaotic system:

ẋ1 (t) = x2 (t)

(11.13)
ẋ2 (t) = 1.1x1 (t) − x13 (t) − 0.4x2 (t) + 1.8cos(1.8t)

Differential flatness of Duffing’s oscillator:

By defining the flat output y = x1 one has x1 = y, x2 = ẏ, and u =
1.8cos(1.8t) = ÿ − 1.1y + y 3 + 0.4 ẏ. Thus all state variables and the control
input of the Duffing oscillator are expressed as functions of the flat output and its
derivatives, and differential flatness of the oscillator is proven. Moreover, by defin-
ing the new control input v = 1.1x1 − x13 − 0.4x2 + u, and the new state variables
y1 = y and y2 = ẏ a description of the oscillator’s dynamics in the linear canonical
(Brunovsky’s) form is obtained
      
ẏ1 01 y1 0
= + v (11.14)
ẏ2 00 y2 1

3. The Genesio-Tesi chaotic system:

ẋ1 (t) = x2 (t)

ẋ2 (t) = x3 (t) (11.15)
ẋ3 (t) = −cx1 (t) − bx2 (t) − αx3 (t) + x12 (t)

where a, b, and c are real constants. When at least one of the system’s Lyapunov
exponents is larger than zero the system is considered to be chaotic. For example,
when a = 1.2, b = 2.92, and c = 6 the system behaves chaotically.
By defining the flat output y = x1 one has x1 = y, x2 = ẏ, x3 = ÿ, and
y (3) = ẋ3 = −cy−b ẏ−a ÿ+y 2 . Thus all state variables of the Genesio-Tesi oscillator
are expressed as functions of the flat output and its derivatives, and differential
flatness of the oscillator is proven. Moreover, by defining as new control input v =
−cy − b ẏ − a ÿ + y 2 , and the new state variables y1 = y, y2 = ẏ and y3 = ÿ a
description of the oscillator’s dynamics in the linear canonical (Brunovsky’s) form
is obtained ⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (11.16)
ẏ3 000 y3 1
11.2 Flatness-Based Control of Chaotic Dynamical Systems 585

4. The Chen chaotic system

ẋ1 (t) = −a(x1 (t) − x2 (t))

ẋ2 (t) = (c − a)x1 (t) + cx2 (t) − x1 (t)x3 (t) (11.17)
ẋ3 (t) = x1 (t)x2 (t) − bx3 (t)

where a, b, and c are positive parameters. When at least one of the system’s Lyapunov
exponents is larger than zero the system is considered to be chaotic. For example,
when a = 40, b = 3, and c = 3 the system behaves chaotically.
By defining the flat output y = x1 one has x1 = y, x2 = a1 ( ẏ + ay) and x3 =
2 )y
− ÿ+(a−c) ẏ−(2ac−a
ay (for y = 0). From the third row of the associated state-space
equations and after intermediate computations, one obtains y (3) = −(a−c) ÿ+(2ac−
2
a 2 ) ẏ+ ẏy ÿ+ ẏy (a−c)−(2ac−a 2 ) ẏ+y 2 ẏ+a y 3 +b ÿ+b(a−c) ẏ−b(2ac−a 2 y). Thus
all state variables of the Chen oscillator are expressed as functions of the flat output
and its derivatives, and differential flatness of the oscillator is proven. Moreover, by
defining the new control input v to be equal to the right part of the previous equation
about y (3) , and the new state variables y1 = y, y2 = ẏ, and y3 = ÿ a description of
the oscillator’s dynamics in the linear canonical (Brunovsky’s) form is obtained
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (11.18)
ẏ3 000 y3 1

It is noted that differential flatness can be proven for several other types of chaotic
oscillators, such as Rössler’s system and Chua’s circuit. By expressing all state vari-
ables and the control input of such systems as functions of the flat output and
their derivatives, it is possible to obtain again a description in the linear canoni-
cal (Brunovsky) form. The evolution in time of the phase diagram of the Duffing
oscillator and of the Genesio-Tesi chaotic system are shown in Fig. 11.2.

11.2.3 Flatness-Based Adaptive Fuzzy Control

for Chaotic Systems

11.2.4 Design of the Feedback Controller

After application of differential flatness theory and of the associated diffeomorphism,

the dynamics of the chaotic dynamical system is written in the form of Eqs. (10.281)
and (10.282), that is
586 11 Differential Flatness Theory for Chaotic Dynamical Systems

Duffing chaotic oscillator Genesio−Tesi chaotic oscillator

(a) 4 (b) 10
3 8
6
2
4
1 2

x2
x2

0 0

−1 −2
−4
−2
−6
−3 −8
−4 −10
−3 −2 −1 0 1 2 3 −4 −2 0 2 4 6
x1 x1

Fig. 11.2 Phase diagram of typical chaotic oscillators. a Duffing’s oscillator. b Genesio-Tesi’s
oscillator

⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 0 1 0 ··· 0 y1 0
⎜ ẏ2 ⎟ ⎜ 0 0 1 ··· 0⎟ ⎜ y2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ · · · ⎟ ⎜· · · ··· ··· ··· · · ·⎟ ⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ · · · ⎟ + ⎜ · · ·⎟ v
⎜ · · · ⎟ = ⎜· · · ··· ··· ··· ⎟ ⎜
· · ·⎟ ⎜ · · · ⎟ ⎜ ⎟ (11.19)
⎜ ⎟ ⎜ ⎟ ⎜ · · ·⎟
⎝ ẏn−1 ⎠ ⎝ 0 0 0 ··· 1 ⎠ ⎝ yn−1 ⎠ ⎝0⎠
ẏn 0 0 0 ··· 0 yn 1

where v = f (x, t) + g(x, t)(u + d̃) is the control input for the linearized model, and
d̃ denotes additive input disturbances. Thus one can use that the chaotic system’s
model is finally written in the form:

y (n) = f (x, t) + g(x, t)(u + d̃) (11.20)

For this system, as well as for the description of the Lorenz oscillator’s dynamics
given in Eq. (11.5), and for uncertain functions f (x, t) and g(x, t) an appropriate
control law is
1 (n)
u= [y − fˆ(x, t) − K T e + u c ] (11.21)
ĝ(x, t) d

with e = [e, ė, ë, . . . , e(n−1) ]T and e = y − yd , K T = [kn , kn−1 , . . . , k1 ], such that
the polynomial e(n) + k1 e(n−1) + k2 e(n−2) + · · · + kn e is Hurwitz. The control law
of Eq. (11.21) results into

e(n) = −K T e + u c + [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + g(x, t)d̃

(11.22)
11.2 Flatness-Based Control of Chaotic Dynamical Systems 587

where the supervisory control term u c aims at the compensation of the approximation
error
w = [ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u (11.23)

as well as of the additive disturbance term d1 = g(x, t)d̃. The above relation
can be written in a state-equation form. The state vector is taken to be e T =
[e, ė, . . . , e(n−1) ], which after some operations yields

ė = (A − B K T )e + Bu c + B{[ f (x, t) − fˆ(x, t)] + [g(x, t) − ĝ(x, t)]u + d1 }

(11.24)

where ⎛ ⎞ ⎛ ⎞
0 1 0 ··· ··· 0 0
⎜0 0 1 ··· ··· 0⎟ ⎜0⎟
⎜ ⎟ ⎜ ⎟
⎜· · · ··· ··· ··· ··· · · ·⎟ ⎜ ⎟
A=⎜ ⎟ , B = ⎜· · ·⎟ (11.25)
⎜· · · ··· ··· ··· ··· · · ·⎟ ⎜· · ·⎟
⎜ ⎟ ⎜ ⎟
⎝0 0 0 ··· ··· 1⎠ ⎝0⎠
0 0 0 ··· ··· 0 1

and K = [kn , kn−1 , . . . , k2 , k1 ]T . As explained in the previous chapters, the control

signal u c is an auxiliary control term, used for the compensation of d̃ and w, which
can be selected according to H∞ control theory:

u c = − r1 B T Pe (11.26)

11.2.5 Approximators of Unknown System Dynamics

For approximating the unknown dynamics of the chaotic system, one should follow
the procedure that was analyzed in Chap. 3. The approximation of functions f (x, t)
and g(x, t) of Eq. (3.5) can be carried out with neuro-fuzzy networks (Fig. 3.1). To
this end, the weights of the neural networks are trained with the use of gradient-type
algorithms [31, 414]. The estimation of f (x, t) and g(x, t) can be written as [544]:

fˆ(x|θ f ) = θ Tf φ(x), ĝ(x|θg ) = θgT φ(x), (11.27)

where φ(x) are kernel functions with elements

n
i=1 μ A (xi )
l
φ l (x) =  N n i l l = 1, 2, . . . , N (11.28)
l=1 i=1 μ A (xi )
i

It is assumed that the weights θ f and θg vary in the bounded areas Mθ f and Mθg
which are defined as
588 11 Differential Flatness Theory for Chaotic Dynamical Systems

Mθ f = {θ f ∈ R h : ||θ f || ≤ m θ f },
(11.29)
Mθg = {θg ∈ R h : ||θg || ≤ m θg }

with m θ f and m θg positive constants.

The values of θ f and θg that give optimal approximation are

θ ∗f = arg minθ f ∈Mθ [supx∈Ux | f (x) − fˆ(x|θ f )|]

f (11.30)
θg∗ = arg minθg ∈Mθg [supx∈Ux |g(x) − ĝ(x|θg )|]

The approximation error of f (x, t) and g(x, t) is given by

w = [ fˆ(x, |θ ∗f ) − f (x, t)] + [ĝ(x|θ ∗f ) − g(x, t)]u ⇒

w = {[ fˆ(x|θ ∗f ) − fˆ(x|θ f )] + [ fˆ(x|θ f ) − f (x, t)]}+ (11.31)
+{[ĝ(x|θg∗ ) − ĝ(x|θg )] + [ĝ(x|θg ) − g(x, t)]u}

where: (i) fˆ(x|θ ∗f ) is the approximation of f for the best estimation θ ∗f of the weights’
vector θ f , (ii) ĝ(x|θg∗ ) is the approximation of g for the best estimation θg∗ of the
weights’ vector θg .
The approximation error w can be decomposed into wa and wb , where

wa = [ fˆ(x|θ f ) − fˆ(x|θ ∗f )] + [ĝ(x|θg ) − ĝ(x|θg∗ )]u

(11.32)
wb = [ fˆ(x|θ ∗f ) − f (x, t)] + [ĝ(x|θg∗ ) − g(x, t)]u

Finally, the following two parameters are defined:

θ̃ f = θ f − θ ∗f
(11.33)
θ̃g = θg − θg∗ .

11.2.6 Lyapunov Stability Analysis

The stability analysis for the adaptive fuzzy control loop of the chaotic dynamical
system, follows the stages which were explained in Chap. 3. The adaptation law of
the weights θ f and θg as well as of the supervisory control term u c is derived by the
requirement for negative definite derivative of the quadratic Lyapunov function

1 T 1 T 1 T
V = e Pe + θ̃ f θ̃ f + θ̃ θ̃g (11.34)
2 2γ1 2γ2 g
11.2 Flatness-Based Control of Chaotic Dynamical Systems 589

Substituting Eq. (11.24) into Eq. (11.34) and differentiating gives

1 T ˙ 1 T ˙
V̇ = 21 ė T Pe + 21 e T P ė + γ1 θ̃ f θ̃ f + γ2 θ̃g θ̃g ⇒
(11.35)
V̇ = 21 e T {(A − B K T )T P + P(A − B K T )}e+
+B T Pe(u c + w + d1 ) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙ g

Assumption 1: For given positive definite matrix Q and coefficients r and ρ there
exists a positive definite matrix P, which is the solution of the following matrix
equation

T 2 1
(A − B K T ) P + P(A − B K T ) − P B( − 2 )B T P + Q = 0 (11.36)
r ρ

Substituting Eq. (11.36) into V̇ yields after some operations

V̇ = − 21 e T Qe + 21 e T P B( r2 − 1
ρ2
)B T Pe + B T Pe(− r1 e T P B)+
(11.37)
+B T Pe(w + d1 ) + γ11 θ̃ Tf θ̃˙ f + γ12 θ̃gT θ̃˙g

It holds that
θ̃˙ f = θ̇ f − θ˙∗f = θ̇ f
(11.38)
θ̃˙g = θ̇g − θ˙g∗ = θ̇g

The following weight adaptation laws are considered [410]

−γ1 e T P Bφ(x) if ||θ f || < m θ f
θ̇ f = (11.39)
0 ||θ f || ≥ m θ f

−γ2 e T P Bφ(x)u c if ||θg || < m θg
θ̇g = (11.40)
0 ||θg || ≥ m θg

θ̇ f and θ̇g are set to 0, when

||θ f || ≥ m θ f , ||θg || ≥ m θg . (11.41)

The update of θ f stands for a gradient algorithm on the cost function 21 ( f − fˆ)2 .
The update of θg is also of the gradient type, while u c implicitly tunes the adaptation
gain γ2 . Substituting Eqs. (11.39) and (11.40) in V̇ finally gives
590 11 Differential Flatness Theory for Chaotic Dynamical Systems

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PB(w + d1 )−
−e T PB(θ f − θ ∗f )T φ(x) − e T PB(θg − θg∗ )T φ(x)u c ⇒ (11.42)

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PB(w + d1 ) + e T PBwα .

Denoting w1 = w + d1 + wα one gets

V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + e T PBw1 or equivalently,
(11.43)
V̇ = − 21 e T Qe − 1 T
2ρ 2
e PBB T Pe + 21 e T PBw1 + 21 w1T B T Pe

Lemma: The following inequality holds that:

1 T
2e P Bw1 + 21 w1T B T Pe − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (11.44)

Proof : The binomial (ρa − ρ1 b)2 ≥ 0 is considered. Expanding the left part of the
above inequality, one gets

ρ2a2 + 1 2
ρ2
b − 2ab ≥ 0 ⇒ 21 ρ 2 a 2 + 2ρ1 2 b2 − ab ≥ 0 ⇒
(11.45)
ab − 1 2
2ρ 2
b ≤ 21 ρ 2 a 2 ⇒ 21 ab + 21 ab − 2ρ1 2 b2 ≤ 21 ρ 2 a 2

The following substitutions are carried out: a = w1 and b = e T PB and the previous
relation becomes
1 T
2 w1 B T Pe + 21 e T P Bw1 − 1 T
2ρ 2
e PBB T Pe ≤ 21 ρ 2 w1T w1 (11.46)

The previous inequality is used in V̇ , and the right part of the associated inequality
is enforced
1 1
V̇ ≤ − e T Qe + ρ 2 w1T w1 (11.47)
2 2
Equation (11.47) can be used to show that the H∞ performance criterion is satisfied.
The integration of V̇ from 0 to T gives
T T T
0 V̇ (t)dt ≤ − 21 0 ||e||2Q dt + 21 ρ 2 0 ||w1 ||2 dt ⇒
(11.48)
T T
2V (T ) + 0 ||e|| Q dt
2 ≤ 2V (0) + ρ 2 0 ||w1 ||2 dt

Moreover, if there exists a positive constant Mw > 0 such that

∞
0 ||w1 ||2 dt ≤ Mw (11.49)
11.2 Flatness-Based Control of Chaotic Dynamical Systems 591

then one gets ∞

0 ||e||2Q dt ≤ 2V (0) + ρ 2 Mw (11.50)
∞
Thus, the integral 0 ||e||2Q dt is bounded. Moreover, V (T ) is bounded and from the
definition of the Lyapunov function V in Eq. (11.34), it becomes clear that e(t) will
be also bounded since e(t) ∈ Ωe = {e|e T Pe≤2V (0) + ρ 2 Mw }.
According to the above and using Barbalat’s Lemma, one obtains limt→∞
e(t) = 0.

11.2.7 Nonlinear Feedback Control of Chaotic Systems Based

on Fuzzy Local Linearization

The approach of differential flatness theory-based adaptive fuzzy control of chaotic

systems is compared to the control of such systems with the use of fuzzy local
linearization (Takagi-Sugeno fuzzy modelling) and the solution of Linear Matrix
Inequalities (LMIs) [568]. It is shown that the differential flatness theory-based adap-
tive fuzzy control is more efficient because it is a control method based on exact global
linearization of the chaotic oscillators’ dynamics, and thus avoids numerical approx-
imators and truncation of specific terms in the linearized equivalent model of the
system. Besides by succeeding the H∞ tracking performance criterion, it provides
improved robustness to modeling uncertainties and external disturbances.
After linearization round local operating points, the initial nonlinear dynamical
system of the form
ẋ = f (x, u) (11.51)

is written in the form of Takagi-Sugeno fuzzy model, that is [145]

Rule i : IF x1 (t) is M1i AND x2 (t) is M2i AND · · · AND xn (t) is Mni
(11.52)
THEN δx(t) = Ai x(t) + Bi u(t) i = 1, 2, . . . , r

where x j is the jth variable of the state vector, M ij is the ith fuzzy set into which the
value range of the jth input variable is partitioned, x(t) = [x1 (t), . . . , xn (t)]T in R n
is the state vector, u(t) = [u 1 (t), . . . , u m (t)]T ∈ R m is the input vector, while it
holds Ai ∈ R n×n and Bi ∈ R n×m . It is noted that the model described in Eq. (11.52)
includes also the autonomous system dynamics, that is when u(t) = 0.
The output of the Takagi-Sugeno fuzzy model is
r
i=1 wi [A x(t)+Bi u(t)]
δx(t) = ri (11.53)
i=1 wi

where r
wi = k=1 μ K (x k (t))
i (11.54)
592 11 Differential Flatness Theory for Chaotic Dynamical Systems

A condition that assures that the aforementioned system remains stable is stated as
follows [145]:
Condition 1: The equilibrium of an autonomous Takagi-Sugeno fuzzy model is glob-
ally asymptotically stable if there exists a common symmetric positive definite matrix
P, such that
AiT P + P Ai < 0 ∀ i = 1, 2, . . . , r (11.55)

Next, stability conditions are formulated for the case of a nonautonomous system.
A feedback fuzzy controller is designed by using the antecedent part of the Takagi-
Sugeno fuzzy model of the system. The fuzzy controller consists of the following
rules
j j j
Rule j : IF x1 (t) is M1 AND x2 (t) is M2 AND · · · AND xn (t) is Mn
(11.56)
THEN u(t) = K j x(t) j = 1, 2, . . . , r

The aggregate control signal that is applied to the nonlinear system becomes
r
j=1 w j K j x(t)
u(t) = r (11.57)
j=1 w j

By applying the above feedback control, the output of the closed-loop system
becomes r r
j=1 wi w j (Ai +Bi K j )x(t)
δx(t) = i=1  r r
ww
(11.58)
i=1 j=1 i j

Stability conditions for the continuous-time Takagi-Sugeno fuzzy system under state
feedback can be also formulated [145].
Condition 2: The equilibrium of the fuzzy system is globally asymptotically stable
if there exists a common symmetric positive definite matrix P such that

(Ai + Bi K j )T P + P(Ai + Bi K j ) < 0 ∀ i, j = 1, 2, . . . , r (11.59)

Finally, a less conservative condition about the stability of the continuous-time

Takagi-Sugeno system is formulated as follows [145].
Condition 3: The equilibrium state for the continuous-time fuzzy system is globally
asymptotically stable if there exists a common positive definite matrix P such that

(Ai + Bi K i )T P + P(Ai + Bi K i ) < 0, i = 1, 2, . . . , r

(11.60)
(G iTj P + P G i j ) < 0, i < j≤r

where
(Ai +Bi K j )+(A j +B j K i )
Gi j = 2 , i, j≤r (11.61)
11.2 Flatness-Based Control of Chaotic Dynamical Systems 593

11.2.8 Simulation Tests

11.2.8.1 Flatness-Based Control of Chaotic Dynamics

Control of the Lorenz chaotic system is considered as a case study. Among other
application examples, the Lorenz system dynamics is met in chaotic fluid flow mod-
els. The control input is ρ = u. The phase diagrams of the Lorenz system when
constant control input is applied are shown in Fig. 11.1. For the implementation of
the differential flatness theory-based adaptive fuzzy control, one has a fuzzy rule
base of the form

R l : IF x1 is M1l AND x2 is M2l AND x3 is M3l THEN fˆl is bl (11.62)

Regarding the implementation of the flatness-based adaptive fuzzy controller, the

neuro-fuzzy approximators for functions f (x, t) and g(x, t) have now three inputs,
i.e., x1 , x2 , and x3 . Taking that each fuzzy input variable consists of 3 fuzzy sets,
there are now 27 fuzzy rules.
The performance of flatness-based control of the chaotic system was compared
against feedback control that was based on describing the system’s dynamics through
local models and on solving LMIs. The nonlinear dynamics of the Lorenz system
was described in Eq. (11.1) and the associated state vector was x = [x1 , x2 , x3 ] =
[x, y, z]. The initial nonlinear model of the Lorenz system is in the form ẋ = f (x) +
g(x)u. The nonlinear system was decomposed into four local models

ẋ = Ai x + Bi u i = 1, . . . , 4
⎛ ⎞ ⎛ ⎞
0 σ −σ 0 (11.63)
Ai = ⎝−x3 −1 −di ⎠ Bi = ⎝di ⎠
x2 di −β 0

where di are specific points in the range of variation of the state variable x1 = x.
Matrix Ai if the Jacobian Jx f (x) computed at the points di i = 1, . . . , 4. Each
linearization point di is associated with a fuzzy set Mi which is centered at di and
which has spread equal to vi . The Takagi-Sugeno model of the Lorenz system takes
the form 4
μ M (x)Ai x+Bi u
ẋ = i=14 i (11.64)
i=1 μ Mi (x)

The results about the performance of the feedback control loop are depicted in
Figs. 11.3, 11.4, 11.5, 11.6, and 11.7. It can be noticed, that differential flatness
theory-based adaptive fuzzy control, succeeds fast and accurate tracking of the refer-
ence setpoints. The global linearization features of differential flatness theory enables
transformation of the nonlinear system dynamics to an accurate equivalent linear
model, thus also permitting the design of a more efficient state feedback controller.
594 11 Differential Flatness Theory for Chaotic Dynamical Systems

(a) 10
x1 (b) 10

x1
5 5

0 0
0 5 10 15 20 0 5 10 15 20
time time
20 20
x2

x2
0 0

−20 −20
0 5 10 15 20 0 5 10 15 20
time time
80 80
x3

x3
60 60

40 40
0 5 10 15 20 0 5 10 15 20
time time

Fig. 11.3 Tracking of setpoint 1 from state variables xi , i = 1, . . . , 3, a Adaptive fuzzy control
of the Lorenz system, b Takagi-Sugeno model-based feedback control of the Lorenz system

(a) 10 (b) 10
5 5
x1
x1

0 0
0 5 10 15 20 0 5 10 15 20
time time
20 20
0 0
x2

x2

−20 −20
0 5 10 15 20 0 5 10 15 20
time time

50 50
x3
x3

0 0
0 5 10 15 20 0 5 10 15 20
time time

Fig. 11.4 Tracking of setpoint 2 from state variables xi , i = 1, . . . , 3, a Adaptive fuzzy control
of the Lorenz system, b Takagi-Sugeno model-based feedback control of the Lorenz system

It is also pointed out that unlike Takagi-Sugeno model-based control, the adaptive
fuzzy control based on differential flatness theory made no use of the mathematical
model of the chaotic dynamical system (or of the associated linear local models).
Although no knowledge about the model parameters was used, the differential flatness
theory-based adaptive fuzzy control scheme succeeded better tracking performance.
11.2 Flatness-Based Control of Chaotic Dynamical Systems 595

(a) 10 (b) 10

x1
5 5
x1

0 0
0 5 10 15 20 0 5 10 15 20
time time
20 20

x2
x2

0 0

−20 −20
0 5 10 15 20 0 5 10 15 20
time time

50 50

x3
x3

0 0
0 5 10 15 20 0 5 10 15 20
time time

Fig. 11.5 Tracking of setpoint 3 from state variables xi , i = 1, . . . , 3, a Adaptive fuzzy control
of the Lorenz system, b Takagi-Sugeno model-based feedback control of the Lorenz system

(a) 8 (b) 10
x1

x1

6
5
4
0 5 10 15 20 0 5 10 15 20
time time
50 50

0
x2

0
x2

−50 −50
0 5 10 15 20 0 5 10 15 20
time time
80 80
60 60
x3

x3

40 40
20 20
0 5 10 15 20 0 5 10 15 20
time time

Fig. 11.6 Tracking of setpoint 4 from state variables xi , i = 1, . . . , 3, a Adaptive fuzzy control
of the Lorenz system, b Takagi-Sugeno model-based feedback control of the Lorenz system
596 11 Differential Flatness Theory for Chaotic Dynamical Systems

(a) 10 (b) 10

x1
5 5
x1

0 0
0 5 10 15 20 0 5 10 15 20
time time
50 50

x2
0 0
x2

−50 −50
0 5 10 15 20 0 5 10 15 20
time time
100 100

x3
x3

50 50

0 0
0 5 10 15 20 0 5 10 15 20
time time

Fig. 11.7 Tracking of setpoint 5 from state variables xi , i = 1, . . . , 3, a Adaptive fuzzy control
of the Lorenz system, b Takagi-Sugeno model-based feedback control of the Lorenz system

11.3 Differential Flatness Theory for Chaos-Based

Communication Systems

11.3.1 Overview

In this section, nonlinear filtering based on differential flatness theory is applied

to chaotic communication systems. There are specific advantages from the imple-
mentation of chaos-based communication, which belongs to the spread spectrum
communication schemes. Such a type of communication systems is more robust to
interferences and multipath effects, uses signals with lower value of the average spec-
tral density and is suitable for multiple access systems where shared communication
bandwidth is needed. Compared to other spread spectrum systems, chaotic systems
are characterized by higher throughput (as a result of higher SNR) or a lower power
spectral density for the same data throughput (this increases spectral reuse) [24, 25,
80, 142, 529, 548]. A main difference between traditional direct sequence spread
spectrum communication and a coherent chaotic sequence spread spectrum commu-
nication system is the absence of apparent periodicity in the chaotic waveform [127,
147, 158, 294, 604].
In this section, a communication system based on chaotic modulation and demod-
ulation is developed. In the transmitter’s side, the source of information undergoes
modulation in which a chaotic signal (Duffing system) is the carrier. The modulated
signal is transmitted through a communication channel, under the effects of Gaussian
noise. At the receiver’s side, demodulation of the transmitted signal takes place, by
exploiting the estimation provided by the Kalman Filter for the state vector of the
chaotic oscillator [143, 213, 283, 395, 549, 571]. Other results on observer-based
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 597

techniques and Kalman Filtering methods in chaotic communication systems can

be found in [243], where the Extended Fractional Kalman filter is used for retrieval
of the information signal at the receiver’s side and in [241] where a state observer
is applied for demodulation and extraction of the information signal. In [403], an
adaptive sliding mode observer is used for transmitter–receiver synchronization in a
chaotic communication system. Moreover, in [367] a method based on state observer
is proposed for constructing a chaotically synchronized system that is based on the
Lorenz oscillator. Furthermore, in [109, 303, 376] the use of the Extended Kalman
Filter and of the Unscented Kalman Filter, respectively, has been proposed for chaotic
communication systems.
The considered Kalman Filtering, which is the previously analyzed Derivative-
free nonlinear Kalman Filtering, makes use of a linearized model of the chaotic
oscillator that is obtained after transforming the oscillator’s dynamic model into a
linear canonical form through the application of differential flatness theory [424, 427,
429, 434, 446, 455]. To conclude if a potential chaotic modulator is differentially
flat, the following should be examined: (i) The existence of the flat output, which is
a variable that can be written as a function of the system’s state variables and (ii) the
system’s state variables and the input can be written as functions of the flat output
and its derivatives. As explained in previous sections, nonlinear systems satisfying
the differential flatness property can be written in the linear canonical (Brunovsky)
form via a transformation of their state variables and control inputs [152, 263, 286,
340, 465, 495, 535].
On the other hand, it is known that the Extended Kalman Filter is a widely used
nonlinear estimation method that has been extensively applied to state estimation
and signal recovery in communication systems [31, 192, 229, 372, 573]. Thus, one
could have also attempted estimation with the use of the Extended Kalman Filter;
however, this would require the computation of Jacobian matrices and would intro-
duce cumulative errors to the estimation procedure due to approximate linearization
(caused by truncation of higher order terms in Taylor series expansion) [405, 412,
422, 438]. Unlike this, with the proposed estimation scheme, i.e., the Derivative-free
nonlinear Kalman Filter, exact linearization is succeeded through the transformation
of the modulator’s dynamic model into the linear canonical form through a change
of coordinates (diffeomorphism). For the linearized equivalent of the communica-
tion system, it becomes possible to apply the standard Kalman Filter recursion. The
Derivative-free nonlinear Kalman Filter provides estimation of improved accuracy
comparing to the Extended Kalman Filter, while it is also computationally simpler
and faster. Thus it gives an efficient real-time solution to the problem of synchroniza-
tion between the transmitter and the receiver and to the problem of demodulation in
chaotic communication systems.
Another issue that is treated in this section is that of joint synchronization and
channel equalization for the chaotic communication system [208, 218, 536, 617].
To compensate for uncertainties and variations of the communication channel (e.g.,
due to fading and multipath effects), a second Kalman Filter is used in parallel to
the proposed Derivative-free nonlinear Kalman Filter, providing estimates of the
coefficients of the channel’s model. The Derivative-free nonlinear Kalman Filter
598 11 Differential Flatness Theory for Chaotic Dynamical Systems

computes estimates of the state vector of the chaotic system at the receiver’s side
using for this the estimates of the channel’s AR (auto-regressive) coefficients which
are provided by the second Kalman Filter algorithm. In a similar manner, the second
Kalman Filter algorithm provides estimates of the channel’s coefficients using in
the associated computation, the estimates of the state vector of the chaotic system
which are generated by the Derivative-free nonlinear Kalman Filter. Considering a
distorted channel dynamics that is described by a first-order AR model, it is shown
that the proposed synchronization and equalization scheme (blind equalization) can
improve the performance of the chaotic communication system.

11.3.2 Structure of the Chaotic Communication System

11.3.2.1 Modeling of the Communication System

Chaotic communication systems are based on the use of a chaotic signal that modu-
lates (encrypts), the information signal to be finally transmitted over the communi-
cation channel. At the receiver’s side, demodulation takes place which is based on
the extraction of the information signal out of the modulated signal (see Fig. 11.8).
To this end, a chaotic carrier signal at the receiver’s side has to be replicated, and this

Transmitter

Information Chaotic Power

Source Modulation Amplifier

Chaotic
Dynamical
System

Received Chaotic Power

Information Demodulation Amplifier

Receiver

Kalman Filter
Based
Estimator

Fig. 11.8 Communication scheme based on chaotic modulation (encryption) and demodulation
with the use of Kalman Filtering
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 599

should follow with precision the chaotic carrier signal at the transmitter’s side even
if starting from different initial conditions. State estimation methods and filtering are
the recommended approaches for reconstructing the chaotic carrier at the receiver’s
side. The quality of communication is measured in terms of bit error rate (BER) for
digital signal transmission or in terms of mean square error (MSE) in case of analog
signal transmission.

11.3.2.2 Chaotic Oscillators for Modulation in Chaotic

Communication Systems

Chaotic systems exhibit dynamics which are highly sensitive to initial conditions.
Since future dynamics are dependent on initial conditions the evolution in time of
chaotic systems appears to be random. The output or state variable of a chaotic
oscillator can be used for the modulation (encryption) of an information signal and
for its secure transmission through a communication channel. Main types of chaotic
modulators have been described in the previous section (e.g., the Lorenz system, the
Duffing chaotic system, the Genesio-Tesi chaotic system, and the Chen chaotic sys-
tem).
As mentioned before dynamical systems that exhibit chaotic behavior are Chua’s
circuit and Rössler’s system. It has been shown that for chaotic dynamical systems
exact linearization is possible, through the application of a diffeomorphism that
enables their description in new state-space coordinates. As example, the Duffing
oscillator will be used, while it is straightforward to apply the method to other types
of chaotic oscillators.
The structure of the chaotic communication system is depicted in Fig. 11.9. At
the transmitter’s side, the chaotic oscillator operates. The second-state variable of
the oscillator x2 (t) is used for the modulation (encryption) of the information sig-
nal s1 (t). The modulated (encrypted) signal is transmitted through the communi-
cation channel, under Gaussian noise. Due to the chaotic waveform of this signal,

Fig. 11.9 Diagram of the Derivative-free Kalman Filter-based chaotic communication system
600 11 Differential Flatness Theory for Chaotic Dynamical Systems

interception and acquisition of the information signal becomes difficult. The pattern
of the transmitted signal depends on the type of the chaotic oscillator, the initial
conditions of the chaotic oscillator, and on the parameters of the dynamic model of
the oscillator; therefore, tracking of the oscillator’s state variables by the interceptor
and extraction of the information signal out of the modulated signal is a cumber-
some procedure. The first-state variable of the chaotic oscillator is also transmitted
through the communication channel and is used for synchronization between the
transmitter and the receiver. This means that by processing the sequence of samples
of x1 (t), and by knowing the oscillator’s type and parameters at the receiver’s side,
it will become possible to reconstruct with precision the complete state vector of the
oscillator at the receiver’s side. The key element for retrieving the information signal
at the receiver is demodulation (decryption) and this is performed with the use of
a nonlinear Filtering algorithm, such as the Extended Kalman Filter, the Unscented
Kalman Filter, or the Derivative-free nonlinear Kalman Filter. The nonlinear filter
succeeds accurate estimation of the state variables of the chaotic oscillator (modu-
lator), despite the effects of noise in the communication channel. By knowing x̂2 (t)
i.e., an estimate of x2 (t) which had been used for modulation it is possible to retrieve
without error the information signal ŝ1 (t). The quality of communication depends on
the accuracy of demodulation, i.e., on the precision of the filtering method used for
this task. It will be shown that demodulation based on the Derivative-free nonlinear
Kalman Filter outperforms the demodulation that is based on the Extended and the
Unscented Kalman Filter.

11.3.3 Differential Flatness Theory

11.3.3.1 Differential Flatness Theory for Precise State Estimation

Aiming at finding more efficient state estimation methods (comparing to the Extended
Kalman Filter) for demodulating (decrypting) the signal at the receiver’s side, in this
section the derivative-free approach to nonlinear Kalman filtering is introduced. In
the proposed Derivative-free Kalman Filtering method, the model of the chaotic
oscillator that operates at the transmitter’s side is first subjected to a linearization
transformation that is based on the differential flatness theory and next state estima-
tion is performed by applying the standard Kalman Filter recursion to the linearized
model. Unlike the Extended Kalman Filtering approach, the proposed method pro-
vides estimates of the state vector of the chaotic oscillator without the need for
derivatives and Jacobians calculation [424, 429]. By avoiding linearization approx-
imations, the proposed filtering method improves the precision in the estimation of
the oscillator’s state variables, and results in accurate extraction of the information
signal out of the modulated signal that reaches the receiver. Thus, occurrence of
transmission errors in the chaotic communication system is reduced and the quality
of communication is improved.
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 601

This section is concerned with chaotic oscillators used in the modulation part of
chaotic communication systems and the resulting description of such oscillators in
the Brunovksy (canonical) form [400]. At a second stage, and with the use of the
oscillators’ exactly linearized model, Kalman Filtering is proposed for estimating the
nondirectly measurable elements of the state vector of the chaotic oscillators. Actu-
ally, by using the description of the oscillator’s dynamic model in the linear canonical
form it is possible to perform state estimation with the standard Kalman Filter recur-
sion (Derivative-free nonlinear Kalman Filtering) and to avoid the approximative
computations and the cumulative linearization errors of the Extended Kalman Filter.

11.3.4 Estimation in Chaotic Modulators with Nonlinear

Kalman Filter

11.3.4.1 Extended Kalman Filtering for the Duffing Oscillator

To apply Extended Kalman Filtering to the model of the Duffing oscillator, given in
Eq. (11.13), one has to compute the following Jacobian matrices:
 
0 1
Jφ =
1.1 − 3x12 −0.4 (11.65)
 
Jγ = 1 0

For the implementation of the Extended Kalman Filter recursion in discrete time, one
has to substitute the previously computed matrix Jφ with its discrete-time equivalent
i.e., I +Ts Jφ where I ∈R 2 is the identity matrix. In a similar manner one can transform
in the linear canonical form the Genesio-Tesi or the Chen system, as well as other
nonlinear chaotic oscillators.

11.3.4.2 Derivative-Free Nonlinear Kalman Filtering for the Duffing

Oscillator

In case that the Derivative-free nonlinear Kalman Filtering approach is applied to

the model of the Duffing chaotic oscillator one can define as flat output y = x1 . It
holds that all system state variables and the control input can be written as functions
of the flat output and its derivatives. Indeed x1 = y, x2 = ẏ, u(t) = 1.8cos(1.8t) =
ÿ − 1.1y + y 3 − 0.4 ẏ. The new state variables of the system are defined as y1 = y
and y2 = ẏ and the new control input is defined as v = ÿ − 1.1y + y 3 − 0.4 ẏ or
v = ẋ2 − 1.1x1 + x13 − 0.4x2 . Consequently, a description of the system in the linear
canonical (Brunovksy) form is obtained
602 11 Differential Flatness Theory for Chaotic Dynamical Systems

ẏ1 (t) = y2 (t)

(11.66)
ẏ2 (t) = v

where v is as given above and the measurable output is y = x1 . The state-space

equations of oscillator’s model, take the form of Eq. (11.14), i.e.,
      
ẏ1 01 y1 0
= + v
ẏ2 00 y2 1
  (11.67)
  y1
z= 10
y2

For the canonical form model of the Duffing oscillator, one can perform the standard
recursion of the Kalman Filter using y1 (i.e., x1 ) as measured output.
The assessment of nonlinear filters for demodulation in chaotic communication
systems should be performed against two criteria: (i) Accuracy of estimation (trans-
mission error) and (ii) speed of computation (computational complexity). Regarding
(i), it has been shown that the Derivative-free nonlinear Kalman Filter (DKF) outper-
forms the Extended Kalman Filter. Actually, in the range 15–35 dB the transmission
error of the chaotic communication system that uses the DKF for retrieving the infor-
mation signal at the transmitter’s side is improved by 20 % comparing to the chaotic
communication system that uses the EKF. The accuracy of estimation of the DKF is
also improved with respect to the Unscented Kalman Filter (see Refs. [434, 437]).
Regarding (ii), the Derivative-free nonlinear Kalman Filter outperforms in terms of
speed of computation all other nonlinear filters. Thus, the DKF is faster than the
Extended Kalman Filter because it does not need to compute in the filter’s recur-
sion Jacobian matrices; whereas, it is also faster than the Unscented Kalman Filter
because it does not need to compute cross-covariance matrices. It has been observed
that for dynamical systems of order as the one of the chaotic oscillators treated in
the chapter, the Derivative-free nonlinear Kalman Filter is about 20 % faster than the
Extended Kalman Filter (see Refs. [434, 436]).

11.3.5 Channel Equalization and Synchronization Using Dual

Kalman Filtering

The design of the chaotic communication system given in the previous sections
assumed that the transmitter was connected to the receiver through an ideal channel.
However, this assumption does not hold always, and due for example to multipath
propagation effects distortions can appear. Multipath propagation results in a received
signal that is superposition of several delayed and scaled copies of the transmitted
signal and causes high error rate in symbol detection.
Multipath effects may appear in either time-varying or time-invariant channels.
A wireline channel is usually considered to be time-invariant (the coefficients of its
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 603

model do not change in time or vary slowly). That is the case of ATM LANs and
telephone networks (xDSL standards). In wireless communications, the channel is
more likely to a time-variant one (its coefficients change in time). The time-varying
coefficients are usually represented by uncorrelated stationary random processes. In
certain cases, they are taken to be Gaussianly distributed with zero mean (Rayleigh
fading) or nonzero mean (Rician fading), depending on whether line-of-sight prop-
agation takes place or not [617].
Compensation of these types of distortions is possible if estimation is performed
about the unknown coefficients that describe the commonly used AR model of the
communication channel, that is
p
z(k) = i=1 ci (k)x(k − i + 1) + n(k) (11.68)

where x(k) and z(k) denote the kth sample of the transmitted and received signals,
ci (k) are the channel coefficients, and n(k) represents additive measurement noise.
Such an estimation procedure, also known as equalization, can be carried out in the
least squares sense if the sequences of x(k) and y(k) are available. However, during
online transmission the sequence x(k) is not supplied by the transmitter. Therefore,
estimation of ci (k) has to be performed without knowledge of the sequence x(k),
and this procedure is known as blind equalization.
The dynamics of the chaotic system after transformation to the canonical form
and after discretization its discrete-time description becomes

y f (k + 1) = Adf y f (k) + B df v(k)

(11.69)
z f (k) = C df y f (k)

The above equation holds for an ideal communication channel. However, for the
distorted communication channel (e.g., due to multipath and interference effects) the
state vector dynamics and the measured output at the receiver’s side becomes

y f (k + 1) = Adf y f (k) + B df v(k) + n y (k)

p (11.70)
z f (k) = i=1 ci (k)y f (k − i + 1) + n 1 (k)

where n y (k), n 1 (k) stand for the noise sequences, as shown in Fig. 11.10. The coef-
ficients of the channel ci (k) are unknown . In several cases these coefficients can be
considered to be constant or piecewise constant (e.g., in wireline communication);
whereas in the case of wireless communication, it is likely that they vary in time
according to an AR model [617].
One has to handle the dual problem of simultaneous estimation of the state vector
of the chaotic system (transmitter-receiver synchronization) and estimation of the
unknown channel coefficients (equalization). This problem can be solved efficiently
within a dual Kalman Filtering scheme. A first Kalman Filter is designed to solve the
estimation problem for the chaotic system thus providing state vector estimates x̂(k).
A second Kalman Filter is designed to solve the estimation problem for the channel
604 11 Differential Flatness Theory for Chaotic Dynamical Systems

Fig. 11.10 Chaotic communication system over a channel with distortions

Fig. 11.11 Joint estimation

of the state vector of the
chaotic modulator and of the
unknown coefficients of the
communication channel
using dual Kalman filtering

thus providing the coefficients’ estimates ĉ(k). The estimates of the first filter are
used in the computation of the second filter and inversely, as shown in Fig. 11.11.
Considering that the coefficients’ vector is piecewise constant, one can use the
following state-Takagi-Sugeno space description for the vector of the channel’s coef-
ficients
 p = c(k − 1) + n c (k)
c(k)
(11.71)
z f (k) = i=1 ci (k)y f (k − i + 1) + n 1 (k)

where n c (k), n 1 (k) stand again for the noise sequences. In Eq. (11.70) the measure-
ment equation can be written in the vectors’ product form z f (k) = c(k)·y f (k)T +
n 1 (k); whereas in Eq. (11.71), the measurement equation can be written in the vec-
tors’product form z f (k) = y f (k)·c(k)T + n 1 (k).
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 605

Moreover, according to Fig. 11.10, after transmission through the distorted chan-
nel the following sequence of measurements of the modulated information signal
arrives at the receiver
p
z m (k) = i=1 ci (k)[x2 (k − i + 1) + s1 (k − i + 1)] + n 2 (k) or
p (11.72)
z m (k) = i=1 ci (k)w(k − i + 1) + n 2 (k)

From the measurements of z m (k) and the estimates of the state vector variable x̂2 (k −
i +1), i = 1· · · p, as well as the estimates of the channel coefficients ĉi (k) it becomes
possible to solve Eq. (11.72), recursively, with respect to s1 (k) and thus to obtain at
the receiver estimates of the information signal ŝ1 (k).

11.3.6 Simulation Tests

11.3.6.1 Testing of the Kalman Filter-Based Chaotic

Communication Scheme

An ideal communication channel is assumed first. The chaotic communication sys-

tem that is based on the Derivative-free nonlinear Kalman Filter is tested against
the communication system that uses at the transmitter’s side the Extended Kalman
Filter. To encrypt the information signal s1 (t), a new signal is generated through the
addition of the second-state variable of the chaotic oscillator x2 (t) (Duffing oscil-
lator). Thus the transmitted signal is w(t) = s1 (t) + x2 (t). This is sent through the
communication channel, under noise n(t). Signal w(t) exhibits a chaotic evolution
in time, which depends on the initial conditions of the chaotic oscillator at the trans-
mitter’s side, on the parameters of the chaotic oscillator as well as on the parameters
defining the oscillator’s input u(t). For the synchronization between the transmitter
and the receiver the first-state variable of the chaotic system x1 (t) is used. With-
out dependence on the initial conditions x̂i (0) i = 1, 2 the state variables of the
stochastic filter at the receiver’s side (Extended Kalman Filter, Unscented Kalman
Filter or Derivative-free nonlinear Kalman Filter) converge asymptotically to the
state variables of the chaotic oscillator at the transmitter’s side. Thus, to retrieve the
information signal s1 (t) from the modulated (encrypted) signal w(t), the following
computation has to be performed ŝ1 (t) = w(t) − x̂2 (t). In place of additive modu-
lation (encryption), multiplicative modulation of the information signal could have
been also used. The more accurate the estimate x̂2 (t) is, the better the reconstruction
of the information signal ŝ1 (t) is at the receiver’s side. The quality of the established
communication can be measured with the use of the bit error rate (BER) for the case
of digital signal transmission and with the use of the mean square error (MSE) in the
case of analog signal transmission.
In the performed evaluation tests, the information signal was taken to be a sinu-
soidal signal s1 (t) = 0.2sin(2π t). The parameters of the chaotic modulator (Duffing
606 11 Differential Flatness Theory for Chaotic Dynamical Systems

(a) 0.5 (b) 0.5

s1
s1

0 0

−0.5 −0.5
0 5 10 15 20 2 4 6 8 10 12 14 16 18 20
t (sec) t (sec)
3 3
2 2
1 1

w1
w1

0 0
−1 −1
−2 −2
−3 −3
0 5 10 15 20 2 4 6 8 10 12 14 16 18 20
t (sec) t (sec)

Fig. 11.12 Information signal s1 and transmitted (modulated) signal w1 in case of a digital signal
transmission, b analog signal transmission

oscillator) were the ones given in Eq. (11.13). The process and measurement noise
covariance matrix used by the Extended Kalman Filter and the Derivative-free non-
linear Kalman Filter were Q = 10−4 I2×2 and R = 10−3 I1×1 . The coefficients of the
communication channel were considered to be c1 = 1 and c2 = 0, which means that
the channel was considered to be undistorted and not to introduce any time delays.
In Fig. 11.12 the initial information signal s1 (t) is depicted as well as the mod-
ulated signal that is transmitted through the communication channel, both in the
case of digital signal transmission and in the case of analog signal transmission. In
Fig. 11.13 the synchronization succeeded between the transmitter and the receiver is
shown. First, results are given about a receiver that made use of the Derivative-free
nonlinear Kalman Filter for reconstructing the state variables of the chaotic system,
while next these results are repeated for a receiver that used the Extended and the
Unscented Kalman Filter. In Fig. 11.14 the quality of the communication that is estab-
lished between the transmitter and the receiver is presented. First, the case of digital
signal transmission is considered and BER diagrams are presented for a receiver
making use of the Derivative-free nonlinear Kalman Filter (DKF), and receivers
making use of the Extended/Unscented Kalman Filter (EKF/UKF). Next the case
of analog signal transmission is considered and mean square error (MSE) diagrams
are presented for the DKF-based receiver and for EKF/UKF-based receivers. In all
cases, it is clear the Derivative-free nonlinear Kalman Filter contributes significantly
to the improvement of the quality of communication between the transmitter and
the receiver.
Digital signal transmission was based on discretization of the aforementioned
sinusoidal signal using a four-bit digital representation of each sampled value. The
quality of transmission was measured in BER for signal to noise ratio (SNR) varying
between 15 and 35 dB. It can be seen that the use of the Derivative-free nonlinear
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 607

(a) 2 (b) 2

x1 EKF UKF
1 1
x1 DKF

0 0

−1 −1

−2 −2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
4 4

x2 EKF UKF
2 2
x2 DKF

0 0

−2 −2

−4 −4
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 11.13 Synchronization of transmitter and receiver (convergence of x̂i i = 1, 2 to xi i = 1, 2)

using the a Derivative-free nonlinear Kalman Filter (blue line) b Extended Kalman Filter (green
line) and Unscented Kalman Filter (blue line)

DKF EKF UKF DKF EKF UKF

(a) 0.04
DKF (b) 0.03
EKF DKF
UKF 0.028 EKF
0.035 UKF
0.026

0.024
0.03
0.022
BER

MSE

0.025 0.02

0.018
0.02
0.016

0.014
0.015
0.012
0.01 0.01
15 20 25 30 35 15 20 25 30 35
SNR in dB SNR

Fig. 11.14 Performance of the demodulator that used the Derivative-free nonlinear Kalman Filter-
based in case of, a digital signal transmission, b analog signal transmission (blue line EKF-based
communication systems, green line DKF-based communication system)

Kalman Filter at the receiver’s side resulted in lower BER comparing to the one
succeeded with the use of Extended Kalman Filter or the Unscented Kalman Filter.
This is because the Derivative-free nonlinear Kalman Filtering is based on an exact
linearization method and avoids cumulative numerical errors. Moreover, analog sig-
nal transmission was tested in the SNR range between 15 and 35 dB. The improved
performance of the Derivative-free nonlinear Kalman Filter was again confirmed as
608 11 Differential Flatness Theory for Chaotic Dynamical Systems

DKF EKF UKF

0.7
DKF
EKF
0.6
UKF

MSE − DKF, EKF, UKF 0.5

0.4

0.3

0.2

0.1

−0.1
0 2 4 6 8 10 12 14 16 18 20
time (sec)

Fig. 11.15 Estimation error curve of the information signal for DKF, EKF, and UKF and its
convergence speed to zero

Table 11.1 Improved % estimation accuracy of DKF

SNR 10.03 13.44 16.84 21.61 28.84 34.22 43.59 49.19
(dB)
DKF– 13.49 15.75 14.87 18.35 20.67 21.08 22.17 24.39
EKF(%)
DKF– 6.14 8.92 6.69 9.48 10.19 10.17 10.32 11.88
UKF(%)

described by the mean square error (MSE) between the initial information signal
s1 (t) and its estimate at the receiver’s side ŝ1 (t).
The Derivative-free nonlinear Kalman Filter is also proven to be computation-
ally faster than the Extended Kalman Filter. This is because it does not require the
online computation of Jacobian matrices. It is also faster than the Unscented Kalman
Filter because it does need to compute cross-covariance matrices. In Fig. 11.15 it is
shown that the estimation error curve of the Derivative-free nonlinear Kalman Filter
converges faster to zero than the associated curves of EKF and UKF.
The improved performance of the DKF over EKF and UKF for different values
of the SNR, is shown in Table 11.1. This confirms that, comparing to EKF, the
Derivative-free nonlinear Kalman Filter (DKF), succeeds a more accurate estimation
of the information signal at the receiver’s side. It is also shown that the DKF has better
estimation accuracy than the one of the UKF. From Fig. 11.14a, b, as well as from
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 609

Table 11.1 it can be noticed that for SNR between 25 and 35 dB the improvement of the
performance of the chaotic communication system based on DKF is of about 20 %.

11.3.6.2 Channel Equalization and Synchronization Using a Dual

Kalman Filtering Scheme

In the results presented above the communication channel was considered to be

an ideal one, i.e., there was no fading and no multipath effects. The design of the
transmitter with use of the Derivative-free nonlinear Kalman Filter can be further
extended so as to perform not only synchronization between the transmitter and the
receiver but also adaptive channel equalization that will enable to compensate for
disturbances affecting the communication channel. Using the approach described in
Sect. 11.3.5 it has been considered that the dynamics of the communication channel
is described by the AR model of Eq. (11.68) with p = 2, and thus it comprises two
unknown coefficients ci , i = 1, 2.
In this second set of experiments, the process and measurement noise covariance
matrices for both Kalman Filters involved in the dual Kalman Filtering scheme were
Q = 10−4 I2×2 and R = 10−3 I1×1 , respectively. The communication channel was
considered to be subjected to perturbations and multipath effects and its coefficients
varied as follows: (i) In test case 1 the channels coefficients were initially [c1 , c2 ] =
[0.6, 0.5] and next changed into [c1 , c2 ] = [0.8, 0.7], (ii) in test case 2 the channels
coefficients were initially [c1 , c2 ] = [0.9, 0.8] and next changed into [c1 , c2 ] =
[0.6, 0.5].
The results shown in Figs. 11.16 and 11.17 confirm the efficiency of the pro-
posed dual Kalman Filtering scheme to perform simultaneously (i) synchronization

(a) DKF (b)

2 DKF
s1 − s1est

1
0
0.8
c1 − c1est

−2 0.6
2 4 6 8 10 12 14 16 18 20
0.4
t (sec)
0.2
EKF
2 0
s1 − s1est

2 4 6 8 10 12 14 16 18 20
0 t (sec)
−2
2 4 6 8 10 12 14 16 18 20 1
t (sec) 0.8
c2 − c2est

UKF 0.6
2
s1 − s1est

0.4
0 0.2
−2 0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
t (sec) t (sec)

Fig. 11.16 Joint channel equalization and synchronization—test 1: a information signal at the
receiver’s side after demodulation, b estimation of the unknown channel coefficients
610 11 Differential Flatness Theory for Chaotic Dynamical Systems

(a) DKF
(b)
DKF
2
s1− s1est
1
0 0.8

c1− c1est
−2 0.6
2 4 6 8 10 12 14 16 18 20
0.4
t (sec)
0.2
EKF 0
s1− s1est

2 2 4 6 8 10 12 14 16 18 20
0 t (sec)
−2
2 4 6 8 10 12 14 16 18 20 1
t (sec) 0.8

c2 − c2 est
UKF 0.6
s1− s1est

2
0.4
0 0.2
−2 0
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
t (sec) t (sec)

Fig. 11.17 Joint channel equalization and synchronization—test 2: a information signal at the
receiver’s side after demodulation, b estimation of the unknown channel coefficients

DKF EKF UKF DKF EKF UKF

(a) 0.02 DKF
(b) 0.02
DKF
EKF EKF
0.018 UKF 0.018 UKF

0.016 0.016

0.014 0.014
MSE
MSE

0.012 0.012

0.01 0.01

0.008 0.008

0.006 0.006
18 20 22 24 26 28 30 32 34 36 38 18 20 22 24 26 28 30 32 34 36
SNR SNR

Fig. 11.18 MSE of the information signal at the receiver under dual Kalman filtering, a test case
1, b test case 2

between the transmitter and the receiver and reconstruction of the information sig-
nal at the receiver’s side, and (ii) estimation of the unknown coefficients in the AR
model of the distorted communication channel (blind equalization). It can be noticed
that synchronization between the transmitter and the receiver is succeeded and that
the demodulated information signal ŝ1 (k) at the receiver converges accurately to the
transmitted information signal s1 (k). Moreover, it can be observed that the unknown
coefficients of the AR model of the communication channel ci i = 1, 2 are esti-
mated with accuracy. The proposed dual estimation scheme has fast convergence
11.3 Differential Flatness Theory for Chaos-Based Communication Systems 611

properties and can be also applied to communication channels with rapid time-
varying characteristics.
The transmission error, under a distorted communication channel is depicted
in Fig. 11.18. The diagrams correspond to the aforementioned two different cases
of parametric variations for the communication channel (changes of the channel’s
coefficients ci , i = 1, 2) and present the communication error (MSE) when the
Derivative-free nonlinear Kalman Filter and the Extended Kalman Filter are used at
the receiver’s side to separate the information signal from the chaotic carrier signal.
Chapter 12
Differential Flatness Theory for Distributed
Parameter Systems

12.1 Introduction

The previous chapters of the book have been mostly concerned with nonlinear
systems described by ordinary differential equations. Another significant class
of control problems is associated with distributed parameter systems, the latter
described by nonlinear partial differential equations. It will also be shown that differ-
ential flatness theory can be used for solving efficiently control and filtering problems
in distributed parameter systems.
First, the chapter presents a pointwise control method for a 1D nonlinear wave
equation and a filtering approach for estimating the dynamics of such a system
from measurements provided from a small number of sensors. It is shown that the
numerical solution of the associated partial differential equation results in a set of
nonlinear ordinary differential equations. With the application of a diffeomorphism
that is based on differential flatness theory it is shown that an equivalent description of
the system in the linear canonical (Brunovsky) form is obtained. This transformation
enables to obtain estimates about the state vector of the system through the application
of the Derivative-free nonlinear Kalman Filter. For the local subsystems, into which
the nonlinear wave equation is decomposed, it becomes possible to apply pointwise
state estimation-based feedback control. The efficiency of the proposed filtering
and control approach for nonlinear systems described by 1D Partial Differential
Equations of the wave type (e.g., sine-Gordon PDE) is confirmed through simulation
experiments. It is shown that, by applying feedback control, the nonlinear wave-type
dynamics can be made to track any reference setpoint.
Moreover, the chapter studies control of the nonlinear heat diffusion PDE, using as
a case study the gas metal arc welding process. Following the procedure for numerical
solution of the nonlinear heat diffusion PDE, a set of coupled nonlinear ordinary
differential equations is obtained and written in a state-space form [255, 388]. For the
latter state-space description, differential flatness properties are proven. By applying a

© Springer International Publishing Switzerland 2015 613

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_12
614 12 Differential Flatness Theory for Distributed Parameter Systems

change of coordinates (diffeomorphism) which is based on differential flatness theory

it is shown that the state-space model of the heat diffusion PDE can be written in a
linear form, in which the previously noted nonlinear ordinary differential equations
are now transformed into linear ones. Next, feedback control is applied to the heat
diffusion PDE. By computing the control input of the heat diffusion PDE, which
varies both in space and time one can also compute the velocity that the welding
torch should have at a specific time instant and point of the cartesian frame, so that
the temperature distribution of the workpiece converges to the reference setpoints.
Finally, to implement feedback control of the gas metal arc welding process using
measurements from only a small number of sensors the Derivative-free nonlinear
Kalman Filter is used.
Additionally, the chapter proposes the Derivative-free nonlinear Kalman Filter for
state estimation and fault diagnosis in distributed parameter systems and particularly
in dynamical systems described by partial differential equations of the nonlinear
wave type. At the first stage, the Derivative-free nonlinear Kalman Filter is used to
obtain estimates about the state vector of the monitored system. At the second stage,
the local statistical approach to fault diagnosis is used to perform fault diagnosis
for the distributed parameters system by processing with elaborated statistical tools
the differences (residuals) between the output of the Kalman Filter and the mea-
surements obtained from the distributed parameter system. Optimal selection of the
fault threshold is succeeded by using the local statistical approach to fault diagnosis.
The efficiency of the proposed filtering approach for performing fault diagnosis in
distributed parameters systems is confirmed through simulation experiments.
Finally, the chapter proposes structural condition monitoring for buildings and
mechanical structures using the previously analyzed Derivative-free nonlinear
Kalman Filtering. The considered filter makes use of an exact linearization trans-
formation of the structure’s dynamical model in accordance to differential flatness
theory and of an inverse transformation that enables to obtain estimates for the state
vector elements of the initial model. In this approach, a multi-DOF building or
mechanical structure is modeled as a set of coupled nonlinear oscillators that can
be subjected to external excitation. The response of the structure is recorded and
compared against the response generated by the Derivative-free nonlinear Kalman
Filter under the assumption of a damage-free model. By comparing the two sig-
nals, residuals sequences are generated. Moreover, the filter provides estimates of
the state vector elements of the structure which cannot be directly measured, while
it can also provide estimates of unknown excitation inputs. The processing of the
residuals with the use of statistical decision-making criteria provides an indication
about the existence of parametric changes (damages) in the structure which other-
wise could not have been detected. The efficiency of the proposed method is tested
through simulation experiments.
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 615

12.2 Pointwise Flatness-Based Control of Distributed

Parameter Systems

12.2.1 Overview

The problem of state estimation and control of distributed parameter systems

described by partial differential equations of the parabolic, hyperbolic, and ellip-
tic type are of advanced difficulty comparing to the control of lumped parameter
systems, the latter being modeled in the form of ordinary differential equations
[427, 453]. The problem becomes more difficult in case that the distributed para-
meter systems are characterized by nonlinearities [21, 35, 133, 187, 267, 306, 320,
547, 557, 558, 618]. This section treats the problem of piecewise control of non-
linear wave-type partial differential equations. Control of nonlinear wave dynamics
is met in several applications. For instance, nonlinear wave-type differential equa-
tions are found in communication systems (transmission lines, optical fibers, and
electromagnetic waves propagation), in electronics (Josephson junctions), in fluid
flow models, in structural engineering (dynamic analysis of buildings under seismic
waves, mechanical structures subjected to vibrations, pendulum chains), in biomed-
ical systems (voltage propagation and variations in neuron’s membrane), etc. Solving
nonlinear estimation and control problems for such systems is important for modify-
ing their dynamics and for succeeding their functioning according to specifications
[186, 253, 254, 320, 466, 518].
Following the procedure for numerical solution of the nonlinear PDE of the wave-
type dynamics, a set of coupled nonlinear ordinary differential equations is obtained
and written in a state-space form [388]. For the latter state-space description, dif-
ferential flatness properties are proven. Thus, it is shown that all state variables and
the control inputs of the state-space model can be written as functions of a vector of
algebraic variables that constitute the flat output and of the flat output’s derivatives
[16, 55, 152, 286, 362, 465]. By applying a change of coordinates (diffeomorphism)
which is based on differential flatness theory it is shown that the state-space model of
the wave-type PDE can be written in the linear canonical (Brunovsky) form, in which
the previously noted nonlinear ordinary differential equations are now transformed
into linear ones. Next, pointwise feedback control is applied to the wave-type PDE.
For each local linear model of the aforementioned differential equations the state
feedback control is selected such that asymptotic stability is assured. This can be
done using for instance pole placement methods.
Another objective of this section is to implement state feedback control of the
nonlinear wave dynamics using measurements from a small number of sensors
[335, 437]. This implies that for state vector elements of the PDE’s state-space
description which cannot be measured directly, state estimation with filtering meth-
ods has to be applied. Filtering for nonlinear distributed parameter systems is again a
nontrivial problem [42, 81, 467, 559, 592]. Both observer-based and Kalman Filter-
based approaches have been proposed [38, 81, 124, 198]. To this end, in this section,
the Derivative-free nonlinear Kalman Filtering, is used. As explained in previous
616 12 Differential Flatness Theory for Distributed Parameter Systems

chapters, the filter consists of the standard Kalman Filter recursion applied to the
linear equivalent state-space model of the wave PDE [31, 408, 414]. Moreover, an
inverse transformation which is based on differential flatness properties enables to
obtain estimates of the state variables of the initial system’s description.

12.2.2 Nonlinear 1D Wave-Type Partial

Differential Equations

12.2.2.1 Sine-Gordon Nonlinear PDE in Coupled Nonlinear Pendula

Nonlinear 1D wave-type partial differential equations appear in models of cou-

pled oscillators. One can consider for example the forced damped sine-Gordon
equation [466]
∂2φ
∂t 2
+ c ∂φ ∂2φ
∂t − ∂ x 2 + sin(φ) = l
(12.1)

where c and l are constants. As explained in Sect. 12.2.2, this type of PDE appears
in many physical phenomena, such as nonlinear resonant optics and Josephson
junctions, or as a dynamic model of electrons in a crystal lattice. Equation (12.97)
describes the motion of an array of pendula each of which is coupled to its nearest
neighbors by a torsional spring with a coupling coefficient k. Each pendulum is sub-
ject to a constant torque l and to a viscous drag force with coefficient c. The angle
xi of the ith pendulum and the vertical axis evolve according to Eq. (12.97).
By considering a one-dimensional grid of N sample points and by computing the
second-order derivative

∂2φ
 1 (φ
∂ x 2 h 2 i+1
− 2φi + φi−1 ) (12.2)

one has the decomposition of the nonlinear partial differential equation into a set of
nonlinear ordinary differential equations of the form

ẍi + c ẋi + εsin(xi ) = k(xi+1 − 2xi + xi−1 + l) i = 1, 2, . . . (12.3)

One can also use a periodicity condition

xi+N = xi + 2π (12.4)

The periodicity condition reduces the system from infinite many pendula into one
with N degrees of freedom

ẍ1 + cx1 + εsin(x1 ) = x2 + x N − 2x1 + l − 2π

ẍ2 + cx2 + εsin(x2 ) = x3 + x1 − 2x2 + l
(12.5)
···
ẍ N + cx N + εsin(x N ) = x1 + x N − 2x N + l + 2π
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 617

12.2.3 Sine-Gordon Nonlinear PDE in the Model

of the Josephson Junction

A transmission line is considered where transverse electromagnetic waves propagate

as shown in Fig. 12.1. The transmission line consists of inductors, capacitors, and
Josephson junctions, such that for a length d x of the transmission line the capacitance
is dC = Cd x, the inductance is d L = Ld x, and the critical current is d I0 = I0 d x
[387, 466, 530]. Dividing the differential voltage drop d V and the shunt current d I
by d x, one obtains
∂V ∂I
∂ x = −L ∂t (12.6)
∂I ∂V
∂ x = −C ∂t − I0 sin(φ)

where φ is the difference φ = φupper − φlower of the superconducting phases. The

voltage is related to the rate of change of φ through the Josepshon equation
∂φ
∂t = 2eV
h
(12.7)

and equivalently
∂φ
∂x = − 2eV
h I
(12.8)

Fig. 12.1 Josephson transmission line described by Josephson junctions (top), a capacitance per
unit length C, an inductance per unit length L, and a critical current per unit length I0
618 12 Differential Flatness Theory for Distributed Parameter Systems

Thus, one arrives at the equation

1 ∂2φ ∂2φ
c2 ∂t 2
− ∂x2
+ 1
λ2j
sin(φ) =0 (12.9)

where c = (LC)−1/2 is the Swihart velocity and λ j = (/2eL I0 )1/2 is the Josephson
length. By rescaling lenghts λ j and λ j /c one arrives at the sine-Gordon equation

∂2φ
∂t 2
+ c ∂φ
∂t −
∂2φ
∂x2
+ sin(φ) = l (12.10)

12.2.4 Current Equation in a Josepshon Transmission Line

A wire of length l is considered which is interrupted by identical Josephson junctions

at positions denoted as x1 , x2 , . . . , x N . The partial differential equation governing the
dynamics of the Josephson transmission line can be determined by considering the
finite element representation shown in Fig. 12.2. Each inductor–capacitor segment
represents a short length Δx of the wire. Some segments also contain a Josephson
junction in series with the inductor.
At the ith segment it holds that

Ii−1 = q̇i + Ii (12.11)

where qi is the charge on the ith capacitor and Ii is the current in the ith inductor.
Equating voltage drops of the two paths from node i to ground it holds


qi
CΔx = LΔx I˙i + qi+1
CΔx + 2e φ̇ j δi, j (12.12)

Fig. 12.2 Finite element

representation of the
Josephson transmission line
(× denotes a Josephson
junction)
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 619

where  is Planck’s constant divided by 2π , e is the magnitude of the electronic

charge, and φ j is the difference in the phase of the wave function across the jth
junction. The Kronecker δ term denotes the voltage drop across the jth Josephson
junction if it appears in segment i of the transmission line. From Eqs. (12.11) and
(12.12) one obtains
v2 Ii+1 −2I i +Ii−1
Δx 2

= I¨i + 2eLΔx φ̈ j δi, j (12.13)

v2 = 1/(LC). Setting the limit Δx→0 yields an equation for the current I (x, t)

∂2 I N 
− v2 ∂∂ x I2 +
2
∂t 2 i=1 2eL φ̈ j δ(x − xj) = 0 (12.14)

The associated boundary conditions are I (0, t) = I (l, t) = Ib . Moreover, about the
dynamics of phase φ at the jth Josephson junction one has

C J
2e φ̈ j + 2eR j φ̇ j + Ic sin(φ j ) = I (x j , t) (12.15)

where I (x j , t) is the current through the jth junction. The parameters of the junction
are the capacitance C j , the resistance R j , and the current Ic .

12.2.5 State-Space Description of the Nonlinear

Wave Dynamics

The previous PDE of the Josephson junction dynamics is generalized in the form of
the following nonlinear wave equation considered

∂2φ
= k ∂∂ xφ2 + f (φ) + u(x, t)
2
(12.16)
∂t 2

where u(x, t) is assumed to be the external control input. Using the approximation
for the partial derivative
∂2φ φi+1 −2φi +φi−1
2 =
(12.17)
∂x 2 Δx

and considering spatial measurements of variable φ along axis x at points x0 +

iΔx, i = 1, 2, . . . , N one has

∂ 2 φi
∂t 2
= K
φ
Δx 2 i+1
− 2K
φ
Δx 2 i
+ K
φ
Δx 2 i−1
+ f (φi ) + u(xi , t) (12.18)

By considering also the associated samples of φ given by φ0 , φ1 , . . . , φ N , φ N +1

one has
620 12 Differential Flatness Theory for Distributed Parameter Systems

∂ 2 φ1
∂t 2
= K
φ
Δx 2 2
− 2K
φ
Δx 2 1
+ K
φ
Δx 2 0
+ f (φ1 ) + u(x1 , t)
∂ 2 φ2
∂t 2
= K
φ
Δx 2 3
− 2K
φ
Δx 2 2
+ K
φ
Δx 2 1
+ f (φ2 ) + u(x2 , t)
∂ 2 φ3
∂t 2
= K
φ
Δx 2 4
− 2K
φ
Δx 2 3
+ K
φ
Δx 2 2
+ f (φ3 ) + u(x3 , t)
(12.19)
···
∂ 2 φ N −1
∂t 2
= K
φ − Δx
Δx 2 N
2K
2 φ N −1 + Δx 2 φ N −2 + f (φ N −1 ) + u(x N1 , t)
K

∂ 2 φN
∂t 2
= Δx
K
2 φ N +1 − Δx 2 φ N + Δx 2 φ N −1 + f (φ N ) + u(x N , t)
2K K

By defining the following state vector

 
x T = φ1 , φ2 , . . . , φ N (12.20)

one obtains the following state-space description

ẍ1 = K
x
Δx 2 2
− 2K
x
Δx 2 1
+ K
φ
Δx 2 0
+ f (x1 ) + u(x1 )
ẍ2 = K
x
Δx 2 3
− 2K
x
Δx 2 2
+ K
x
Δx 2 1
+ f (x2 ) + u(x2 )
ẍ3 = K
x
Δx 2 4
− 2K
x
Δx 2 3
+ K
x
Δx 2 2
+ f (x3 ) + u(x3 )
(12.21)
···
ẍ N −1 = K
x − Δx
Δx 2 N
2K
2 x N −1 + Δx 2 x N −2 + f (x N −1 ) + u(x N −1 )
K

ẍ N = Δx
K
2 φ N +1 − Δx 2 x N + Δx 2 x N −1 + f (x N ) + u(x N )
2K K

Next, the following state variables are defined

y1,i = xi
(12.22)
y2,i = ẋi

and the state-space description of the system becomes as follows:

ẏ1,1 = y2,1
ẏ2,1 = K
y
Δx 2 1,2
− 2K
Δx 1,1
2 y + ΔxK
2 φ0 + f (y1,1 ) + u(y1,1 )
ẏ1,2 = y2,2
ẏ2,2 = K
y
Δx 2 1,3
− 2K
y + Δx
Δx 2 1,2
K
2 y1,1 + f (y1,2 ) + u(y1,2 )
ẏ1,3 = y2,3
ẏ2,3 = K
y
Δx 2 1,4
− 2K
Δx 2 y1,3 + Δx
K
2 y1,2 + f (y1,3 ) + u(y1,3 )
(12.23)
···
···
ẏ1,N −1 = y2,N −1
ẏ2,N −1 = ΔxK
2 y1,N − Δx 2 y1,N −1 + Δx 2 y1,N −2 + f (y1,N −1 ) + u(y1,N −1 )
2K K

ẏ1,N = y2,N
ẏ2,N = ΔxK
2 φ N +1 − Δx 2 y1,N + Δx y1,N −1 + f (y1,N ) + u(y1,N )
2K K
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 621

The dynamical system described in Eq. (12.23) is a differentially flat one with flat
output defined as the vector ỹ = [y1,1 , y1,2 , . . . , y1,N ]. Indeed all state variables can
be written as functions of the flat output and its derivatives.
Moreover, by defining the new control inputs

v1 = K
y − Δx
Δx 2 1,2
2K
2 y1,1 + Δx 2 φ0 + f (y1,1 ) + u(y1,1 )
K

v2 = K
y − Δx
Δx 2 1,3
2K
2 y1,2 + Δx 2 y1,1 + f (y1,2 ) + u(y1,2 )
K

v3 = K
y − Δx
Δx 2 1,4
2K
2 y1,3 + Δx 2 y1,2 + f (y1,3 ) + u(y1,3 )
K
(12.24)
···
v N −1 = K
y
Δx 2 1,N
− Δx
2K
2 y1,N −1 + Δx 2 y1,N −2 +
K
f (y1,N −1 ) + u(y1,N −1 )
v N = ΔxK
2 φ N +1 − Δx 2 y1,N + Δx 2 y1,N −1 +
2K K
f (y1,N ) + u(y1,N )

the following state-space description is obtained

⎛ ⎞ ⎛ ⎞
⎛ ⎞ 0 1 0 0 ··· 0 0 0 0 ⎛ ⎞ 0 0 0 ··· 0 0
ẏ1,1 ⎜ 0 0 0 0 ··· 0 0 0 0 ⎟ y1,1 ⎜ 1 0 0 ··· 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜ ẏ2,1 ⎟ ⎜ 0 0 0 1 · · · 0 0 0 0 ⎟ ⎜ y2,1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎛ ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ ẏ1,2 ⎟ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎜ y1,2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ2,2 ⎟ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎜ y2,2 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ v3 ⎟
⎜ ··· ⎟=⎜ 0 0 0 0 ··· 0 0 0 0 ⎟⎜ ··· ⎟+⎜ 0 0 1 ··· 0 0 ⎟⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ1,N −1 ⎟ ⎜· · · · · · · · · · · · · · · · · ·⎟ ⎜ y1,N −1 ⎟ ⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎝v N −1 ⎠
⎜ ẏ2,N −1 ⎟ ⎜ 0 0 0 0 · · · 0 1 0 0 ⎟ ⎜ y2,N −1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ẏ1,N ⎠ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎝ y1,N ⎠ ⎜ 0 0 0 · · · 1 0 ⎟ v N
⎜ ⎟ ⎜ ⎟
ẏ2,N ⎝ 0 0 0 0 ··· 0 0 0 1 ⎠ y2,N ⎝ 0 0 0 ··· 0 0 ⎠
0 0 0 0 ··· 0 0 0 0 0 0 0 ··· 0 1
(12.25)

By selecting measurements from a subset of points x j j∈[1, 2, . . . , m], the associated

observation (measurement) equation becomes
⎛ ⎞
⎛ ⎞ y1,1
⎛ ⎞ 1 0 0 ··· 0 0 ⎜ ⎟
⎜ y2,1 ⎟
z1 ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ y1,2 ⎟
⎜ z2 ⎟ ⎜ ⎟⎜ ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ y2,2 ⎟ (12.26)
⎝· · ·⎠ ⎜ ⎟⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎜ ⎟
0 0 0 · · · 0 0 ⎝ y1,N ⎠
y2,N

Thus, in matrix form one has the following state-space description of the system

ỹ˙ = A ỹ + Bv
(12.27)
z̃ = C ỹ
622 12 Differential Flatness Theory for Distributed Parameter Systems

Moreover, denoting a = Dx K
2 and b = − Dx 2 , the initial description of the system
2K

given in Eq. (12.25) is rewritten as follows

⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 ··· 0 0 0 0 0 0 0 0 0 ··· 0 0
⎛ ⎞ ⎛ ⎞
ẏ1,1 ⎜b 0 a 0 0 0 0 · · · 0 0 0 0 0 0⎟ y1,1 ⎜ 1 0 0 ··· 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜ ẏ2,1 ⎟ ⎜ 0 0 0 1 0 0 0 · · · 0 0 0 0 0 0 ⎟ ⎜ y2,1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎛ ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ ẏ1,2 ⎟ ⎜a 0 b 0 a 0 0 · · · 0 0 0 0 0 0⎟ ⎜ y1,2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ2,2 ⎟ ⎜ 0 0 0 0 0 1 0 · · · 0 0 0 0 0 0 ⎟ ⎜ y2,2 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ v3 ⎟
⎜ · · · ⎟ = ⎜0 0 a 0 b 0 a · · · 0 0 0 0 0 0 ⎟ ⎜ · · · ⎟ + ⎜ 0 0 1 · · · 0 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ1,N −1 ⎟ ⎜ · · · · · · · · · · · · · · · · · · · · · ⎟ ⎜ y1,N −1 ⎟ ⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎝v N −1 ⎠
⎜ ẏ2,N −1 ⎟ ⎜ 0 0 0 0 0 0 0 · · · 0 0 0 1 0 0 ⎟ ⎜ y2,N −1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ vN
⎝ ẏ1,N ⎠ ⎜0 0 0 0 0 0 0 · · · a 0 b 0 a 0⎟ ⎝ y1,N ⎠ ⎜ 0 0 0 · · · 1 0 ⎟
⎜ ⎟ ⎜ ⎟
ẏ2,N ⎝0 0 0 0 0 0 0 · · · 0 0 0 0 0 1⎠ y2,N ⎝ 0 0 0 ··· 0 0 ⎠
0 0 0 0 0 0 0 ··· 0 0 a 0 b 0 0 0 0 ··· 0 1
(12.28)

The associated control inputs are defined as

v1 = Δx K
2 φ0 + f (y1,1 ) + u(y1,1 )
v2 = f (y1,2 ) + u(y1,2 )
v3 = f (y1,3 ) + u(y1,3 )
(12.29)
···
v N −1 = f (y1,N −1 ) + u(y1,N −1 )
v N = Δx K
2 φ N +1 + f (y1,N ) + u(y1,N )

12.2.6 Solution of the Control and Estimation Problem

for Nonlinear Wave Dynamics

12.2.6.1 Solution of the Control Problem

It holds that the dynamics of the linearized equivalent model of the nonlinear wave-
type PDE takes the form

ÿ1,1 = by1,1 + ay1,2 + v1

ÿ1,2 = ay1,1 + by1,2 + ay1,3 + v2
ÿ1,3 = ay1,2 + by1,3 + ay1,4 + v3
··· (12.30)
···
ÿ1,N −1 = ay1,N −2 + by1,N −1 + ay1,N + v N −1
ÿ1,N = ay1,N −1 + by1,N + v N
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 623

For the dynamics given in Eq. (12.30), the feedback control law that assures tracking
of the reference setpoint y d = [y1,1
d , yd , yd , . . . , yd
1,2 1,3 1,N −1 , y1,N ] is
d T

v1 = ÿ1,1
d − by
1,1 − ay1,2 − K d1,1 ( ẏ1,1 − ẏ1,1 ) − K p1,1 (y1,1 − y1,1 )
d d d d

v2 = ÿ1,2
d − ay
1,1 − by1,2 − ay1,3 − K d1,2 ( ẏ1,2 − ẏ1,2 ) − K p1,2 (y1,2 − y1,2 )
d d d d

v3 = ÿ1,3
d − ay
1,2 − by1,3 − ay1,4 − K d1,3 ( ẏ1,3 − ẏ1,3 ) − K p1,3 (y1,3 − y1,3 )
d d d d

···
···
v N −1 = ÿ1,N
d
−1 − ay1,N −2 − by1,N −1 − ay1,N −
−K d1,N −1 ( ẏ1,N
d
−1 − ẏ1,N −1 ) − K p1,N −1 (y1,N −1 − y1,N −1 )
d d d

v N = ÿ1,N
d − ay
1,N −1 − by1,N − K d1,N ( ẏ1,N − ẏ1,N ) − K p1,N (y1,N − y1,N )
d d d d

(12.31)

Next, using Eq. (12.31) one can compute the control action that is applied to the
wave-type dynamics

v1 = K
φ
Δx 2 0
+ f (y1,1 ) + u(y1,1 ) ⇒ u(y1,1 ) = v1 − K
φ
Δx 2 0
− f (y1,1 )
v2 = f (y1,2 ) + u(y1,2 ) ⇒ u(y1,2 ) = v2 − f (y1,2 )
v3 = f (y1,3 ) + u(y1,3 ) ⇒ u(y1,3 ) = v3 − f (y1,3 )
···
···
v N −1 = f (y1,N −1 ) + u(y1,N −1 ) ⇒ u(y1,N −1 ) = v N −1 − f (y1,N −1 )
vN = K
φ
Δx 2 N +1
+ f (y N ) + u(y1,N ) ⇒ u(y1,N ) = v N − K
φ
Δx 2 N +1
− f (y1,N )
(12.32)

12.2.6.2 Solution of the Estimation Problem

By selecting measurements from a subset of points x j j ∈ [1, 2, . . . , m], the associ-

ated observation (measurement) equation remains as in Eq. (12.26), i.e.,
⎛ ⎞
⎛ ⎞ y1,1
⎛ ⎞ 1 0 0 ··· 0 0 ⎜ ⎟
⎜ y2,1 ⎟
z1 ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ y1,2 ⎟
⎜ z2 ⎟ ⎜ ⎟⎜ ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ y2,2 ⎟ (12.33)
⎝· · ·⎠ ⎜ ⎟⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎜ ⎟
0 0 0 · · · 0 0 ⎝ y1,N ⎠
y2,N
624 12 Differential Flatness Theory for Distributed Parameter Systems

For the description of the system in the form of Eq. (12.27) one can perform estimation
using the Derivative-free nonlinear Kalman Filter recursion. In the filter’s algorithm,
the previously defined matrices A, B, and C are substituted by their discrete-time
equivalents Ad , Bd , and Cd . The discrete-time Kalman filter can be decomposed into
two parts: (i) time update (prediction stage), and (ii) measurement update (correction
stage).
measurement update:

K (k) = P − (k)CdT [C·P − (k)CdT + R]−1

ŷ(k) = ŷ − (k) + K (k)[z(k) − Cd ŷ − (k)] (12.34)
P(k) = P − (k) − K (k)Cd P − (k)

time update:
P − (k + 1) = Ad (k)P(k)AdT (k) + Q(k)
(12.35)
ŷ − (k + 1) = Ad (k) ŷ(k) + Bd (k)u(k)

Therefore, by taking measurements of φ(x, t) at time instant t at a small number of

measuring points j = 1, . . . , n 1 it is possible to estimate the complete state vector,
i.e., to get values of φ in a mesh of points that covers efficiently the variations of
φ(x, t). By processing a sequence of output measurements of the system, one can
obtain local estimates of the state vector ŷ. The measuring points (active sensors)
can vary in time provided that the observability criterion for the state-space model
of the PDE holds.
According to Eqs. (12.34) and (12.35) an estimate of the system’s state vector ŷ
is obtained. Should one want to obtain an estimate of the state vector x of the initial
state-space description of the nonlinear wave-type dynamics, he should apply the
inverse differential flatness transformation connecting xi ’s to yi ’s. In the examined
model of the wave PDE it holds that

x̂1 = ŷ1,1 x̂2 = ŷ1,2 x̂3 = ŷ1,3

··· ··· ··· (12.36)
x̂ N −2 = ŷ1,N −2 x̂ N −1 = ŷ1,N −1 x̂ N = ŷ1,N

Also in the case of distributed parameter systems, the proposed Derivative-free non-
linear Kalman Filter is of improved precision because unlike other nonlinear filtering
schemes, e.g., unlike the Extended Kalman Filter it does not introduce cumula-
tive numerical errors due to approximative linearization of the system’s dynamics.
Besides, it is computationally more efficient (faster) because it does not require to
calculate Jacobian matrices and partial derivatives.
12.2 Pointwise Flatness-Based Control of Distributed Parameter Systems 625

Fig. 12.3 Grid points for

measuring φ(x, t)

If state estimation-based control is applied to the state-space model of the non-

linear wave PDE then the associated control inputs are (Fig. 12.3)

v1 = K
φ +
Δx 2 0
f ( ŷ1,1 ) + u( ŷ1,1 ) ⇒ u( ŷ1,1 ) = v1 − Δx
K
2 φ0 − f ( ŷ1,1 )
v2 = f ( ŷ1,2 ) + u( ŷ1,2 ) ⇒ u( ŷ1,2 ) = v2 − f ( ŷ1,2 )
v3 = f ( ŷ1,3 ) + u( ŷ1,3 ) ⇒ u( ŷ1,3 ) = v3 − f ( ŷ1,3 )
···
···
v N −1 = f ( ŷ1,N −1 ) + u( ŷ1,N −1 ) ⇒ u( ŷ1,N −1 ) = v N −1 − f ( ŷ1,N −1 )
v N = Δx K
2 φ N +1 + f ( ŷ N ) + u( ŷ1,N ) ⇒ u( ŷ1,N ) = v N − Δx 2 φ N +1 − f ( ŷ1,N )
K

(12.37)

12.2.7 Simulation Tests

12.2.7.1 State Estimation-Based Control of the Nonlinear Wave PDE

The performance of the proposed control scheme has been tested in the model of the
sine-Gordon nonlinear wave-type partial differential equation. A discretization grid
of the PDE consisting of N = 50 points was considered. At each grid point (see
for instance the grid of Fig. 12.3) the local control input u(xi , t) was exerted to the
system. The extended state-space description of the wave-type PDE comprised a state
vector of dimension y ∈ R 100×1 . The number of measurement points was m ≤ N
where m was selected such that the observability of the linearized state-space model
626 12 Differential Flatness Theory for Distributed Parameter Systems

φ
Fig. 12.4 a Nonlinear wave dynamics φ(x, t) without control, b nonlinear wave dynamics φ(x, t)
made to track a sinusoidal setpoint after applying control

of the PDE is preserved. The obtained results are presented in Figs. 12.4, 12.5, 12.6,
and 12.7. It can be observed that through the application of suitable control inputs the
wave function φ(x, t) and its derivative φ̇(x, t) can be made to track the desirable
sinusoidal setpoints. Moreover, in Figs. 12.8, 12.9, 12.10, and 12.11 diagrams are
provided about the tracking accuracy of the proposed control method at specific
points of the discretization grid. It can be noticed again that the state estimation-
based control method resulted in fast and accurate tracking of the reference setpoints
(Fig. 12.12).
φ

Fig. 12.5 State estimation-based control with the use of the Derivative-free nonlinear Kalman
Filter: a nonlinear wave dynamics φ(x, t) without control, b nonlinear wave dynamics φ(x, t)
made to track a sinusoidal setpoint after applying control
12.3 Control of Heat Diffusion in Arc Welding Using Differential Flatness … 627

φ
Fig. 12.6 a First time derivative of the nonlinear wave dynamics φ̇(x, t) without control, b first
time derivative of the nonlinear wave dynamics φ̇(x, t) made to track a sinusoidal setpoint after
applying control
φ

Fig. 12.7 State estimation-based control with the use of the Derivative-free nonlinear Kalman Filter:
a first time derivative of the nonlinear wave dynamics φ̇(x, t) without control, b first time derivative
of the nonlinear wave dynamics φ̇(x, t) made to track a sinusoidal setpoint after applying control

12.3 Control of Heat Diffusion in Arc Welding Using

Differential Flatness Theory and Nonlinear Kalman
Filtering

12.3.1 Overview

The previous results on flatness-based control of nonlinear PDEs will be extended

to the control of the nonlinear heat diffusion equation, using as a case study control
628 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
4 1 4 2

3 0.5 2
1
5

6
2 0 0

0
1 −0.5 −2

0 −1 −4 −1
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
6 1 4 2

0.5 2 1
4
7

8
0 0 0

2
−0.5 −2 −1

0 −1 −4 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.8 Control and estimation of wave dynamics at grid points 3 (first row of diagrams) to 4
(second row of diagrams), a function φ(x, t) and its derivative when no control input is applied, b
function φ(x, t) and its derivative under state feedback control

(a) (b)
6 2 4 2

2
4 1 1
17

18

17

18

2 0 0
−2

0 −1 −4 −1
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
6 2 4 2

1 2 1
4
19

20

19

20

0 0 0

2
−1 −2 −1

0 −2 −4 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.9 Control and estimation of wave dynamics at grid points 9 (first row of diagrams) to 10
(second row of diagrams), a function φ(x, t) and its derivative when no control input is applied, b
function φ(x, t) and its derivative under control

of the gas metal arc welding process (GMAW). Arc welding is one of the primary
tasks in shipbuilding [532] and in other industrial applications. Automated welding
in shipbuilding is required because of the low productivity of hand welding, which
is the result of the severe environmental conditions produced in the intense heat and
the fumes that are generated by the welding process. The dynamics of the welding
12.3 Control of Heat Diffusion in Arc Welding Using Differential Flatness … 629

(a) (b)
15 2 4 2

1 2
10 1
57

58

57

58
0 0

5 0
−1 −2

0 −2 −4 −1
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
15 2 4 2

1 2 1
10
59

60

59

60
0 0 0

5
−1 −2 −1

0 −2 −4 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.10 Control and estimation of wave dynamics at grid points 29 (first row of diagrams) to
30 (second row of diagrams), a function φ(x, t) and its derivative when no control input is applied,
b function φ(x, t) and its derivative under state feedback control

(a) (b)4
6 4 2

2 2
4 1
85

86

85

86

0 0

2 0
−2 −2

0 −4 −4 −1
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
6 4 4 2

2 2 1
4
87

88

87

88

0 0 0

2
−2 −2 −1

0 −4 −4 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.11 Control and estimation of wave dynamics at grid points 43 (first row of diagrams) to
44 (second row of diagrams), a function φ(x, t) and its derivative when no control input is applied,
b function φ(x, t) and its derivative under state feedback control

process is given by a diffusion-type partial differential equation which describes the

spatiotemporal variations of the temperature distribution in the welded workpiece.
The parameters (inputs) that affect this temperature distribution are the velocity of
the torch and the heat input power. It is important to control heat diffusion in the
630 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
10 10 4 2

5 2
5 1
93

94

93

94
0 0

0 0
−5 −2

−5 −10 −4 −1
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 20 4 2

2 1
5 10
95

96

95

96
0 0

0 0
−2 −1

−5 −10 −4 −2
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.12 Control and estimation of wave dynamics at grid points 47 (first row of diagrams) to
48 (second row of diagrams), a function φ(x, t) and its derivative when no control input is applied,
b function φ(x, t) and its derivative under state feedback control

welding process and the associated temperature distribution in the welded material,
because this finally determines the quality, strength, and endurance of the weld.
According to welding theory, the temperature of the heat-affected zone determines
the quality of the weld [257, 266, 489, 567]. The heat-affected zone is defined as the
area round the weld bead where the temperature of the melted material varies between
a lower and upper limit, where each limit is associated with transition to a different
phase and structure of the material. The structure of the material that is formed after
welding as well as the defects appearing in the weld depend on the temperature that is
developed in the heat-affected zone and its variations during the welding process. For
the monitoring of the thermal distribution in the welded workpiece several methods
have been implemented, such as the use of thermocouples, infrared thermometers
and infrared cameras. Efficient control of the thermal distribution during welding
is still an open problem. In this section a solution will be developed based on the
previous results on control and state estimation for distributed parameter systems
with the use of differential flatness theory [427, 453].
Feedback control of diffusion-type (parabolic) PDEs has been a subject of exten-
sive research and several remarkable results have been produced [320, 463, 557,
559, 619]. Following the procedure for numerical solution of the nonlinear PDE of
the nonlinear heat diffusion, a set of coupled nonlinear ordinary differential equations
is obtained and written in a state-space form [175, 388]. For the latter state-space
description, differential flatness properties are proven. Thus, it is shown that all state
variables and the control inputs of the state-space model can be written as functions
12.3 Control of Heat Diffusion in Arc Welding Using Differential Flatness … 631

of a vector of algebraic variables that constitute the flat output and of the flat out-
put’s derivatives [55, 152, 286, 362, 465]. By applying a change of coordinates
(diffeomorphism) which is based on differential flatness theory it is shown that the
state-space model of the heat diffusion PDE can be written in a linear form, in which
the previously noted nonlinear ordinary differential equations are now transformed
into linear ones. Next, feedback control is applied to the heat diffusion PDE. For
each local linear model of the aforementioned differential equations the state feed-
back control is selected such that asymptotic stability is assured. This can be done
using for instance pole placement methods. By computing the control input of the
heat diffusion PDE, which varies both in space and time one can also compute the
velocity that the welding torch should have a specific time instant at a point of the
cartesian frame, so that the temperature distribution of the workpiece converges to
the reference setpoints.
Another aim of this section is to implement state feedback control of the nonlinear
heat diffusion PDE using measurements from a small number of sensors [335, 437].
This implies that for state vector elements of the PDE’s state-space description which
cannot be measured directly, state estimation with filtering methods has to be be
applied. Filtering for nonlinear distributed parameter systems is again a nontrivial
problem [42, 467, 559, 592]. To solve it, both observer-based and Kalman Filter-based
approaches have been proposed [81, 124, 185, 198]. To this end, in this section, a
new nonlinear filtering method, under the name Derivative-free nonlinear Kalman
Filtering, is proposed. The filter consists of the standard Kalman Filter recursion
applied to the linear equivalent state-space model of the heat diffusion PDE and of
an inverse transformation that enables to obtain estimates of the state variables in the
initial nonlinear description [31, 408, 414].

12.4 Dynamic Model of the Arc Welding Process

The reference frame of Fig. 12.3 is introduced and the following nonlinear heat dif-
fusion equation is considered, describing the spatiotemporal variations of the tem-
perature in the welded workpiece [266, 489]

∂φ
= K ∂∂ xφ2 + f (φ) + u(x, t)
2
∂t
(12.38)

where K is the heat conductivity and u(x, t) is a term associated with the partial
derivative of the temperature distribution with respect to the space variable x as well
as with the velocity of the welding torch. This control input is given by
∂φ(x,t)
u(x, t) = ∂ x vs (t)
(12.39)
632 12 Differential Flatness Theory for Distributed Parameter Systems

Fig. 12.13 Control scheme of the nonlinear heat diffusion in the welding process

with vs (t) to stand for the velocity of the torch at time instant t. Moreover, about the
nonlinear term f (x, t) this is given by

f (x, t) = q(x)h(t) (12.40)

The term h(t) stands for the heating power provided by the torch. The term q(x)
denotes the spatial distribution of the heating input and can be approximated by a
Gaussian, that is [257, 567]
2
− (x−xs ) (12.41)
q(x) = ae σ 2

where a and σ are constant parameters and xs is the position of the torch in the
reference system used for the welding process. The previous dynamic model of the
welding process is the result of the energy conservation principle. In the simplest
scenario of welding the following assumptions are made: (i) the thermal conduc-
tivity coefficient K remains constant throughout the process and is not affected by
temperature variations of the welded material, (ii) the only heat source provided to
the welded material is the one given by the torch, (iii) no heat is either produced or
lost at any other parts of the workpiece (Fig. 12.13).
12.5 State-Space Description of the Nonlinear Heat Diffusion Dynamics 633

12.5 State-Space Description of the Nonlinear Heat

Diffusion Dynamics

Using the approximation for the partial derivative in the partial differential equation
given in Eq. (12.38) one has

∂2φ φi+1 −2φi +φi−1

∂x2
 = Δx 2
(12.42)

and considering spatial measurements of variable φ along axis x at points x0 +

iΔx, i = 1, 2, . . . , N one has
∂φi
∂t = K
φ
Δx 2 i+1
− 2K
φ
Δx 2 i
+ K
φ
Δx 2 i−1
+ f (φi ) + u(xi , t) (12.43)

By considering the associated samples of φ given by φ0 , φ1 , . . . , φ N , φ N +1 one has

∂φ1
∂t = K
φ
Δx 2 2
− 2K
φ
Δx 2 1
+ K
φ
Δx 2 0
+ f (φ1 ) + u(x1 , t)
∂φ2
∂t = K
φ
Δx 2 3
− 2K
φ
Δx 2 2
+ K
φ
Δx 2 1
+ f (φ2 ) + u(x2 , t)
∂φ3
∂t = K
φ
Δx 2 4
− 2K
φ
Δx 2 3
+ K
φ
Δx 2 2
+ f (φ3 ) + u(x3 , t)
(12.44)
···
∂φ N −1
∂t = K
φ − Δx
Δx 2 N
2K
2 φ N −1 + Δx 2 φ N −2 + f (φ N −1 ) + u(x N −1 , t)
K

∂φ N
∂t = ΔxK
2 φ N +1 − Δx 2 φ N + Δx 2 φ N −1 + f (φ N ) + u(x N , t)
2K K

By defining the following state vector

 
x T = φ1 , φ2 , . . . , φ N (12.45)

one obtains the following state-space description

ẋ1 = K
x
Δx 2 2
− 2K
x
Δx 2 1
+ K
φ
Δx 2 0
+ f (x1 ) + u(x1 )
ẋ2 = K
x
Δx 2 3
− 2K
x
Δx 2 2
+ K
x
Δx 2 1
+ f (x2 ) + u(x2 )
ẍ3 = K
x
Δx 2 4
− 2K
x
Δx 2 3
+ K
x
Δx 2 2
+ f (x3 ) + u(x3 )
(12.46)
···
ẋ N −1 = K
x − Δx
Δx 2 N
2K
2 x N −1 + Δx 2 x N −2 + f (x N −1 ) + u(x N −1 )
K

ẋ N = Δx
K
2 φ N +1 − Δx 2 x N + Δx 2 x N −1 + f (x N ) + u(x N )
2K K

Next, the following state variables are defined

y1,i = xi
(12.47)
y2,i = ẋi
634 12 Differential Flatness Theory for Distributed Parameter Systems

and the state-space description of the system becomes as follows:

ẏ1,1 = K
y − Δx
Δx 2 1,2
2K
2 y1,1 + Δx 2 φ0 + f (y1,1 ) + u(y1,1 )
K

ẏ1,2 = K
y − Δx
Δx 2 1,3
2K
2 y1,2 + Δx 2 y1,1 + f (y1,2 ) + u(y1,2 )
K

ẏ1,3 = K
y − Δx
Δx 2 1,4
2K
2 y1,3 + Δx 2 y1,2 + f (y1,3 ) + u(y1,3 )
K

(12.48)
···
···
ẏ1,N −1 = K
y
Δx 2 1,N
− Δx2K
2 y1,N −1 + Δx 2 y1,N −2
K
+ f (y1,N −1 ) + u(y1,N −1 )
ẏ1,N = ΔxK
2 φ N +1 − Δx 2 y1,N + Δx y1,N −1
2K K
+ f (y1,N ) + u(y1,N )

The dynamical system described in Eq. (12.48) is a differentially flat one with flat
output defined as the vector ỹ = [y1,1 , y1,2 , . . . , y1,N ]. Indeed all state variables can
be written as functions of the flat output and its derivatives.
Moreover, by defining the new control inputs

v1 = Δx K
2 φ0 + f (y1,1 ) + u(y1,1 )
v2 = f (y1,2 ) + u(y1,2 )
v3 = f (y1,3 ) + u(y1,3 )
(12.49)
···
v N −1 = f (y1,N −1 ) + u(y1,N −1 )
v N = Δx K
2 φ N +1 + f (y1,N ) + u(y1,N )

the following state-space description is obtained

⎛ 2K ⎞
− Δx 2 Δx
K
2 0 0 ··· 0 0 0 0
⎛ ⎜
⎞ ⎜ K ⎟⎛ ⎞
ẏ1,1 ⎜ Δx 2 − Δx 2 Δx 2 0 · · ·
2K K
0 0 0 0 ⎟ ⎟ y1,1
⎜ ẏ1,2 ⎟ ⎜ ⎟⎜
y1,2 ⎟
⎜ ⎟ ⎜ 0 K
− Δx
2K
2 Δx 2 · · ·
K
0 0 0 0 ⎟ ⎟⎜ ⎟
⎜ ··· ⎟ = ⎜ Δx 2 ⎜
⎟⎜ ··· ⎟
⎜ ⎟ ⎜ ⎟+
⎝ ẏ1,N −1 ⎠ ⎜
⎜ ··· ··· ··· ··· ··· ··· ··· ··· ··· ⎟ ⎟ ⎝ y1,N −1 ⎠
⎜ K ⎟
ẏ1,N ⎜ 0 0 0 0 ··· 0 K
− Δx
2K ⎟ y1,N
⎝ Δx 2 2 Δx 2 ⎠
0 0 0 0 ··· 0 0 Δx K
2 − Δx 2
2K

⎛ ⎞⎛ ⎞
1 0 0 ··· 0 0 v1
⎜ 0 1 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟⎜ ⎟
⎜ 0 0 1 · · · 0 0 ⎟ ⎜ v3 ⎟
⎜
+⎜ ⎟ ⎜ ⎟
⎟⎜ ⎟
⎜· · · · · · · · · · · · ⎟ ⎜ · · · ⎟
⎝ 0 0 0 · · · 1 0 ⎠ ⎝v N −1 ⎠
0 0 0 ··· 0 1 vN
(12.50)
12.5 State-Space Description of the Nonlinear Heat Diffusion Dynamics 635

Assuming that all measurements from the set of points x j j ∈ [1, 2, . . . , m] are
available, the associated observation (measurement) equation becomes
⎛ ⎞
⎛ ⎞ y1,1
⎛ ⎞ 1 0 0 ··· 0 0 ⎜
z1 ⎟
y1,2
⎜ ⎟⎜ ⎟
⎜ z2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟ ⎜ ⎟
y1,3
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ ⎟ (12.51)
⎝ · · ·⎠ ⎜ ⎟ ⎜ ⎟
···
⎝ 0 0 0 ··· 1 0 ⎠⎜ ⎟
zm ⎝ y1,N −1 ⎠
0 0 0 ··· 0 1
y1,N

Thus, in matrix form one has the following state-space description of the system

ỹ˙ = A ỹ + Bv
(12.52)
z̃ = C ỹ

Moreover, denoting a = Dx K
2 and b = − Dx 2 , the initial description of the system
2K

given in Eq. (12.50) is rewritten as follows:

⎛ ⎞ ⎛ ⎞⎛ ⎞
⎛ ⎞ b a 0 0 0 ··· 0 0 0 0 0 0 ⎛ ⎞ 1 0 0 ··· 0 0 v1
ẏ1,1 ⎜ ⎟ y 1,1 ⎜ ⎟⎜ ⎟
⎜ ẏ1,2 ⎟ ⎜ a b a 0 0 · · · 0 0 0 0 0 0 ⎟ ⎜ y1,2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ 0 0 a b a ··· 0 0 0 0 0 0 ⎟⎜ ⎟ ⎜ 0 0 1 · · · 0 0 ⎟ ⎜ v3 ⎟
⎜ ··· ⎟=⎜ ⎟⎜ ··· ⎟ + ⎜ ⎟⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎝ ẏ1,N −1 ⎠ ⎜· · · · · · · · · · · · · · · · · · · · ·⎟ ⎝ y1,N −1 ⎠ ⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟
⎝ 0 0 0 0 0 0 0 ··· a b a 0 ⎠ ⎝ 0 0 0 · · · 1 0 ⎠ ⎝v N −1 ⎠
ẏ1,N y1,N
0 0 0 0 0 0 0 ··· 0 0 a b 0 0 0 ··· 0 1 vN
(12.53)

12.6 Solution of the Control and Estimation Problem for

Nonlinear Heat Diffusion

12.6.1 Solution of the Control Problem

It holds that the dynamics of the linearized equivalent model of the nonlinear heat
diffusion PDE takes the form

ẏ1,1 = by1,1 + ay1,2 + v1

ẏ1,2 = ay1,1 + by1,2 + ay1,3 + v2
ẏ1,3 = ay1,2 + by1,3 + ay1,4 + v3
··· (12.54)
···
ẏ1,N −1 = ay1,N −2 + by1,N −1 + ay1,N + v N −1
ẏ1,N = ay1,N −1 + by1,N + v N
636 12 Differential Flatness Theory for Distributed Parameter Systems

For the dynamics given in Eq. (12.54), the feedback control law that assures tracking
of the reference setpoint y d = [y1,1
d , yd , yd , . . . , yd
1,2 1,3 1,N −1 , y1,N ] is
d T

d − by
v1 = ẏ1,1 d d
1,1 − ay1,2 − K p1,1 (y1,1 − y1,1 )
d d
v2 = ẏ1,2 − ay1,1 − by1,2 − ay1,3 − K p1,2 (y1,2 − y1,2d )
d d d
v3 = ẏ1,3 − ay1,2 − by1,3 − ay1,4 − K p1,3 (y1,3 − y1,3 )
··· (12.55)
···
d
v N −1 = ẏ1,N − − d
− ay1,N − K p1,N −1 (y1,N d
−1 ay 1,N −2 by 1,N −1 −1 − y1,N −1 )
d d
v N = ẏ1,N − ay1,N −1 − by1,N − K p1,N (y1,N − y1,N )d

Next, using Eq. (12.55) one can compute the control action that is applied to the heat
diffusion dynamics

v1 = K 2 φ0 + f (y1,1 ) + u(y1,1 ) ⇒ u(y1,1 ) = v1 − K 2 φ0 − f (y1,1 )

Δx Δx
v2 = f (y1,2 ) + u(y1,2 ) ⇒ u(y1,2 ) = v2 − f (y1,2 )
v3 = f (y1,3 ) + u(y1,3 ) ⇒ u(y1,3 ) = v3 − f (y1,3 )
··· (12.56)
···
v N −1 = f (y1,N −1 ) + u(y1,N −1 ) ⇒ u(y1,N −1 ) = v N −1 − f (y1,N −1 )
v N = K 2 φ N +1 + f (y N ) + u(y1,N ) ⇒ u(y1,N ) = v N − K 2 φ N +1 − f (y1,N )
Δx Δx

The control input u(x, t) is a heat distribution that is generated by the moving welding
torch. By knowing the spatiotemporal variations u(x, t) of the heat distribution that is
provided by the torch and the associated partial derivative ∂∂ux one can also compute the
speed of the torch v that is needed for making the heat distribution of the welded area
reach the desirable setpoints. The proposed control scheme is depicted in Fig. 12.14.
For the computation of the setpoints of the speed of the welding torch one should
proceed as follows: At each time instant ti and for the complete sequence of the grid
points x1 , x2 , . . . , x N one has the set of equations:

∂φ(x,t)
∂ x |x=x1 ,t=ti ·v(ti ) = u(x1 , ti )
∂φ(x,t)
∂ x |x=x2 ,t=ti ·v(ti ) = u(x2 , ti )
··· (12.57)
∂φ(x,t)
∂ x |x=x N −1 ,t=ti ·v(ti ) = u(x N −1 , ti )
∂φ(x,t)
∂ x |x=x N ,t=ti ·v(ti ) = u(x N , ti )

By substituting the values of x j , j = 1, 2, . . . , x N , and ti in Eq. (12.57) the compu-

tation of v(ti ), that is v(ti ) can be performed in the least squares sense.
12.6 Solution of the Control and Estimation Problem for Nonlinear Heat Diffusion 637

Fig. 12.14 Control scheme of the nonlinear heat diffusion in the welding process

12.6.2 Solution of the Estimation Problem

Next, measurements are selected from a subset of points x j j ∈ [1, 2, . . . , N ] so

as the observability of the state-space model of the welding process to be preserved
(Fig. 12.3). For instance, the associated observation (measurement) equation may
take the form ⎛ ⎞
⎛ ⎞ y1,1
⎛ ⎞ 1 0 0 ··· 0 0 ⎜
z1 ⎟
⎜ ⎟ ⎜ y1,2 ⎟
⎜ z 2 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ y1,3 ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ ⎟ (12.58)
⎝ · · ·⎠ ⎜ ⎟⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎝ y1,N −1 ⎠
0 0 0 ··· 0 0
y1,N

For the description of the system in the form of Eq. (12.53) one can perform estimation
using the Derivative-free nonlinear Kalman Filter recursion. In the filter’s algorithm,
the previously defined matrices A, B, and C are substituted by their discrete-time
equivalents Ad , Bd , and Cd . The discrete-time Kalman filter can be decomposed into
two parts: (i) time update (prediction stage), and (ii) measurement update (correction
stage).
638 12 Differential Flatness Theory for Distributed Parameter Systems

measurement update:

K (k) = P − (k)CdT [C·P − (k)CdT + R]−1

ŷ(k) = ŷ − (k) + K (k)[z(k) − Cd ŷ − (k)] (12.59)
P(k) = P − (k) − K (k)Cd P − (k)

time update:
P − (k + 1) = Ad (k)P(k)AdT (k) + Q(k)
(12.60)
ŷ − (k + 1) = Ad (k) ŷ(k) + Bd (k)u(k)

Therefore, by taking measurements of φ(x, t) at time instant t at a small number of

measuring points j = 1, . . . , n 1 it is possible to estimate the complete state vector,
i.e., to get values of φ in a mesh of points that covers efficiently the variations of
φ(x, t). By processing a sequence of output measurements of the system, one can
obtain estimates of the state vector ŷ. The measuring points (active sensors) can vary
in time provided that the observability criterion for the state-space model of the PDE
holds.
According to Eqs. (12.59) and (12.60) an estimate of the system’s state vector ŷ
is obtained. Should one want to obtain an estimate of the state vector x of the initial
state-space description of the nonlinear heat diffusion, he should apply the inverse
differential flatness transformation connecting xi ’s to yi ’s. In the examined model of
the nonlinear heat diffusion it holds that

x̂1 = ŷ1,1 x̂2 = ŷ1,2 x̂3 = ŷ1,3

··· ··· ··· (12.61)
x̂ N −2 = ŷ1,N −2 x̂ N −1 = ŷ1,N −1 x̂ N = ŷ1,N

As explained in previous sections, the Derivative-free nonlinear Kalman Filter is of

improved precision because unlike other nonlinear filtering schemes, e.g., unlike the
Extended Kalman Filter it does not introduce cumulative numerical errors due to
approximative linearization of the system’s dynamics. Besides, it is computationally
more efficient (faster) because it does not require to calculate Jacobian matrices and
partial derivatives.
If state estimation-based control is applied to the state-space model of the non-
linear heat diffusion equation of the welding process, then the associated control
inputs are

v1 = K 2 φ0 + f ( ŷ1,1 ) + u( ŷ1,1 ) ⇒ u( ŷ1,1 ) = v1 − K 2 φ0 − f ( ŷ1,1 )

Δx Δx
v2 = f ( ŷ1,2 ) + u( ŷ1,2 ) ⇒ u( ŷ1,2 ) = v2 − f ( ŷ1,2 )
v3 = f ( ŷ1,3 ) + u( ŷ1,3 ) ⇒ u( ŷ1,3 ) = v3 − f ( ŷ1,3 )
··· (12.62)
···
v N −1 = f ( ŷ1,N −1 ) + u( ŷ1,N −1 ) ⇒ u( ŷ1,N −1 ) = v N −1 − f ( ŷ1,N −1 )
v N = K 2 φ N +1 + f ( ŷ N ) + u( ŷ1,N ) ⇒ u( ŷ1,N ) = v N − K 2 φ N +1 − f ( ŷ1,N )
Δx Δx
12.7 Simulation Tests 639

12.7 Simulation Tests

The performance of the proposed control scheme has been tested in the model of
the nonlinear heat diffusion PDE that is associated with the arc welding process.
A discretization grid of the PDE consisting of N = 50 points, along the x-axis
(Fig. 12.13) was considered. At each grid point the local control input u(xi , t) was
exerted to the system. The state-space description of the wave-type PDE comprised
also a state vector of dimension y ∈ R 50×1 . The number of measurement points was
m ≤ N where m was selected such that the observability of the linearized state-space
model of the PDE is preserved (actually measurements were sampled from half of
the number of grid points). The obtained results are presented in Figs. 12.15, 12.16,
12.17, 12.18, 12.19, 12.20, and 12.21. It can be observed that through the application
of suitable control inputs the wave function φ(x, t) and its derivative φ̇(x, t) can be
made to track the desirable sinusoidal setpoints. Actually, the tracking of a sinusoidal
reference setpoint by the distribution φ(x, t) of the nonlinear heat diffusion PDE is
shown in Fig. 12.15a. Additionally in Fig. 12.15b the variations of the control input
term ∂φ(x,t)
∂ x u(t) are shown. Moreover, in Figs. 12.16, 12.17, 12.18, 12.19, 12.20,
and 12.21 diagrams are provided about the tracking accuracy of the proposed control
method at specific points of the discretization grid. It can be noticed again that the
state estimation-based control method resulted in fast and accurate tracking of the
reference setpoints.
φ

Fig. 12.15 Control of the arc welding process. a Controlled distribution φ(x, t) of the nonlinear
heat diffusion PDE tracking a sinusoidal setpoint. b Control input ∂φ(x,t)
∂ x u(t)
640 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
10 10 10 10

5 5 5 5
1

6
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
3

8
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.16 a Grid points p1 to p4 and b grid points p5 to p8 of the nonlinear heat diffusion PDE:
tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

(a) (b)
10 10 10 10

5 5 5 5
10

13

14
9

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
11

12

15

16

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.17 a Grid points p9 to p12 and b grid points p13 to p16 of the nonlinear heat diffusion
PDE: tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

12.8 Fault Detection and Isolation in Distributed Parameter

Systems

12.8.1 Overview

Fault diagnosis in distributed parameter systems is a complicated problem that has

been little explored up to now. A reason for this is that state estimation methods used
12.8 Fault Detection and Isolation in Distributed Parameter Systems 641

(a) (b)
10 10 10 10

5 5 5 5
17

18

21

22
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
19

20

23

24
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.18 a Grid points p17 to p20 and b Grid points p21 to p24 of the nonlinear heat diffusion
PDE: tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

(a) (b)
10 10 10 10

5 5 5 5
25

26

29

30

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
27

28

31

32

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.19 a Grid points p25 to p28 and b grid points p29 to p32 of the nonlinear heat diffusion
PDE: tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

for residual generation in distributed parameter systems and in infinite dimensional

systems described by partial differential equations are much more complicated than
state estimation methods for lumped parameter systems [42, 467, 559, 562, 592]. Of
particular interest is state estimation of wave-type nonlinear phenomena, appearing
in several engineering applications [81, 124, 185, 198]. The present section treats
642 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
10 10 10 10

5 5 5 5
33

34

37

38
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
35

36

39

40
0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.20 a Grid points p33 to p36 and b Grid points p37 to p40 of the nonlinear heat diffusion
PDE: tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

(a) (b)
10 10 10 10

5 5 5 5
41

42

45

46

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
10 10 10 10

5 5 5 5
43

44

47

48

0 0 0 0

−5 −5 −5 −5
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.21 a Grid points p41 to p44 and b grid points p45 to p48 of the nonlinear heat diffusion
PDE: tracking of the reference setpoints (red line) by the value of the distribution φ(x, t) (blue line)

the problem of estimation and fault diagnosis for 1D nonlinear infinite dimensional
systems, which are described by Partial Differential Equations (PDEs) of the wave
type.
12.8 Fault Detection and Isolation in Distributed Parameter Systems 643

At the first stage the dynamics of the PDE model is computed through a state
estimator that processes measurements from a small number of sensors. To this end
the following steps are applied. Using the method for numerical solution of the PDE
through discretization the initial partial differential equation is decomposed into a
set of nonlinear ordinary differential equations with respect to time [388]. Next,
each one of the local models associated with the ordinary differential equations is
transformed into a model of the linear canonical (Brunovsky) form through a change
of coordinates (diffeomorphism) which is based on differential flatness theory. This
transformation provides an extended model of the nonlinear PDE for which state
estimation is possible by the application of the standard Kalman Filter recursion [55,
152, 286, 335, 362, 465]. As explained in Sects. 12.2 and 12.3, unlike other nonlinear
estimation methods (e.g., Extended Kalman Filter) the application of the standard
Kalman Filter recursion to the linearized equivalent of the nonlinear PDE system does
need the computation of Jacobian matrices and partial derivatives [408, 427, 437].
At the second stage, the chapter proposes the local statistical approach to fault
diagnosis for detecting changes and faults in the distributed parameter system [31, 32,
36, 414, 416, 600]. Residuals are generated by comparing the outputs measured from
the distributed parameter system against the outputs obtained from the Derivative-
free nonlinear Kalman Filter. The processing of these differences through the local
statistical approach to fault diagnosis provides clear indications about the existence
of incipient changes in the model of the monitored PDE. Fault diagnosis with the
Local Statistical Approach has two significant advantages: (i) it provides a credible
criterion (χ 2 test) to detect if faults have taken place in the distributed parameters
system. This criterion is more efficient than the normalized square error and mean
error tests since it employs the modeling error derivative and records the tendency
for change. Thus early change detection for the filter’s parameters becomes possible
(ii) it recognizes the parameters of the PDE model which are responsible for the
deviation of the filter’s estimates from the real output of the monitored dynamical
system.

12.8.2 Estimation of Nonlinear Wave Dynamics

As explained in Sect. 12.2.5 the nonlinear wave dynamics of Eq. (12.16) can be
transformed into the state-space form given in Eqs. (12.25) and (12.26). Thus, in
matrix form one has the following state-space description of the system

ỹ˙ = A ỹ + Bv
(12.63)
z̃ = C ỹ
644 12 Differential Flatness Theory for Distributed Parameter Systems

Denoting a = Dx K
2 and b = − Dx 2 , the initial description of the system given in
2K

Eq. (12.25) is rewritten as follows:

⎛ ⎞ ⎛ ⎞
0 1 0 0 0 0 0 ··· 0 0 0 0 0 0 0 0 0 ··· 0 0
⎛ ⎞ ⎛ ⎞
ẏ1,1 ⎜b 0 a 0 0 0 0 · · · 0 0 0 0 0 0 ⎟ y1,1 ⎜ 1 0 0 ··· 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜ ẏ2,1 ⎟ ⎜ 0 0 0 1 0 0 0 · · · 0 0 0 0 0 0 ⎟ ⎜ y2,1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎛ ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ ẏ1,2 ⎟ ⎜a 0 b 0 a 0 0 · · · 0 0 0 0 0 0⎟ ⎜ y1,2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ2,2 ⎟ ⎜ 0 0 0 0 0 1 0 · · · 0 0 0 0 0 0 ⎟ ⎜ y2,2 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ v3 ⎟
⎜ · · · ⎟ = ⎜0 0 a 0 b 0 a · · · 0 0 0 0 0 0 ⎟ ⎜ · · · ⎟ + ⎜ 0 0 1 · · · 0 0 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ1,N −1 ⎟ ⎜ · · · · · · · · · · · · · · · · · · · · · ⎟ ⎜ y1,N −1 ⎟ ⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎝v N −1 ⎠
⎜ ẏ2,N −1 ⎟ ⎜ 0 0 0 0 0 0 0 · · · 0 0 0 1 0 0 ⎟ ⎜ y2,N −1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎝ ẏ1,N ⎠ ⎜0 0 0 0 0 0 0 · · · a 0 b 0 a 0⎟ ⎝ y1,N ⎠ ⎜ 0 0 0 · · · 1 0 ⎟ v N
⎜ ⎟ ⎜ ⎟
ẏ2,N ⎝0 0 0 0 0 0 0 · · · 0 0 0 0 0 1⎠ y2,N ⎝ 0 0 0 ··· 0 0 ⎠
0 0 0 0 0 0 0 ··· 0 0 a 0 b 0 0 0 0 ··· 0 1
(12.64)

The associated control inputs are defined again as

v1 = Δx K
2 φ0 + f (y1,1 )
v2 = f (y1,2 )
v3 = f (y1,3 )
(12.65)
···
v N −1 = f (y1,N −1 )
v N = f (y1,N )

By selecting measurements from a subset of points x j j ∈ [1, 2, . . . , m], the associ-

ated observation (measurement) equation remains as in Eq. (12.26), i.e.,
⎛
⎞
⎛ y1,1
⎞
⎛ ⎞ 1 0 0 ··· 0 0 ⎜ ⎟
⎜ y2,1 ⎟
z1 ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ y1,2 ⎟
⎜ z2 ⎟ ⎜ ⎟⎜ ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ y2,2 ⎟ (12.66)
⎝· · ·⎠ ⎜ ⎟ ⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎜ ⎟
0 0 0 · · · 0 0 ⎝ y1,N ⎠
y2,N

For the linear description of the system in the form of Eq. (12.64) one can perform
estimation using the standard Kalman Filter recursion. The discrete-time Kalman
filter can be decomposed into two parts: (i) time update (prediction stage), and (ii)
measurement update (correction stage), as described in Eqs. (12.34) and (12.35).
Therefore, by taking measurements of φ(x, t) at time instant t at a small number
of measuring points j = 1, . . . , n 1 it is possible to estimate the complete state vector,
i.e., to get values of φ in a mesh of points that covers efficiently the variations of
φ(x, t). By processing a sequence of output measurements of the system, one can
12.8 Fault Detection and Isolation in Distributed Parameter Systems 645

obtain local estimates of the state vector ŷ. The measuring points (active sensors)
can vary in time provided that the observability criterion for the state-space model
of the PDE holds (see for example Fig. 12.3).

12.8.3 Equivalence Between Kalman Filters and Regressor

Models

12.8.3.1 Equivalence Between the Standard Kalman Filter and Linear

Regressor Models

For fault diagnosis purposes it is convenient to turn the Kalman Filter model of dis-
tributed parameter systems into equivalent ARMAX (autoregressive moving average
model with auxiliary input) models. An ARMAX model is an input–output model
of the form [31]
A(z)Yk = C(z)Uk + B(z){εk } (12.67)

A, B, and C are polynomial matrices in the backwards shift operator z −1 :

p
A(z) = A0 − i=1 Ai z −i
q
B(z) = j=0 B j z − j (12.68)

C(z) = ll=1 Cl z −l

such that A has nonsingular constant term A0 and where εk is a white noise sequence
with covariance matrix R. A state-space model and particularly the Kalman Filter
estimator can be written in the form of an ARMAX model. For linear systems, the
Kalman Filter (for the single-input case) can be written in the form [31]
⎛ ⎞ ⎛ ⎞⎛ ⎞
x̂1 (k + 1) α1 10 · 0 x̂1 (k)
⎜ x̂2 (k + 1)⎟ ⎜ · · · · ·⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ x̂2 (k)⎟
⎜ ··· ⎟ = ⎜ · · · · ⎜
·⎟ ⎜ · · · ⎟
⎟
⎜ ⎟ ⎜ ⎟+
⎝ ··· ⎠ ⎝ 0 · · · 1⎠ ⎝ · · · ⎠
x̂n (k + 1) αn · · · 0 x̂n (k)
⎛ ⎞ ⎛ ⎞ (12.69)
g1 κ1 (k)
⎜·⎟ ⎜ · ⎟
⎜ ⎟ ⎜ ⎟
⎜
+ ⎜ · ⎟ Uk + ⎜
⎟
⎜ · ⎟ εk
⎟
⎝·⎠ ⎝ · ⎠
gn κn (k)
 
Ŷ (k) = 1 0 · · 0 X̂ (k − 1) + J Uk + εk (12.70)
646 12 Differential Flatness Theory for Distributed Parameter Systems

Using successive substitutions this can be rewritten as a time-varying ARMAX

model:
A(z)Yk = C(z)Uk + B(k, z)εk (12.71)

where
A(z) = 1 − α1 z −1 − · · · − αn z −n
C(z) = g1 z −1 + · · · + gn z −n + J A(z −1 )
(12.72)
B(κ, z) = 1 + [κ1 (k − 1) − α1 ]z −1 + · · · +
+[κn (k − n) − αn ]z −n

Matrix B(κ, z) is time-varying because the Kalman Filter gain K k is time-varying.

But, under the conditions of the stability theorem, K and B are asymptotically con-
stant. Thus, the ARMAX description of the Kalman Filter becomes

A(z)Yk = C(z)Uk + B(z)εk (12.73)

This approach also holds for multi-input systems and one can transform again the
state-space representation into an ARMAX model.

12.8.4 Change Detection with the Local Statistical Approach

12.8.4.1 The Global χ 2 Test for Change Detection

Fault diagnosis for distributed parameter systems is based on the processing of the
residuals, i.e., of the differences between the outputs of the PDE model and the outputs
of the associated Kalman Filter. First, the residual ei is defined as the difference
between the Kalman filter output ŷi and the physical system output yi , i.e., ei =
ŷi − yi . It is also acceptable to define the residual as the difference between the
Kalman Filter output and the exact model output, where the exact model replaces the
physical system and has the same number of parameters as the Kalman Filter (see
Fig. 12.22). The partial derivative of the residual square is:

1 ∂ei
2
H (θ, yi ) = 2 ∂θ = ei ∂∂θŷi (12.74)

The local statistical approach to fault diagnosis is a statistical method of fault diagno-
sis which can be used for consistency checking of the fuzzy Kalman Filter. Based on
a small parametric disturbance assumption, the proposed FDI method aims at trans-
forming complex detection problems concerning a parameterized stochastic process
into the problem of monitoring the mean of a Gaussian vector. The local statistical
approach consists of two stages : (i) the global test which indicates the existence of a
change in some parameters of the distributed parameter system, (ii) the diagnostics
tests (sensitivity or min–max) which isolate the parameter affected by the change.
The method’s stages are analyzed first, following closely the method presented in
[31, 600].
12.8 Fault Detection and Isolation in Distributed Parameter Systems 647

Fig. 12.22 Residual

between the output of the
distributed parameter system
and the Kalman Filter that
models the PDE dynamics in
the fault-free case

As shown in Fig. 12.22 the concept of this FDI technique is as follows: there is
a Kalman Filter (ARMAX) model that represents the unchanged PDE dynamics. At
each time instant the output of the aforementioned reference model is compared to the
output of the Kalman Filter (ARMAX) model that represents the actual dynamics of
the PDE. The difference ei between these two output measurements is called residual.
The statistical processing of a sufficiently large number of residuals through an FDI
method provides an index variable that is compared against a fault threshold and
which can give early indication about deviation of the PDE model from its fault-
free condition. Under certain conditions (detectability of changes) the proposed FDI
method enables also fault isolation, i.e., to identify the source of fault within the
distributed parameter system.
Considering the representation of the Kalman Filter as an ARMAX model, the
partial derivative of the residual square is:

1 ∂ei
2
H (θ, ŷi ) = 2 ∂θ = ei ∂∂θŷi (12.75)

where θ is the vector of model’s parameters. Vector H having as elements the above
H (θ, ŷi ) is called primary residual. The gradient of the output with respect to the
ARMAX model parameters are given by

∂ ŷ
= xi (12.76)
∂wi

Next, having calculated the partial derivatives of Eq. (12.76), the rows of the Jacobian
matrix J are found by 
∂ ŷk (θ ) 
J (θ0 , ŷk ) = (12.77)
∂θ  θ=θ0
648 12 Differential Flatness Theory for Distributed Parameter Systems

where θ0 represents the nominal value of the parameters. The problem of change
detection with the χ 2 test consists of monitoring a change in the mean of the Gaussian
variable which for the one-dimensional parameter vector θ is formulated as

1  ∂ ŷk
N
X=√ ek ∼N(μ, σ 2 ) (12.78)
N i=1 ∂θ

where ŷk is the output of the Kalman Filter (ARMAX) model generated by the input
pattern xk , ek is the associated residual, and θ is the vector of the model’s parameters.
It is noted that X is the monitored parameter for the FDI test, which means that when
the mean value of X is 0 the system is in the fault-free condition, while when the
mean value of X has moved away from 0 the system is in a faulty condition. For a
multivariable parameter vector θ should hold X ∼N(Mδθ, S), where S denotes the
covariance matrix of X . In order to decide if the system (Kalman Filter) is in fault-
free operating conditions, given a set of data of N measurements, let θ∗ be the value
of the parameters vector μ minimizing the RMSE. The notation is introduced only
for the convenience of problem formulation, and its actual value does not need to be
known. Then the model validation problem amounts to make a decision between the
two hypotheses:
H0 : θ∗ = θ0
H1 : θ∗ = θ0 + √1 δθ (12.79)
N

where δθ
= 0. It is known from the Central Limit Theorem that for a large data sam-
ple, the normalized residual given by Eq. (12.78) asymptotically follows a Gaussian
distribution when N → ∞ [31, 32, 36, 414, 416, 600]. More specifically, the hypoth-
esis that has to be tested is:

H0 : X ∼ N(0, S)
H1 : X ∼ N(Mδθ, S)

where M is the sensitivity matrix (see Eq. (12.80)), δθ is the change in the parameters’
vector, and S is the convariance matrix (see Eq. (12.81)). The product Mδθ denotes
the new center of the monitored Gaussian variable X , after a change on the system’s
parameter θ . The sensitivity matrix M of √1 X is defined as the mean value of the
N
partial derivative with respect to θ of the primary residual defined in Eq. 12.75, i.e.,
∂
E{ ∂θ H (θ, ŷk )} and is approximated by [31]:

∂ 1 N
M(θ0 ) ∂θ N k=1 H (θ0 , ŷk ) N
1
JT J (12.80)
12.8 Fault Detection and Isolation in Distributed Parameter Systems 649

The covariance matrix S is defined as E{H (θ, yk )H T (θ, ŷk+m )}, m = 0, ±1, . . .
and is approximated by [32]:
N
S =  k=1 [H (θ0 , ŷk )H T (θ0 , ŷk )]+
I 1  N −m
+ m=1 N −m k=1 [H (θ0 , ŷk )H (θ0 , ŷk+m )+
T (12.81)
+H (θ0 , ŷk+m )H (θ0 , ŷk )]
T

where an acceptable value for I is 3. The decision tool is the likelihood ratio
p
s(X ) = ln pθθ1 (x)
(x)
(12.82)
0

where
T −1
pθ1 (X ) = e[X −μ(X )] S [X −μ(X )]
T −1 (12.83)
pθ0 (X ) = e X S X .

The center of the Gaussian distribution of the changed system is denoted as μ(X ) =
Mδθ where δθ is the change in the parameters vector. The Generalized Likelihood
Ratio (GLR) is calculated by maximizing the likelihood ratio with respect to δθ [32].
This means that the most likely case of parameter change is taken into account. This
gives the global χ 2 test t:

t = X T S −1 M(M T S −1 M)−1 M T S −1 X (12.84)

Since X asymptotically follows a Gaussian distribution, the statistics defined in

Eq. (12.84) follows a χ 2 distribution with n degrees of freedom. Mapping the change
detection problem to this χ 2 distribution enables the choice of the change threshold.
Assume that the desired probability of false alarm is α then the change threshold λ
should be chosen from the relation
∞
λ χn2 (s)ds = α, (12.85)

where χn2 (s) is the probability density function (p.d.f.) of a variable that follows the
χ 2 distribution with n degrees of freedom.

12.8.4.2 Statistical Fault Isolation with the Sensitivity Test

Fault isolation is needed to identify the source of faults and parametric changes
in the PDE system. This means that the fault diagnosis method should be able to
find out, among the complete set of parameters of the PDE, which are the ones
that are subject to change with respect to their nominal values. A first approach to
change isolation is to focus only on a subset of the parameters while considering
that the rest of the parameters remain unchanged [32]. The parameters vector η can
be written as η = [φ, ψ]T , where φ contains those parameters to be subject to the
650 12 Differential Flatness Theory for Distributed Parameter Systems

isolation test ,while ψ contains those parameters to be excluded from the isolation
test. Mφ contains the columns of the sensitivity matrix M which are associated with
the parameters subject to the isolation test. Similarly, Mψ contains the columns of
M that are associated with the parameters to be excluded from the sensitivity test.
Assume that among the parameters η, it is only the subset φ that is suspected
to have undergone a change. Thus η is restricted to η = [φ, 0]T . The associated
columns of the sensitivity matrix are given by Mφ and the mean of the Gaussian to
be monitored is μ = Mφ φ, i.e.,

μ = M Aφ, A = [0, I ]T (12.86)

Matrix A is used to select the parameters that will be subject to the fault isolation
test. The rows of A correspond to the total set of parameters while the columns of A
correspond only to the parameters selected for the test. Thus the fault diagnosis (χ 2 )
test of Eq. (12.84) can be restated as:

tφ = X T S −1 Mφ (MφT S −1 Mφ )−1 MφT S −1 X (12.87)

12.8.4.3 Statistical Fault Isolation with the Min–Max Test

In this approach the aim is to find a statistic that will be able to detect a change on
the part φ of the parameters vector η and which will be robust to a change in the
nonobserved part ψ [32]. Assume the vector partition η = [φ, ψ]T . The following
notation is used:  
I I
M T S −1 M = ϕϕ ϕψ (12.88)
Iψϕ Iψψ
 T    
ϕ Iϕϕ Iϕψ ϕ
γ = · · (12.89)
ψ Iψϕ Iψψ ψ

where S is the previously defined covariance matrix. The min–max test aims to
minimize the noncentrality parameter γ with respect to the parameters that are not
suspected for change. The minimum of γ with respect to ψ is given for:
−1
ψ ∗ = arg min γ = ϕ T (Iϕϕ − Iϕψ Iψψ Iψϕ )ϕ (12.90)
ψ

and is found to be
−1
γ ∗ = min γ = ϕ T (Iϕϕ − Iϕψ Iψψ Iψϕ )ϕ =
ψ
 T     (12.91)
ϕ Iϕϕ Iϕψ ϕ
= −1 −1
−Iψψ Iψϕ ϕ Iψϕ Iψψ −Iψψ Iψϕ ϕ
12.8 Fault Detection and Isolation in Distributed Parameter Systems 651

Table 12.1 Stages of the local statistical approach for FDI

1. Generate the residuals partial derivative given by Eq. (12.75)
2. Calculate the Jacobian matrix J given by Eq. (12.77)
3. Calculate the sensitivity matrix M given by Eq. (12.80)
4. Calculate the covariance matrix S given by Eq. (12.81)
5. Apply the χ 2 test for change detection of Eq. (12.84)
6. Apply the change isolation tests of Eqs. (12.87) or (12.96)

which results in
−1 −1
γ ∗ = ϕ T {[I, −Iϕψ Iψψ ]M T Σ −1 } Σ −1 {Σ −1 M[I, −Iϕψ Iψψ ]}ϕ (12.92)

The following linear transformation of the observations is considered:

 
−1
X φ∗ = I, −Iϕψ Iψψ M T Σ −1 X (12.93)

The transformed variable X φ∗ follows a Gaussian distribution N (μ∗φ , Iφ∗ ) with mean:

μ∗ϕ = Iϕ∗ ϕ (12.94)

and with covariance:

−1
Iϕ∗ = Iϕϕ − Iϕψ Iψψ Iψϕ (12.95)

The max–min test decides between the hypotheses:

H0∗ : μ∗ = 0
H1∗ : μ∗ = Iϕ∗ ϕ

and is described by:

τϕ∗ = X ϕ∗ T Iϕ∗ −1 X ϕ∗ (12.96)

The stages of fault detection and isolation (FDI), for the PDE system, with the use
of the local statistical approach are summarized in the following table (Table 12.1).

12.8.5 Simulation Tests

12.8.5.1 Fault Detection in a Sensor’s Network that Monitors a Wave

PDE

The proposed filtering scheme was tested in estimation and fault diagnosis for a wave
equation of the form of Eq. (12.16) under unknown boundary conditions. Nonlinear
652 12 Differential Flatness Theory for Distributed Parameter Systems

1D wave-type partial differential equations of this type appear in models of coupled

oscillators. One can consider for example the nonlinear PDE (forced damped sine-
Gordon equation) [186, 466]

∂φ
+ c ∂φ ∂ φ
2
∂t − K ∂ x 2 + εsin(φ) = l
(12.97)
∂t 2

where c, ε, and l are constants.

To perform state estimation of the distributed parameter system of the Eq. (12.97)
a grid consisting of n = 50 points was considered. The number of measurement
points was n 1 = 25. In general, the number of measurements of the state variable
x1 = φ(x, t) which can be used by the estimation algorithm is n 1 ≤ n, where n is
the number of grid points, and the criterion to select the number of grid points where
measurements will be taken is to maintain the system’s observability.
In Fig. 12.23 the monitored wave function φ(x, t) and the associated first deriv-
ative in time φ̇(x, t) are depicted. Indicative results about the estimation obtained
at local grid points is given in Figs. 12.24 and 12.27. It was assumed that erroneous
measurements were provided by one specific sensor used in the monitoring of the
PDE system (e.g., sensor monitoring output 22 of the PDE system) (Figs. 12.25 and
12.26).
The experimental results show that state estimates in the vicinity of the sensor
subjected to fault (sensor monitoring output 22, i.e., state vector element 85) exhibit
significant deviations comparing to the real values of the state vector. As moving far
from the faulty sensor (e.g., state vector elements 61–64) the estimated and the real
values of the state vector elements exhibit smaller differences. Thus, the comparison
of the measurements recorded by the sensors against the estimated values from
the Derivative-free nonlinear Kalman Filter provides a clear indication about the
malfunctioning sensor. Thus it becomes possible to isolate this sensor from the rest
of the sensors set.

Fig. 12.23 The monitored wave function φ(x, t)

12.8 Fault Detection and Isolation in Distributed Parameter Systems 653

(a) (b)
20 20 20 20

10 10 10 10
61

62

61

62
0 0 0 0

−10 −10 −10 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
20 10 20 10

5 5
10 10
63

64

63

64
0 0

0 0
−5 −5

−10 −10 −10 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.24 Monitoring of state vector elements y61 to y64 of the DPS, a when all sensors are
fault-free, b when a fault appears at sensor monitoring output 22, i.e., state variable y85

(a) (b) 40
40 20 20

20 10 20 10
81

82

81

82

0 0 0 0

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
40 10 40 10

5 5
20 20
83

84

83

84

0 0

0 0
−5 −5

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.25 Monitoring of state vector elements y81 to y84 of the DPS, a when all sensors are
fault-free, b when a fault appears at sensor monitoring output 22, i.e., state variable y85

12.8.5.2 Change Detection in the Distributed Parameter System

It is also important to detect changes in the coefficients of the distributed parameter

model. To this end the following state-space equation of the nonlinear wave PDE
given in Eq. (12.97) is used. The state-space equation is written in the form of an
ARMAX model which enables fault diagnosis with the use of the local statistical
approach.
654 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
40 20 60 20

40
20 10 10
85

86

85

86
20

0 0 0
0

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
40 10 40 10

5 5
20 20
87

88

87

88
0 0

0 0
−5 −5

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.26 Monitoring of state vector elements y85 to y88 of the DPS, a when all sensors are
fault-free, b when a fault appears at sensor monitoring output 22, i.e., state variable y85

(a) (b)
40 20 40 20

20 10 20 10
89

90

89

90

0 0 0 0

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
40 10 40 10

5 5
20 20
91

92

91

92

0 0

0 0
−5 −5

−20 −10 −20 −10

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 12.27 Monitoring of state vector elements y89 to y92 of the DPS, a when all sensors are
fault-free, b when a fault appears at sensor monitoring output 22, i.e., state variable y85

Without loss of generality it is assumed that the wave propagation coefficient K

appearing in Eq. (12.97) remains the same at all points of the grid. Using only the
last subsystem in description of dynamics of the PDE one has

ẏ1,N = y2,N
(12.98)
ẏ2,N = K
φ
Δx 2 N +1
− 2K
y
Δx 2 1,N
+ ΔxK
2 y1,N −1 + f (y1,N )
12.8 Fault Detection and Isolation in Distributed Parameter Systems 655

Defining v N = Δx K
2 φ N +1 + Δx 2 y1,N −1 + f (y1,N ) the following discrete-time
K

description of the system is obtained

      
y1,N (k + 1) 1 Ts y1,N (k) 0
= + v N (k) (12.99)
y2,N (k + 1) − Dx
2K
2 T s 1 y2,N (k) Ts

 
  y1,N (k)
z 1,N (k) = 1 0 (12.100)
y2,N (k)

The associated ARMA model is found to be

Ts2 2K
z 1,N (k + 2) = 2z 1,N (k + 1) − (1 + Dx 2
)z 1,N (k) + v N (k) (12.101)

In a similar manner, the discrete-time model of the Kalman Filter is written as

    
ŷ1,N (k + 1) 1 Ts ŷ1,N (k)
= +
ŷ2,N (k + 1) − Dx
2K
2Ts 1 ŷ (k)
    2,N (12.102)
0 κ (k)
+ v(k) + 1 ê(k)
Ts κ2 (k)
 
  ŷ1,N (k)
ẑ(k) = 1 0 (12.103)
ŷ2,N (k)

and the associated ARMAX model is found to be

ẑ 1,N (k + 2) = 2ẑ 1,N (k + 1) − (1 + Ts2 Dx

2K
2 )ẑ 1,N (k)+
(12.104)
+Ts v N (k) + κ1 (k)ê(k + 1) + (κ1 (k) − κ2 (k)Ts )ê(k)

where ê(k) is the estimation error (innovation). The output of the ARMAX model
can be also written in the product form

ẑ 1,N (k + 1) = w(k)·X T (k) (12.105)

where the weights vector is defined as

 
w(k) = 2, −(1 + Ts2 Dx
2K
2 ), Ts , κ1 (k), (κ1 (k) − κ2 (k)Ts ) (12.106)

and the regressor vector is defined as

 
X (k) = ẑ 1,N (k), ẑ 1,N (k − 1), v N (k − 1), ê(k), ê(k − 1) (12.107)

Indicative results about the performance of the global χ 2 in detection of incipient

changes of parameter K of the nonlinear wave PDE are depicted in Fig. 12.28. The
fault threshold is set equal to the number of parameters in the associated ARMAX
656 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b) 200

200

180 180

160 160

140 140

120 120
χ2

χ2
100 100

80 80

60 60

40 40

20 20

0 0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0 0.05 0.1 0.15 0.2 0.25

ΔK% ΔK%

Fig. 12.28 Detection of incipient changes in the coefficient K of the nonlinear wave equation using
the χ 2 test, a nominal value K = 0.04050, b nominal value K = 0.05050 (The red line denotes
the fault threshold, while the blue dots denote the value of the change detection test)

model, i.e., η = 5. It can be observed that the proposed FDI method was capable
of detecting changes in parameter K which were less than 1 % of the coefficient’s
nominal value. For small deviations from the parameter’s nominal value the χ 2 test
obtained a value that was several times larger than the fault threshold.

12.9 Application to Condition Monitoring of Civil and

Mechanical Structures

12.9.1 Overview

In this section, differential flatness theory-based methods of filtering will be used for
condition monitoring of civil and mechanical structures. Such systems are known
to have dynamics that approximate the one of distributed parameter systems. Dam-
ages in civil or mechanical structures (e.g., buildings, oil rigs, etc.) cannot always
be detected by visual inspection. Therefore, it is necessary to develop elaborated
methods for structural health monitoring that will be capable of diagnosing the
existence of parametric changes and failures [64, 519, 563]. Previous results in the
area of Kalman Filter-based structural condition monitoring can be found in [234,
273, 274, 611]. In the present section, the development of a new structural health
monitoring approach is proposed making use of nonlinear Kalman Filtering and of
statistical signal processing.
The multi-DOF building or mechanical structure is modeled as a set of coupled
nonlinear oscillators that can be subjected to external excitation. First, it is shown that
the dynamic model of the monitored structure is differentially flat and admits static
12.9 Application to Condition Monitoring of Civil and Mechanical Structures 657

feedback linearization [55, 152, 495, 427, 464]. It is also shown that the dynamic
model can be written in the linear Brunovsky canonical form for which a state
feedback controller can be easily designed, thus also enabling active control of the
structure. Next, to estimate the building’s or mechanical structure’s motion character-
istics from indirect measurements the previously analyzed Derivative-free nonlinear
Kalman Filtering is introduced [408, 436, 437]. The considered filter makes use of
the aforementioned exact linearization transformation of the structure’s dynamical
model, in accordance to differential flatness theory, and of an inverse transformation
that enables to obtain estimates for the state vector elements of the initial model. The
filter provides estimates of the state vector elements of the structure which cannot be
directly measured. Moreover, by redesigning the Derivative-free nonlinear Kalman
Filter as a disturbance observer, it is shown that it is possible to estimate unknown
external inputs applied to the structure.
The response of the structure is recorded through suitable sensors (in the form of a
wireless sensors network deployed at specific measurement points) and is compared
against the response generated by nonlinear Kalman Filtering under the assumption
of a damage-free model (to improve accuracy it is possible to apply a sensor fusion-
based implementation of the Kalman Filter). By comparing the two signals, residuals
sequences are generated. The processing of the residuals with the use of statistical
decision making criteria provides an indication about the existence of parametric
changes (damages) in the structure, which otherwise could not have been detected
[31, 414, 600]. Thus, by applying fault detection tests based on the χ 2 distribution it
can be concluded if the structure remains healthy and if the nominal parameter values
for its model still hold. Otherwise, a failure can be detected. Moreover by applying
the χ 2 tests in compartments of the building’s model, the faulty substructure can be
also isolated.

12.9.2 Dynamical Model of the Building—Mechanical

Structure

The model of the n-storey building (Fig. 12.29) is that of coupled oscillators. In a first
modeling approach these oscillators are taken to be linear. Considering an inertial
measurement system, the terms xi i = 1, 2, . . . , n denote the displacement of the ith
floor, the terms ẋi i = 1, 2, . . . , n denote the linear velocity of the ith floor, while the
terms ẍi i = 1, 2, . . . , n denote the linear acceleration of the ith floor. Equivalently,
the term x g denotes the displacement of the ground, the term ẋ g denotes the linear
velocity of the ground, and the term ẍ g denotes the linear acceleration of the ground
(e.g., in case of earthquake excitation).
658 12 Differential Flatness Theory for Distributed Parameter Systems

Fig. 12.29 Diagram of the n-storey building with elasticity and dissipation elements

Ground floor
m 1 ẍ1 + c1 (ẋ1 − ẋ g ) + k1 (xi − x g ) + c2 (ẋ1 − ẋ2 ) + k2 (x1 − x2 ) + f 1 (x, x g ) = u 1
··· ···

Middle floor i
m i ẍi + ci (ẋi − ẋi−1 ) + ki (xi − xi−1 ) + ci+1 (ẋi − ẋi+1 ) + ki+1 (xi − xi 1 ) + f i (x) = u i

··· ···
Top floor n
m n ẍn + cn (ẋn − ẋn−1 ) + kn (xn − xn−1 ) + f n (x) = u n (12.108)

Finally, effects of nonlinear friction and nonlinear forces can be considered to affect
the displacement of each floor, and these are described by the terms f i (x), i =
1, 2, . . . , n. In case that active control is exerted on the building (e.g., through actua-
tors which compensate for the displacement of each floor), the model is completed by
the inclusion of the associated control input terms u i i = 1, 2, . . . , n. It is considered
that full actuation is possible, although the results can be generalized for the case of
underactuation.
12.9 Application to Condition Monitoring of Civil and Mechanical Structures 659

12.10 Differential Flatness of the Multi-DOF Building’s

Structure

As explained in Sect. 12.9.2, the complete dynamical model of the building is given by

m 1 ẍ1 + c1 (ẋ1 − ẋ g ) + k1 (x1 − x g ) + c2 (ẋ1 − ẋ2 ) + k2 (x1 − x2 ) + f 1 (x) = u 1

m 2 ẍ2 + c2 (ẋ2 − ẋ1 ) + k2 (x2 − x1 ) + c3 (ẋ2 − ẋ3 ) + k3 (x2 − x3 ) + f 2 (x) = u 2
···
···
m i ẍi + ci (ẋi − ẋi−1 ) + ki (xi − xi−1 ) + ci+1 (ẋi − ẋi+1 ) + ki+1 (xi − xi+1 ) + f i (x) = u i
···
···
m n ẍn + cn (ẋn − ẋn−1 ) + kn (xn − xn−1 ) + f n (x) = u n
(12.109)

where x is the state vector x = [x1 , x2 , . . . , xn , ẋ1 , ẋ2 , . . . , ẋn ]. Next, the following
notation for the system’s state variables is used: x1,1 = x1 , x1,2 = ẋ1 , x2,1 =
x2 , x2,2 = ẋ2 , . . ., xi,1 = xi , xi,2 = ẋi , . . . xn−1,1 = xn−1 , xn−1,2 = ẋn−1 , xn,1 =
xn , xn,2 = ẋn .
Then the model of the building can be written in the form

ẋ1,1 = x1,2
c1 k1 c2
ẋ1,2 = − (x1,2 − ẋ g ) − (x1,1 − x g ) − (x1,2 − x2,2 )−
m1 m1 m1
k2 1
− (x1,1 − x2,1 ) − f 1 (x) + u 1
m1 m1
ẋ2,1 = x2,2
c2 k2 c3
ẋ2,2 = − (x2,2 − x1,2 ) − (x2,1 − x3,1 ) − (x2,2 − x3,2 )−
m2 m2 m2
k3 1
− (x2,1 − x1,1 ) − f 2 (x) + u 2
m2 m2
··· (12.110)
ẋi,1 = xi,2
ci ki ci+1
ẋi,2 = − (xi,2 − xi−1,2 ) − (xi,1 − xi−1,1 ) − (xi,2 − xi+1,2 )−
mi mi mi
ki+1 1
− (xi,1 − xi+1,1 ) − f i (x) + u i
mi m1
···
ẋn,1 = xn,2
cn kn 1
ẋn,2 = − (xn,2 − xn−1,2 ) − (xn,1 − xn−1,1 ) − f n (x) + u n
mn mn mn
660 12 Differential Flatness Theory for Distributed Parameter Systems

The state vector of the system is

x = [x1,1 , x1,2 , x2,1 , x2,2 , . . . , xi,1 , xi,2 , . . . , xn−1,1 , xn−1,2 , xn,1 , xn,2 ] (12.111)

The flat output of the system is taken to be the vector

Y = [x1,1 , x2,1 , . . . , xi,1 , . . . , xn−1,1 , xn,1 ] (12.112)

From Eq. (12.110) and using x1,2 = ẋ1,1 , x2,2 = ẋ2,1 , . . ., xi,2 = ẋi,1 , . . ., xn−1,2 =
ẋn−1,1 , and xn,2 = ẋn,1 one has x1,2 = ẏ1 , x2,2 = ẏ2 , . . ., xi,2 = ẏi , . . . xn−1,2 =
ẏn−1 , and xn,2 = ẏn . Therefore, the state vector elements xi,2 , x2,2 , . . ., xi,2 , . . .,
xn−1,2 , xn,2 are expressed as functions of the flat outputs and its derivatives. Moreover,
by solving the even rows of Eq. (12.110) with respect to the control inputs u i , i =
1, 2, . . . , n one has again the control inputs u i that are functions of the flat output and
its derivatives. Consequently, the dynamical system of the building is a differentially
flat one. Next, it will be shown that the building’s model can be transformed into the
linear canonical (Brunovsky) form. The following control inputs are defined:

v1 = − mc11 (x1,2 − ẋ g ) − mk11 (x1,1 − x g ) − mc21 (x1,2 − x2,2 )−

− mk21 (x1,1 − x2,1 ) − m11 f 1 (x) + u 1
v2 = − mc22 (x2,2 − x1,2 ) − mk22 (x2,1 − x1,1 ) − mc32 (x2,2 − x3,2 )−
− mk32 (x2,1 − x3,1 ) − m12 f 2 (x) + u 2
··· (12.113)
vi = − mcii (xi,2 − xi−1,2 ) − mkii (xi,1 − xi−1,1 ) − cmi+1i (xi,2 − xi+1,2 )−
− kmi+1i (xi,1 − xi+1,1 ) − m11 f i (x) + u i
···
vn = − mcnn (xn,2 − xn−1,2 ) − mknn (xn,1 − xn−1,1 ) − m1n f n (x) + u n

Thus, the following description of the system is obtained:

ẋ1,1 = x1,2 ẋ1,2 = v1 ẋ2,1 = x2,2 ẋ2,2 = v2

··· ··· ··· ···
ẋi,1 = xi,2 ẋi,2 = vi ẋi+1,1 = xi+1,2 ẋi+1,2 = vi+1 (12.114)
··· ··· ··· ···
ẋn−1,1 = xn−1,2 ẋn−1,2 = vn−1 ẋn,1 = xn,2 ẋn,2 = vn

The above description can be also written in the state-space form

12.10 Differential Flatness of the Multi-DOF Building’s Structure 661

⎛ ⎞ ⎛ ⎞
0 1 0 0 ··· 00 ··· 0000 000 ··· 0 ··· 000
⎛ ⎞ ⎛ ⎞
ẋ1,1 ⎜0 0 0 0 · · · 00 ··· 0 0 0 0⎟ x1,1 ⎜1 0 0 ··· 0 ··· 0 0 0⎟
⎜ ⎟ ⎜ ⎟⎛ ⎞
⎜ ẋ1,2 ⎟ ⎜0 0 0 1 · · · 00 ··· 0 0 0 0⎟ ⎜ x1,2 ⎟ ⎜0 0 0 ··· 0 ··· 0 0 0 ⎟ v1
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ ẋ2,1 ⎟ ⎜0 0 0 0 · · · 00 ··· 0 0 0 0⎟ ⎜ ⎟ ⎜ ··· 0 ··· 0 0 0⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ x2,1 ⎟ ⎜0 1 0 ⎟ ⎜ v2 ⎟
⎜ ẋ2,2 ⎟ ⎜ · · · · · · ··· ··· ··· ⎟ ⎜ x2,2 ⎟ ⎜ · · · ··· ··· ··· ⎟ ⎟⎜ ··· ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎜ ⎟
⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ··· ⎟ ⎜ ⎟ ⎜ ··· ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ··· ⎟ ⎜ ··· ··· ··· ⎟⎜ ··· ⎟
⎜ ẋi,1 ⎟ ⎜0 0 0 0 · · · 01 ··· 0 0 0 0⎟ ⎜ xi,1 ⎟ ⎜0 0 0 ··· 0 ··· 0 0 0⎟ ⎜ vi ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟+⎜ ⎟⎜ ⎟
⎜ ẋi,2 ⎟ = ⎜0 0 0 0 · · · 00 ··· 0 0 0 0⎟ ⎜ ⎟ ⎜ ··· 0 ··· 0 0 0⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜ xi,2 ⎟ ⎜0 0 0 ⎟ ⎜ vi+1 ⎟
⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ··· ⎟ ⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ⎟ ⎜ ··· ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ẋn−1,1 ⎟ ⎜ · · · · · · ··· ··· ··· ⎟ ⎜ ⎟ ⎜ ··· ⎟ ⎜ ⎟
⎜ ⎟ ⎜ ⎟ ⎜xn−1,1 ⎟ ⎜ · · · ··· ··· ⎟⎜ ··· ⎟
⎜ẋ ⎟ ⎜ ⎟
0 1 0 0⎟ ⎜xn−1,2 ⎟
⎜ ⎜ ⎟
0 0 0⎟ vn−1 ⎠
⎝
⎜ n−1,2 ⎟ ⎜0 0 0 0 · · · 00 ··· ⎟ ⎜0 0 0 ··· 0 ···
⎝ ẋ ⎠ ⎜0 0 0 0 · · · ··· 0 0 0 0⎟ ⎝ x ⎠ ⎜0 0 0 ··· 0 ··· 0 1 0⎟
n,1 ⎜ 00 ⎟ n,1 ⎜ ⎟ vn
ẋn,2 ⎝0 0 0 0 · · · 00 ··· 0 0 0 1⎠ x ⎝0 0 0 ··· 0 ··· 0 0 0⎠
n,2
0 0 0 0 ··· 00 ··· 0000 0 0 0 ··· 0 ··· 001
(12.115)

while the state-space description of the system is

⎛ ⎞
x1,1
⎜ x1,2 ⎟
⎜ ⎟
⎛ ⎞ ⎛ ⎞⎜ x2,1 ⎟
x1,1 1 0 0 0 ··· 00 ··· 0000 ⎜ ⎜ ⎟
x2,2 ⎟
⎜ x2,1 ⎟ ⎜0 0 1 0 · · · 00 ··· ⎟ ⎜
0 0 0 0⎟ ⎜ ⎟
⎜ ⎟ ⎜ ··· ⎟
⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ··· ⎟⎜⎟ ⎜ ⎟
⎜ ⎟ ⎜ xi,1 ⎟
Y =⎜ ⎟ ⎜
⎜ xi,1 ⎟ = ⎜0 0 0 0 · · · 10 ··· ⎟ ⎜
0 0 0 0⎟ ⎜ ⎟
⎟ (12.116)
⎜ ··· ⎟ ⎜ ··· ··· ⎟ ⎜ xi,2 ⎟
⎜ ⎟ ⎜ ··· ··· ··· ⎟⎜ ⎟
⎝xn−1,1 ⎠ ⎝0 0 0 0 · · · ⎠ ⎜ ··· ⎟
00 ··· 1000 ⎜ ⎟
⎜ x n−1,1 ⎟
xn,1 0 0 0 0 ··· 00 ··· 0010 ⎜ ⎟
x
⎜ n−1,2 ⎟
⎝ xn,1 ⎠
xn,2

Thus the system is finally written in the state-space form

ẋ = Ax + Bv
(12.117)
y = Cx

It can be confirmed that the above model is observable and thus state estimation can
be performed with the use of a state observer. The selection of the observer’s gain K
can be carried out using the standard Kalman Filter recursion. Prior to this, matrices
A, B, and C are brought to the discrete-time form Ad , Bd , and Cd using common
discretization methods. The discrete-time Kalman Filter recursion is [31, 408, 414]
measurement update:

K (k) = P − (k)CdT [Cd P − (k)CdT + R]−1

x̂(k) = x̂ − (k) + K (k)[y(k) − ŷ(k)] (12.118)
P(k) = P − (k) − K (k)Cd P − (k)
662 12 Differential Flatness Theory for Distributed Parameter Systems

time update:
P − (k + 1) = Ad P(k)AdT + Q
(12.119)
x̂(k + 1) = Ad x̂(k) + Bd v(k)

12.10.1 Damage Detection with the Use of Statistical Criteria

12.10.1.1 Fault Detection

The residuals’ sequence (differences between the real output of the monitored struc-
ture and the one estimated by the Kalman Filter) is a discrete error process ek with
dimension m × 1 (here m = N ). Actually, it is a zero mean Gaussian white noise
process with covariance given by E k . A conclusion can be stated based on a measure
of certainty that the parameters of the dynamic model of the multistorey building
remain unchanged. To this end, the following normalized error square (NES) is
defined [442]
εk = ekT E k−1 ek (12.120)

The normalized error square follows a χ 2 distribution. An appropriate test for the
normalized error sum is to numerically show that the following condition is met
within a level of confidence (according to the properties of the χ 2 distribution which
are shown in Fig. 12.30)
E{εk } = m (12.121)

(a) (b)
0.14 0.5
p=2
0.45 p=4
0.12 p=6
0.4 p=8
p=10
0.1 0.35 p=12
p=14
p=16
0.3
0.08
pdf f(x)

p=18
pdf f(x)

p=20
0.25
p=24
0.06 p=28
0.2 p=32
p=36
0.04 0.15 p=40
p=44
0.1 p=48
0.02 p=52
0.05

0 0
0 10 20 30 40 50 0 10 20 30 40 50
x x

Fig. 12.30 a Probability density function of the χ 2 distribution for p = 6 degrees of freedom,
b Probability density function of the χ 2 distribution for several values of the degrees of freedom
(variable p)
12.10 Differential Flatness of the Multi-DOF Building’s Structure 663

This can be succeeded using statistical hypothesis testing, which are associated with
confidence intervals. A 95 % confidence interval is frequently applied, which is spec-
ified using 100(1 − a) with a = 0.05. Actually, a two-sided probability region is
considered cutting off two end tails of 2.5 % each. For M runs the normalized error
square that is obtained is given by

1 1 T
M M
ε̄k = εk (i) = ek (i)E k−1 (i)ek (i) (12.122)
M M
i=1 i=1

where εi stands for the ith run at time tk . Then M ε̄k will follow a χ 2 density with Mm
degrees of freedom. This condition can be checked using a χ 2 test. The hypothesis
holds, if the following condition is satisfied

ε̄k ∈ [ζ1 , ζ2 ] (12.123)

where ζ1 and ζ2 are derived from the tail probabilities of the χ 2 density. For exam-
ple, for m = 20 and M = 100 one has χ Mm 2 (0.025) = 1878 and χ 2 (0.975) =
Mm
2126. Using M = 100 one obtains ζ1 = χ Mm 2 (0.025)/M = 18.78 and ζ =
2
2 (0.975)/M = 21.26.
χ Mm

12.10.1.2 Fault Isolation

By applying the statistical test into the n decoupled substructures of the building
it is also possible to find out the substructure within the multistorey building that
has suffered the damage [204, 275, 519]. In the case of a single fault one has to
carry out n χ 2 statistical change detection tests, where each test is applied to the
substructure that comprises stores i − 1, i, and i + 1, i = 1, 2, . . . , n. Actually, out
of the n χ 2 statistical change detection tests, the two that exhibit the highest score
(or equivalently indicate the largest parameter deviation from the nominal value) are
those that identify the floor that has been subjected to damage (the faulty components
for this floor are the changed elasticity and damping coefficients).
In the case of multiple faults one can identify the subset of floors that have been
subjected to parametric change by applying the χ 2 statistical change detection test
according to a combinatorial sequence. This means that
 
n
= n!
k!(n−k)! (12.124)
k

tests have to take place, for all substructures in the building that comprise n, n − 1,
n −2, . . ., 2, 1 floors. Again the χ 2 tests that give the highest scores indicate the floors
which are most likely to have been subjected to structural damage. This approach
enables to assess the magnitude of stressing and parametric change that the building
664 12 Differential Flatness Theory for Distributed Parameter Systems

has undergone and to focus on the part that exhibits the most significant structural
damage.

12.10.2 Disturbances Estimation with the Derivative-Free

Nonlinear Kalman Filter

Next, it is considered that the system’s dynamics described in Eqs. (12.109) and
(12.110) is affected by additive input disturbances:

ẋ1,1 = x1,2 ẋ1,2 = v1 + d̃1 ẋ2,1 = x2,2 ẋ2,2 = v2 + d̃2

··· ···
ẋi,1 = xi,2 ẋi,2 = vi + d̃i ẋi+1,1 = xi+1,2 ẋi+1,2 = vi+1 + d̃i+1
··· ···
ẋn−1,1 = xn−1,2 ẋn−1,2 = vn−1 + d̃n−1 ẋn,1 = xn,2 ẋn,2 = vn + d̃n
(12.125)

It is considered that the dynamics of each perturbation term is described by its

nth order derivative, that is d̃ (n) = f d (t), and of the associated initial conditions.
However, the reconstruction of the signals d̃i i = 1, . . . , n will be performed with the
use of Kalman Filtering, and the convergence of the latter estimation method does
not depend on initial conditions. Therefore, initial conditions are finally unnecessary
for estimating the disturbance terms’ evolution in time.
According to the above and without loss of generality it is assumed that d̃ (n) =
f d (t) with n = 2. Next , the state vector of the system is extended by introducing as
additional state variables the disturbance terms and their derivatives

z 1,1 = x1,1 z 1,2 = x1,2 z 2,1 = x2,1 z 2,2 = x2,2

··· ··· ··· ···
z i,1 = xi,1 z i,2 = xi,2 z i+1,1 = xi+1,1 z i+1,2 = xi+1,2
··· ··· ··· ···
z n−1,1 = xn−1,1 z n−1,2 = xn−1,2 z n,1 = xn,1 z n,2 = xn,2
z n+1,1 = d̃1 z n+1,2 = d̃˙1 z n+2,1 = d̃2 z n+2,2 = d̃˙2 (12.126)
··· ··· ··· ···
z 2i+1,1 = d̃i z 2i+1,2 = d̃i z 2i+2,1 = d̃i+1 z 2i+2,2 = d̃˙i+1
˙
··· ··· ··· ···
z2n−1,1 = d̃
n−1 z2n−1,2 = d̃˙
n−1 z
2n,1 = d̃
n z2n,2 = d̃˙n

For the extended state-space description of the system new matrices A, B, and C are
formulated, comprising a double number of rows comparing to the description given
in Eqs. (12.115) and (12.116). However, even in this extended state-space form the
system remains observable.
12.10 Differential Flatness of the Multi-DOF Building’s Structure 665

For example, in the case of a five-storey building the extended state-space descrip-
tion of the system becomes the extended state vector of the system

z e = [z 1,1 , z 1,2 , z 2,1 , z 2,2 , z 3,1 , z 3,2 , z 4,1 , z 4,2 , z 5,1 , z 5,2 ,
(12.127)
z 6,1 , z 6,2 , z 7,1 , z 7,2 , z 8,1 , z 8,2 , z 9,1 , z 9,2 , z 10,1 , z 10,2 ]T .

The measurable state variables are z 1,1 , z 2,1 , z 3,1 , z 4,1 , and z 5,1 . By denoting the
extended state vector as z e ∈R 12×1 (and after omitting the disturbance functions
f d,i , i = 1, . . . , 5 from the control vector) one has the linear state-space equation
in the form
ż e = Ae z e + Be v
(12.128)
z emeas = Ce z e

where
⎛ ⎞ ⎛ ⎞
01000000001000000000 00000
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜1 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 1 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 1 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 1 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 1⎟
⎜
Ae = ⎜ ⎟ ⎜ ⎟
⎟ Be = ⎜0 0 0 0 0⎟ (12.129)
⎜0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 ⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ ⎜0 0 0 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎝0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1⎠ ⎝0 0 0 0 0 ⎠
00000000000000000000 00000
⎛ ⎞
10000000000000000000
⎜0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟
⎜ ⎟
Ce = ⎜ ⎟
⎜0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0⎟ (12.130)
⎝0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0⎠
00000000100000000000

A state estimator for the system of Eq. (12.128) has the form
666 12 Differential Flatness Theory for Distributed Parameter Systems

ẑ˙ = Ae ẑ + Be v + K (z meas − ẑ meas )

(12.131)
ẑ meas = Ce ẑ

Moreover, using common discretization methods the discrete-time equivalents of

matrices Ae , Be , and Ce are obtained, and these are are denoted as Ae,d , Be,d and
Ce,d . The computation of the estimator’s gain K can be performed with the use of
the Kalman Filter recursion, which consists of two stages: the measurement update
and the time update
K = P − Ce,d T [C − T
e,d P C e,d + R]
−1
−
ẑ e = ẑ e + K f (z meas − ẑ meas ) (12.132)
P(k) = P − (k) − K (k)Ce,d P − (k)

P − (k + 1) = Ae,d P(k)Ae,d T +Q
(12.133)
ẑ(k + 1) = Ae,d ẑ e (k) + Be,d v(k)

12.10.3 Simulation Tests

The proposed structural integrity method was tested in the model of a 5-DOF storey
building Fig. 12.29. The response of the building to the simulated external exci-
tation (e.g., earthquake) is shown in Fig. 12.31. The capability of the Derivative-
free nonlinear Kalman Filter for estimating the state vector of the building out of

(a) (b) DKF − suivi de l integrite structurelle de batiments

2 2

1.5 1.5
Y

1 1

0.5 0.5

0 0
−1 −0.5 0 0.5 1 −1 −0.5 0 0.5 1
X

Fig. 12.31 a A 5-DOF storey building structure at its initial position under no external excitation,
b displacement instant of the structure during external excitation (e.g., earthquake)
12.10 Differential Flatness of the Multi-DOF Building’s Structure 667

(a) (b)
0.3 0.3
position et1 position et2
x1 − x1−est

x3 − x3−est
0.2 DKF 0.2 DKF

0.1 0.1

0 0

−0.1 −0.1
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.6 0.6
vitesse et1 vitesse et2
x2 − x2−est

x4 − x4−est
0.4 DKF 0.4 DKF

0.2 0.2

0 0

−0.2 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 12.32 Derivative-free nonlinear Kalman Filter: a Estimation of the displacement and velocity
of floor 1. b Estimation of the displacement and velocity of floor 2

(a) (b)
0.15 0.15
position et3 position et4
x5 − x5−est

x7 − x7−est

0.1 DKF 0.1 DKF

0.05 0.05

0 0

−0.05 −0.05
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
0.6 0.6
vitesse et3 vitesse et4
x8 − x8−est
x6 − x6−est

0.4 DKF 0.4 DKF

0.2 0.2

0 0

−0.2 −0.2
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 12.33 Derivative-free nonlinear Kalman Filter: a Estimation of the displacement and velocity
of floor 3. b Estimation of the displacement and velocity of floor 4

measurements of the floors displacement is depicted in Figs. 12.32, 12.33, and 12.34a.
The active control inputs exerted on the floors are shown in Fig. 12.34b.
The nominal parameters of the building (coupled nonlinear oscillators model)
were as follows: elasticity coefficients k1 = 0.8, k2 = 0.8, k3 = 0.9, k4 = 10.1,
k5 = 10.2 and damping coefficients c1 = 1.5, c2 = 1.5, c3 = 1.5, c4 = 0.5,
c5 = 1.5. In the first damage case the changed parameters of the building’s dynamical
model were: elasticity coefficients k f1 = 2.8, k f 2 = 3.8, k f3 = 4.9, k f4 = 15.1,
668 12 Differential Flatness Theory for Distributed Parameter Systems

(a)
0.15 (b)
position et5
1

u1
0.1 DKF 0
x9 − x9−est

−1
0 5 10 15 20
0.05
t (sec)
1

u2
0 0
−1
0 5 10 15 20
−0.05
0 5 10 15 20 t (sec)
t (sec) 1

u3
0
0.6 −1
0 5 10 15 20
vitesse et5
t (sec)
x10 − x10−est

0.4 DKF
1

u4
0
0.2 −1
0 5 10 15 20
t (sec)
0
1

u5
0
−0.2 −1
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 12.34 Derivative-free nonlinear Kalman Filter: a Estimation of the displacement and velocity
of floor 5. b Active control inputs u i , ı = 1, . . . , 5

(a) (b)
10.6 11
test

10.4
10.5
2
mean value of the χ

10.2
score

10 10

9.8
9.5
9.6

9.4 9

0 5 10 15 20 0 5 10 15 20
2 2
No of χ test i No of χ test i

Fig. 12.35 Damage-free structure (confidence interval 98 % denoted with red lines): a Individual
values of the χ 2 tests, b mean value of the χ 2 test denoted with green line

k f5 = 14.2 and damping coefficients c f1 = 0.5, c f2 = 0.5, c f3 = 0.5, c f4 = 1.5,

c f5 = 0.5.
In the second damage case the changed parameters of the building’s dynamical
model were: elasticity coefficients k f1 = 3.8, k f2 = 4.8, k f3 = 5.9, k f4 = 17.1,
k f5 = 12.2 and damping coefficients c f1 = 0.7, c f2 = 0.8, c f3 = 0.9, c f4 = 1.2,
c f5 = 1.5.
Finally, in the third damage case the changed parameters of the building’s dynam-
ical model were: elasticity coefficients k f1 = 5.8, k f2 = 6.8, k f3 = 7.9, k f4 = 12.1,
12.10 Differential Flatness of the Multi-DOF Building’s Structure 669

(a) (b)
10.6 11

mean value of the χ test

10.4
10.5

2
10.2
score

10 10

9.8
9.5
9.6

9.4 9

0 5 10 15 20 0 5 10 15 20
2 2
No of χ test i No of χ test i

Fig. 12.36 Damaged structure—case 1 (confidence interval 98 % denoted with red lines): a Indi-
vidual values of the χ 2 tests, b mean value of the χ 2 test denoted with green line

(a) (b)
10.6 11
mean value of the χ 2 test

10.4
10.5
10.2
score

10 10

9.8
9.5
9.6

9.4 9

0 5 10 15 20 0 5 10 15 20
2 2
No of χ test i No of χ test i

Fig. 12.37 Damaged structure—case 2 (confidence interval 98 % denoted with red lines): a Indi-
vidual values of the χ 2 tests, b mean value of the χ 2 test denoted with green line

k f5 = 11.2 and damping coefficients c f1 = 1.5, c f2 = 2.5, c f3 = 1.5, c f4 = 2.5,

c f5 = 0.5.
The sampling period was Ts = 0.01 s. The fault thresholds are determined by
the confidence intervals of the χ 2 distribution. The χ 2 distribution has d = 10
degrees of freedom. The number of iterations was M = 2000. Thus, for an 98 %
confidence interval the associated upper and lower fault thresholds are U = 10.2340
and L = 9.7685. It can be observed that in case of a damage-free structure the mean
value of the χ 2 test remained between the upper and lower fault thresholds. On the
contrary when a parametric change had taken place the value of the χ 2 test exceeded
670 12 Differential Flatness Theory for Distributed Parameter Systems

(a) (b)
10.6 11

10.4

mean value of the χ2 test

10.5
10.2
score

10 10

9.8
9.5
9.6

9.4 9

0 5 10 15 20 0 5 10 15 20
2
No of χ test i
2
No of χ test i

Fig. 12.38 Damaged structure—case 3 (confidence interval 98 % denoted with red lines): a Indi-
vidual values of the χ 2 tests, b mean value of the χ 2 test denoted with green line

(a) (b)
d1 − d1−est

2 0.5
d/dt d1

1 disturbance et1 0 d/dt disturbance et1

0 DKF −0.5 DKF
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
d2 − d2−est

2 0.5
d/dt d2

1 disturbance et2 0 d/dt disturbance et2

0 DKF −0.5 DKF
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
d3 − d3−est

2 0.5
d/dt d3

1 disturbance et3 0 d/dt disturbance et3

0 DKF −0.5 DKF
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
d4 − d4−est

1 0.5
d/dt d4

0.5 disturbance et3 0 d/dt disturbance et3

0 DKF −0.5 DKF
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)
d5 − d5−est

1 0.5
d/dt d5

0 disturbance et3 0 d/dt disturbance et3

−1 DKF −0.5 DKF
0 5 10 15 20 0 5 10 15 20
t (sec) t (sec)

Fig. 12.39 Derivative-free nonlinear Kalman Filter redesigned as a disturbance observer: a Esti-
mation of additive disturbance terms affecting the dynamical model of the structure, b estimation
of the rate of change of the disturbance terms

clearly the fault boundaries. The associated results are depicted in Figs. 12.35, 12.36,
12.37, and 12.38.
Finally, in Fig. 12.39 it is shown that when the Derivative-free nonlinear Kalman
Filter is redesigned as a disturbance observer, according to the stages described in
Sect. 12.10.1 it made possible to estimate additive input disturbances that affected
the structure’s dynamical model.
Chapter 13
Differential Flatness Theory
in the Background of Other
Control Methods

13.1 Differential Flatness Theory in the Background

of Backstepping Control

13.1.1 Overview

As shown in the previous chapters, differential flatness theory is currently a main

direction in nonlinear control systems analysis and synthesis [285, 352, 420, 427, 453,
465, 495]. A system is considered to be differentially flat if all its state variables and
its control inputs can be expressed as functions of one single algebraic variable which
is the so-called flat output, and also as functions of the flat output’s derivatives [152,
284, 286, 338, 535]. The differential flatness property enables the transformation
of the nonlinear system’s dynamics in the linear canonical form and the design of
a state feedback controller through the application of pole placement techniques
in the linearized equivalent model of the system [55, 103, 351, 494, 516, 613].
In this chapter, a different approach is developed for controller design in nonlinear
dynamical systems which exhibit the differential flatness property. The method makes
use of the initial nonlinear model of the system and of its decomposition into a set
of nonlinear subsystems for which the differential flatness property holds.
The proposed control method is directly applicable to nonlinear systems of the
so-called triangular form, or to system’s which can be transformed into such a form.
Each row of the state-space model of the nonlinear system is considered to be a
subsystem, which satisfies the differential flatness properties. For each subsystem
which is associated with a row of the state-space model a virtual control input is
computed, capable of inverting the subsystem’s dynamics and of eliminating the
subsystem’s tracking error. The control input that is actually applied to the nonlinear
system is found from the last row of the state-space description. This control input
incorporates in a recursive manner all virtual control inputs which were computed
for the individual subsystems associated with the previous rows of the state-space
equation. The control input that should be applied to the nonlinear system so as
© Springer International Publishing Switzerland 2015 671
G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5_13
672 13 Differential Flatness Theory in the Background of Other …

to assure that all its state vector elements will converge to the desirable setpoints,
is obtained at each iteration of the control algorithm, by tracing the rows of the
state-space model backwards. The performance of the proposed control method is
evaluated through simulation experiments, in the case of nonlinear robotic models,
as well as in the case of nonlinear electric power generation systems. Moreover,
the chapter shows how the problem of optimal control can be reformulated using
differential flatness theory terms.
In a following section, the aforementioned implementation of flatness-based con-
trol in cascading loops is used to solve the problem of boundary control of distributed
parameter systems, that is control of nonlinear PDEs using as control inputs their
boundary conditions. The objective of this section is to develop a boundary control
method for a 1D nonlinear wave equation and a filtering approach for estimating
the dynamics of such a system from measurements provided by a small number of
sensors. It is shown that the numerical solution of the associated partial differential
equation results in a set of nonlinear ordinary differential equations. With the appli-
cation of a diffeomorphism that is based on differential flatness theory it is shown
that an equivalent description of the system in the linear canonical (Brunovsky) form
is obtained. This transformation enables to compute estimates about the state vec-
tor of the system through the application of the standard Kalman Filter recursion.
For the local subsystems, into which the nonlinear wave equation is decomposed, it
becomes possible to apply state estimation-based feedback control. The controller
design proceeds by showing that the rows of the state-space model of the nonlinear
PDE stand for differentially flat systems, where the flat output is chosen to be the
associated state variable. Next, for each subsystem which is linked with a row of
the state-space model a virtual control input is computed, that can invert the sub-
system’s dynamics and can eliminate the subsystem’s tracking error. From the last
row of the state-space description, the control input that is actually applied to the
nonlinear system is found. This control input contains recursively all virtual control
inputs which were computed for the individual subsystems associated with the pre-
vious rows of the state-space equation. Thus, by tracing the rows of the state-space
model backwards, at each iteration of the control algorithm, one can finally obtain
the control input that should be applied to the nonlinear PDE so as to assure that all
its state vector elements will converge to the desirable setpoints. It is shown that, by
applying feedback control, the nonlinear wave-type dynamics can be made to track
the desirable setpoint.
Through the previously described stages, the chapter shows the following: (i)
backstepping control stands for a subcase of flatness-based control, (ii) to apply
flatness-based control it is not restrictive to demonstrate that the controlled system
satisfies the Lie-Backlünd condition (the latter meaning that the system can be trans-
formed to the linear canonical form). It is possible to apply flatness-based control in
the cascading loops which are associated with the rows of the state-space description
of the system.
13.1 Differential Flatness Theory in the Background of Backstepping Control 673

13.1.2 Flatness-Based Control Through Transformation

into the Canonical Form

After proving differential flatness properties for a dynamical system, flatness-based

control is usually implemented through its transformation to the linear canonical
form. This is explained in the following example.
Example: The following state-space equation of a nonlinear DC motor, that was
given in Chap. 8, is reminded:

ẋ1 = x2
ẋ2 = k1 x2 + k2 x3 + k3 x32 + k4 T1 (13.1)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u

where, x1 = θ is the rotor’s turn angle, x2 = θ̇ is the rotor’s angular speed, and
x3 = I is the armature’s current. Moreover, T1 the load torque, u is the field voltage,
and ki are coefficients. From the second row of Eq. (13.1) one obtains,

ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 x3 ẋ3 ⇒

ẍ2 = k1 ẋ2 + k2 ẋ3 + 2k3 k5 x2 x3 + 2k3 k6 x2 x32 + (13.2)
+2k3 k7 x32 + 2k3 k8 x3 u

Thus the input–output relation can be written as

ẍ2 = f¯(x) + ḡ(x)u (13.3)

where, f¯(x) = k1 ẋ2 +k2 ẋ3 +2k3 k5 x2 x3 +2k3 k6 x2 x32 +2k3 k7 x32 , and ḡ(x) = 2k3 k8 x3 .
Taking k1 = 0 and considering the torque T1 as external disturbance, the nonlinear
DC motor model of Eq. (13.1) becomes

ẋ1 = x2
ẋ2 = k2 x3 + k3 x32 (13.4)
ẋ3 = k5 x2 + k6 x2 x3 + k7 x3 + k8 u

Then, setting the flat output to be y = x1 one can see that all state variables xi , i =
1, 2, 3 and the control input u can be expressed as functions of this output and its
derivatives. Indeed it holds

x1 = y
x2 = ẏ
−k2 + |k 2 +4k3 ÿ| (13.5)
x3 = 2
2k3
u = ḡ(x)
1
[y (3) − f¯(x)]
674 13 Differential Flatness Theory in the Background of Other …

The aforementioned system of Eq. (13.4) can be written in the trivial (Brunovsky
canonical form):
⎛ ⎞ ⎛ ⎞⎛ ⎞ ⎛ ⎞
ẏ1 010 y1 0
⎝ ẏ2 ⎠ = ⎝0 0 1⎠ ⎝ y2 ⎠ + ⎝0⎠ v (13.6)
ẏ3 000 y3 1

where v = f¯(x, t) + ḡ(x, t)u. Using a control input of the form u = ḡ(x, t)−1 [v −
f¯(x, t)] it is possible to make the electric motor’s angle track any desirable setpoint.
Therefore, a transformation of the system into a linear equivalent model is obtained
and then it is straightforward to design a controller based on linear control theory.
Thus, given the reference trajectory [x1∗ , x2∗ , x3∗ ]T one can find the transformed refer-
ence trajectory [y1∗ , ẏ1∗ , ÿ1∗ ]T and can also select the appropriate control input v that
succeeds tracking. Knowing v, the control u of the initial system can be found.
For the linearized description of the system given above, the state feedback con-
(3)
troller is v = yd − k f1 ( ÿ − ÿd ) − k f2 ( ẏ − ẏd ) − k f3 (y − yd ). This results in tracking
error dynamics of the form e(3) (t) + k f1 ë(t) + k f 2 ė(t) + k f3 e(t) = 0. By selecting
the feedback gains k fi , i = 1, 2, 3 such that the previous characteristic polynomial
is a Hurwitz one, it assured that limt→∞ e(t) = 0.

13.1.3 A New Approach to Flatness-Based Control for

Nonlinear Dynamical Systems

The following nonlinear dynamical system is now examined:

ẋ = f (x) + g(x)u x∈R n u∈R

(13.7)
y = h(x)

Moreover, it is considered that the system can be written in the so-called triangular
(or integral backstepping) form (this is also succeeded through transformation to the
canonical form):

ẋ1 = f 1 (x1 ) + g1 (x1 )x2

ẋ2 = f 2 (x1 , x2 ) + g2 (x1 , x2 )x3
ẋ3 = f 3 (x1 , x2 , x3 ) + g3 (x1 , x2 , x3 )x4
···
(13.8)
ẋi = f i (x1 , x2 , . . . , xi ) + gi (x1 , x2 , . . . , xi )xi+1
···
ẋn−1 = f n−1 (x1 , x2 , . . . , xn−1 ) + gn−1 (x1 , x2 , . . . , xn−1 )xn
ẋn = f n (x1 , x2 , . . . , xn ) + gn (x1 , x2 , . . . , xn )u
13.1 Differential Flatness Theory in the Background of Backstepping Control 675

The following virtual control inputs αi = xi+1 are defined per row of the state-space
model of Eq. (13.8)

ẋ1 = f 1 (x1 ) + g1 (x1 )α1

ẋ2 = f 2 (x1 , x2 ) + g2 (x1 , x2 )α2
ẋ3 = f 3 (x1 , x2 , x3 ) + g3 (x1 , x2 , x3 )α3
···
(13.9)
ẋi = f i (x1 , x2 , . . . , xi ) + gi (x1 , x2 , . . . , xi )αi
···
ẋn−1 = f n−1 (x1 , x2 , . . . , xn−1 ) + gn−1 (x1 , x2 , . . . , xn−1 )αn−1
ẋn = f n (x1 , x2 , . . . , xn ) + gn (x1 , x2 , . . . , xn )u

The system of Eq. (13.9) is a differentially flat one. It is considered that y = x1 is the
flat output of system. It can be easily shown that each virtual control input αi = xi+1 ,
i = 1, 2, . . . can be expressed as a function of the flat output and its derivatives, since
it holds

αi = gi (x1 ,x2 ,...,xi ) ( ẋ i

1
− f (x1 , x2 , . . . , xi )) (13.10)

For i = 1 and a1 = x2 one has

α1 = g1 (x1 ) ( ẋ 1
1
− f 1 (x1 )) (13.11)

which means that α1 is a function of the flat output and its derivatives. For i = 2 one
has

α2 = g2 (x1 ,x2 ) ( ẋ 2
1
− f 2 (x1 , x2 )) (13.12)

which means that α2 = x3 is a function of the flat output y = x1 and its derivatives.
Continuing in a similar manner one has αn−1 = xn and consequently αn = u is
also a function of the flat output y = x1 and its derivatives. According to the above,
one has a nonlinear dynamical system in which, all its state variables and the control
input can be written as functions of the flat output and its derivatives. Therefore, such
a system is differentially flat.
Additionally, by considering each row of the model of Eq. (13.9), one has a set of
n subsystems of the form

ẋi = f i (x1 , x2 , . . . , xi ) + gi (x1 , x2 , . . . , xi )αi (13.13)

where each subsystem describes the dynamics of the single state variable xi . For
each one of these subsystems one can consider the state variable xi as the flat output.
Variables x1 , . . . , xi−1 (being all functions of the flat output y1 = x1 and of its
derivatives) are locally considered to be parameters of the ith per row differential
equation. Obviously, the virtual control input αi is a function of this flat output and
its derivatives. Therefore, each local subsystem is also differentially flat.
676 13 Differential Flatness Theory in the Background of Other …

Next, one can compute the virtual inputs which are applied to each subsystem.
For the first subsystem, which is associated with the first row of Eq. (13.8), and by
defining z i = xi − xid = x1 − αi−1 , the virtual control input is given by

α1 = x2∗ = g1 (x1 ) ( ẋ 1
1 d − f (x1 ) − K 11 (x1 − x1d )) ⇒
(13.14)
α1 = x2∗ = g1 (x1 ) ( ẋ 1
1 d − f (x1 ) − K 11 z 1 )

From the second row of Eq. (13.8), and using z 2 = x2 − x2∗ = x2 − α1 one has

α2 = x3∗ = ∗
g2 (x1 ,x2 ) ( ẋ 2
1
− f 2 (x1 , x2 ) − K 12 (x2 − x2∗ )) ⇒
(13.15)
α2 = x3∗ = g2 (x1 ,x2 ) (α̇1
1
− f 2 (x1 , x2 ) − K 12 z 2 )

From the third row of Eq. (13.8), and using z 3 = x3 − x3∗ = x3 − α2 one has

α3 = x4∗ = ∗
g3 (x1 ,x2 ,x3 ) ( ẋ 3
1
− f 3 (x1 , x2 , x3 ) − K 13 (x3 − x3∗ )) ⇒
(13.16)
α3 = x4∗ = g3 (x1 ,x2 ,x3 ) (α̇2
1
− f 3 (x1 , x2 , x3 ) − K 13 z 3 )

Continuing in a similar manner and from the ith row of the state-space description
of the system given in Eq. (13.8), and while also using z i = xi − xi∗ = xi − αi−1
one obtains
∗ =
αi = xi+1 ∗ − f i (x1 , x2 , . . . , xi ) − K 1i (xi − xi∗ )) ⇒
gi (x1 ,x2 ,...,xi ) ( ẋ i
1

(13.17)
αi = ∗ = gi (x1 ,x2 ,...,xi ) (α̇i−1 − f i (x1 , x2 , . . . , x3 ) − K 1i z i )
1
xi+1

Equivalently, from the n − 1th row of the state-space model of Eq. (13.8) and using
∗
z n−1 = xn−1 − xn−1 = xn−1 − αn−2 one has

αn−1 = xn∗ = ∗ ∗ ))
− f n−1 (x1 , x2 , . . . , xn−1 ) − K 1n−1 (xn−1 − xn−1
gn−1 (x1 ,x2 ,...,xn−1 ) ( ẋ n−1
1

⇒αn−1 = xn∗ = gn−1 (x1 ,x2 ,...,xn−1 ) (α̇n−2

1
− f n−1 (x1 , x2 , . . . , xn−1 ) − K 1n−1 z n−1 )
(13.18)

Finally, from the nth row of the state-space model of Eq. (13.8) and using z n =
xn − xn∗ = xn − αn−1 one has

αn = u = ∗ − f n (x1 , x2 , . . . , xn ) − K 1n (xn − xn∗ )) ⇒

gn (x1 ,x2 ,...,xn ) ( ẋ n
1

(13.19)
αn = u = gn (x1 ,x2 ,...,xn ) (α̇n−1
1
− f n (x1 , x2 , . . . , xn ) − K 1n z n )
13.1 Differential Flatness Theory in the Background of Backstepping Control 677

The computation of the control input u that should be actually applied to the nonlinear
system is performed in a recursive manner by processing backwards the virtual
control inputs described in Eqs. (13.14)–(13.19).
Thus, from the last row of the state-space description the control input that is actu-
ally applied to the nonlinear system is found. This control input contains recursively
all virtual control inputs which were computed for the individual subsystems asso-
ciated with the previous rows of the state-space equation. Consequently, by tracing
the rows of the state-space model backwards, at each iteration of the control algo-
rithm, one can finally obtain the control input that should be applied to the nonlinear
system so as to assure that all its state vector elements will converge to the desirable
setpoints.

13.1.4 Closed-Loop Dynamics

By substituting Eq. (13.19) into the last row of the state-space model of Eq. (13.8),
and using the definition ẋn − ȧn−1 = z n , one obtains:

ẋn = ȧn−1 − K 1n (xn − αn−1 )⇒

(ẋn − ȧn−1 ) + K 1n (xn − αn−1 ) = 0⇒ (13.20)
ż n + K 1n z n = 0

By substituting Eq. (13.18) into the last row of the state-space model of Eq. (13.8),
and using the definition ẋn−1 − ȧn−2 = z n−1 , one obtains:

ẋn−1 = ȧn−2 − K 1n−1 (xn−1 − αn−2 )⇒

(ẋn−1 − ȧn−2 ) + K 1n−1 (xn−1 − αn−2 ) = 0⇒ (13.21)
ż n−1 + K 1n−1 z n−1 = 0

By substituting Eq. (13.17) into the last row of the state-space model of Eq. (13.8),
and using the definition ẋi − ȧi−1 = z i , one obtains:

ẋi = ȧi−1 − K 1n−1 (xi − αi−1 )⇒

(ẋi − ȧi−1 ) + K 1i (xi − αi−1 ) = 0⇒ (13.22)
ż i + K 1i z i = 0

while continuing backward and by substituting Eq. (13.15) into the second row of
the state-space model of Eq. (13.8), and using the definition ẋ2 − ȧ1 = z 2 , one gets:

ẋ2 = ȧ1 − K 12 (x2 − α1 )⇒

(ẋ2 − ȧ1 ) + K 12 (x2 − α1 ) = 0⇒ (13.23)
ż 2 + K 12 z 2 = 0
678 13 Differential Flatness Theory in the Background of Other …

Finally, by substituting Eq. (13.14) into the first row of the state-space model of
Eq. (13.8), one has:

ẋ1 = ẋ1 − K 11 (x1 − x1d )⇒

(ẋ1 − ẋ1d ) + K 11 (x1 − x1d ) = 0⇒ (13.24)
ż 1 + K 11 z 1 = 0

Therefore, after the application of the feedback control law, the closed-loop dynamics
becomes ż 1 + K 11 z 1 = 0, ż 2 + K 12 z 2 = 0, ż 3 + K 13 z 3 = 0, . . ., ż i + K 1i z i = 0, . . .,
ż n−1 + K 1n−1 z n−1 = 0, ż n + K 1n z n = 0.
In matrix form, the closed-loop dynamics is written as
⎛ ⎞ ⎛ ⎞⎛ ⎞
ż 1 −K 11 0 0 · · · 0 · · · 0 0 z1
⎜ ż 2 ⎟ ⎜ 0 − K 2 0 · · · 0 · · · 0 0 ⎟ ⎜ z 2 ⎟
⎜ ⎟ ⎜ 1 ⎟⎜ ⎟
⎜ ż 3 ⎟ ⎜ 0 0 − K 3 · · · 0 · · · 0 0 ⎟ ⎜ z 3 ⎟
⎜ ⎟ ⎜ 1 ⎟⎜ ⎟
⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ··· ⎟⎜ ··· ⎟
⎜ ⎟=⎜ ⎟ ⎜ ⎟ (13.25)
⎜ ż i ⎟ ⎜ 0 0 0 · · · − K i · · · 0 0 ⎟ ⎜ z i ⎟
⎜ ⎟ ⎜ 1 ⎟⎜ ⎟
⎜ ··· ⎟ ⎜ ··· ··· ··· ··· ··· ⎟⎜ ··· ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟
⎝ż n−1 ⎠ ⎝0 0 0 · · · 0 · · · − K n−1 0⎠ ⎝z n−1 ⎠
1
ż n 0 0 0 · · · 0 · · · 0 − K 1n zn

or equivalently

Ż = KZ (13.26)

By selecting the eigenvalues of matrix K to be in the left complex semiplane, one

has

limt→∞ Z = 0n×1 (13.27)

which also implies that limt→∞ x1 = x1d , limt→∞ x2 = α1 = x2d , limt→∞ x3 =

α2 = x3d , . . ., limt→∞ xi = αi−1 = xid , . . ., limt→∞ xn−1 = αn−2 = xn−1
d , and

limt→∞ xn = αn−1 = xn . d

To prove asymptotic stability for the proposed control scheme the following
Lyapunov function can be defined
N
V = 1 2
i=1 2 z i (13.28)

The time derivative of the aforementioned Lyapunov function is

N N
V̇ = i=1 z i ż i ⇒ V̇ = − i=1 K 1i z i2 ⇒V̇ < 0 (13.29)

By selecting the feedback control gains K 1i , i = 1, . . . , n to be K 1i > 0, the asymp-

totic stability of the control loop is assured.
13.1 Differential Flatness Theory in the Background of Backstepping Control 679

13.1.5 Comparison to Backstepping Control

Backstepping control can be applied to nonlinear dynamical systems in the triangular

form. A backstepping control law can be derived for systems of the form [238, 407]:

ẋ = f (x) + g(x)u
(13.30)
y = h(x)

for which holds

∂h(x) ∂h(x)
ẏ = ẋ = [ f (x) + g(x)u] = L f h(x) + L g h(x)u (13.31)
∂x ∂x
where the Lie derivatives are defined as
∂ ∂
L f h(x) = h(x) f (x), L g h(x) = h(x)g(x) (13.32)
∂x ∂x
The system of Eq. (13.30) can be written in cascading form

ẋ1 = f 1 (x1 ) + g1 (x1 )x2

ẋ2 = f 2 (x1 , x2 ) + g2 (x1 , x2 )x3
ẋ3 = f 3 (x1 , x2 , x3 ) + g3 (x1 , x2 , x3 )x4
··· ···
··· ··· (13.33)
x˙n 1 = f n−1 (x1 , x2 , . . . , xn−1 ) + gn−1 (x1 , x2 , . . . , xn−1 )xn
x˙n = f n (x1 , x2 , . . . , xn ) + gn−1 (x1 , x2 , . . . , xn )u

y = h(x1 )

Then, the nth order backstepping SISO controller is given by the recursive relation

α1 = L h(x 1 [ ẏd − L f 1 h(x1 ) − k1 z 1 − n 1 (z 1 )z 1 ]

g1 1)
α2 1
= L h(x ,x ) [α̇1 − f 2 (x1 , x2 ) − L g1 h(x1 )z 1 − k2 z 2 − n 2 (z 2 )z 2 ]
g2 1 2
···
1
αi = L h(x ,x [ α̇ − f (x , . . . , x ) − gi−1 (x1 , x2 , . . . , xi−1 )z i−1 − ki z 1 − n i (z i )z i ]
gi 1 2 ,...,x ) i−1
i
i 1 i
···
1
αn = L h(x ,x [ α̇ − f (x , . . . , x ) − gn−1 (x1 , x2 , . . . , xn−2 )z n−1 − kn z n − n n (z n )z n ]
n n
1 2 ,...,x n )
gn n−1 1

u = αn
(13.34)

with z 1 = h(x1 ) − yd and z i = xi − αi−1 . Such a backstepping controller results in

closed-loop dynamics given by ż = −K (z)z + S(x)z, with
680 13 Differential Flatness Theory in the Background of Other …
⎛ ⎞
k1 + n 1 (z 1 ) 0 ··· 0
⎜ 0 k2 + n 2 (z 2 ) · · · 0 ⎟
K (z) = ⎜
⎝
⎟
⎠ (13.35)
··· ··· ··· ···
0 0 · · · kn + n n (z n )

S(x) =
⎛ ⎞
0 L g1 (x1 ) 0 ··· 0 0 0
⎜−L g1 (x1 ) 0 g (x 1 , x2 ) 0 0 0 ⎟
⎜ 2 ⎟
⎜ 0 −g (x , x ) 0 ··· 0 0 0⎟
⎜ 2 1 2 ⎟
⎜ ··· ··· ··· ··· ··· ··· · · ·⎟
⎜ ⎟
⎜ 0 0 0 ··· 0 gn−1 (x1 , . . . , xn−1 ) 0⎟
⎜ ⎟
⎝ 0 0 0 · · · −gn−1 (x1 , . . . , xn−1 ) 0 0⎠
0 0 0 ··· 0 −gn (x1 , . . . , xn 0)
(13.36)

From above it can be noticed that the so-called backstepping control, which is based
on the recursive computation of the control signal of the system after applying virtual
control inputs to the individual rows of the state-space model, can be completely
substituted by the proposed flatness-based control method. A backstepping control
law can be derived for systems of the triangular form [238, 407, 428]. However, as it
was previously analyzed, by showing that each row of the state-space model stands
for a subsystem that satisfies differential flatness properties one can apply effectively
to each subsystem the controller design stages found in flatness-based (input–output
linearizing) control methods.

13.1.6 Simulation Tests

13.1.6.1 Control of a 1-DOF Robotic Manipulator

The proposed control scheme was applied to the model of a 1-DOF robotic manipu-
lator which is depicted in Fig. 13.1. The model comprises a DC electric actuator and
a link with a mass attached to its end effector.
As in the previously presented example, the definition of the model makes use of
the following parameters: θ : is the rotor’s rotation angle, I is the armature’s current,
K I is a torque constant, K b is a back electromotive coefficient, B is the viscous
bearing friction coefficient, l is the length of the link, M is the mass of the payload,
m is the mass of the link, L is the rotor’s inductance, R is the rotor’s resistance, V is
the input voltage, and J is the inertia of the actuator’s rotor.
The following state variables are defined: x1 = θ , x2 = θ̇, x3 = I . Moreover,
the following parameters are defined: Jl is the aggregate moment of inertia of the
rotating part of the system (rotor of the electric motor, link and mass attached to its
end) and N is the gravity torque exerted on the link and on the mass attached to its
end. Thus, the considered robotic model is [550]:
13.1 Differential Flatness Theory in the Background of Backstepping Control 681

Fig. 13.1 Model of the 1-DOF robotic manipulator

ẋ1 = x2
ẋ2 = − Jl x2 + NJl sin(x1 ) + KJlI x3
B
(13.37)
ẋ3 = − RL x3 − KLb x2 − wL + L1 u

The performance of the proposed control scheme was evaluated in the aforemen-
tioned model of the robotic manipulator, in the case of tracking of several setpoints.
The associated results are presented in Figs. 13.2 to 13.3, in Figs. 13.4 to 13.5, and in
Figs. 13.6 to 13.7. In all cases, the state variables of the robot (blue lines) converged

(a) 5.5 (b) 12

5
10
4.5
8
4

3.5 6
x2
x1

3 4

2.5
2
2
0
1.5

1 −2
0 5 10 15 20 0 5 10 15 20
time time

Fig. 13.2 Setpoint 1. a Convergence of state variable x1 to the desirable setpoint, b Convergence
of state variable x2 to the desirable setpoint
682 13 Differential Flatness Theory in the Background of Other …

(a) 20 (b) 100

80
15
60
10
40
5
20
x3

0 0

u
−20
−5
−40
−10
−60
−15
−80

−20 −100
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
time time

Fig. 13.3 Setpoint 1. a Convergence of state variable x3 to the desirable setpoint, b Control input u

(a) 2.5 (b) 3

2
2
1.5

1
1
0.5
x1

x2

0 0

−0.5
−1
−1

−1.5
−2
−2

−2.5 −3
0 5 10 15 20 0 5 10 15 20
time time

Fig. 13.4 Setpoint 2. a Convergence of state variable x1 to the desirable setpoint, b Convergence
of state variable x2 to the desirable setpoint

with accuracy to the reference setpoints (red lines) while the associated control input
varied smoothly (Figs. 13.8 and 13.9). Besides the proposed control scheme per-
formed well under modeling uncertainty and external disturbances.

13.1.6.2 Control of a Permanent Magnet Synchronous Generator

The proposed flatness-based control method is also applied to the model of a Per-
manent Magnet Synchronous Generator (PMSG). The single-machine infinite-bus
13.1 Differential Flatness Theory in the Background of Backstepping Control 683

(a) 20 (b) 100

80
15
60
10
40
5
20
x3

0 0

u
−20
−5
−40
−10
−60
−15
−80

−20 −100
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
time time

Fig. 13.5 Setpoint 2. a Convergence of state variable x3 to the desirable setpoint, b Control input u

5 5
(a) (b)
4
4
3
3
2

1 2
x1

x2

0 1

−1
0
−2
−1
−3

−4 −2

−5 −3
0 5 10 15 20 0 5 10 15 20

time time

Fig. 13.6 Setpoint 3. a Convergence of state variable x1 to the desirable setpoint, b Convergence
of state variable x2 to the desirable setpoint

model of the PMSG (Fig. 13.10) is described by a nonlinear state-space model of the
form
ẋ = f (x) + g(x)u (13.38)


 T
where the state vector x is defined as x = Δδ, Δω, E q . The vector fields f (x)
and g(x) are defined as
684 13 Differential Flatness Theory in the Background of Other …

(a) 20 (b)100
80
15
60
10
40
5
20
x3

0 0

u
−20
−5
−40
−10
−60
−15
−80

−20 −100
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20

time time

Fig. 13.7 Setpoint 3. a Convergence of state variable x3 to the desirable setpoint, b Control input u

(a) 15 (b) 20
15
10
10

5
5
x1

x2

0
0

−5
−5
−10

−10 −15
0 5 10 15 20 0 5 10 15 20
time time

Fig. 13.8 Setpoint 4. a Convergence of state variable x1 to the desirable setpoint, b Convergence
of state variable x2 to the desirable setpoint

⎛ ⎞
ω − ω0

⎜ D 1 Vs E q ⎟
⎜− 2J (ω − ω0 ) + ω0 2J
Pm
− ω sin(Δδ)⎟
f (x) = ⎜ 0 2J 
xdσ ⎟ (13.39)
⎝  ⎠
1  1 xd −xd
−  E q + Tdo  Vs cos(Δδ)
Td xdσ

T
1
g(x) = 0 0 Tdo
(13.40)
13.1 Differential Flatness Theory in the Background of Backstepping Control 685

(a) 20 (b) 100

80
15
60
10
40
5
20
x3

0 0

u
−20
−5
−40
−10
−60
−15
−80

−20 −100
2 4 6 8 10 12 14 16 18 20 2 4 6 8 10 12 14 16 18 20
time time

Fig. 13.9 Setpoint 4. a Convergence of state variable x3 to the desirable setpoint, b Control input u

Vt VS
XL
XT

Generator Transformer Transmission Line Infinite

bus

Fig. 13.10 Model of the permanent magnet synchronous generator connected to an infinite buss

with control input u = E f being the field voltage (equivalent voltage in the excitation
coil) and measurable output the rotation speed of the rotor

y = h(x) = δ − δ0 (13.41)

The obtained results are depicted in Figs. 13.11 and 13.12. It can be noticed that the
proposed control method succeeds fast and there is accurate tracking of the reference
setpoints for the state variables of the generator. The associated variations of the
control input also remain smooth. This confirms the previously given theoretical
analysis about the stability and convergence of the proposed control scheme and
shows its potential application to electric power generation systems.
686 13 Differential Flatness Theory in the Background of Other …

(a) 2.5 (b) 100

80
2
60

40
1.5
20
x2

1 0

u
−20
0.5
−40

−60
0
−80

−0.5 −100
0 5 10 15 20 2 4 6 8 10 12 14 16 18 20
time time

Fig. 13.11 a The deviation Δω of the PMSG rotor angular velocity from the synchronous speed
converges to 0, b associated control input E f

(a) 1.2 (b) 100

80
1
60

0.8 40

20
0.6
x2

0
u

0.4
−20

0.2 −40

−60
0
−80

−0.2 −100
0 5 10 15 20 2 4 6 8 10 12 14 16 18 20
time time

Fig. 13.12 a The deviation Δω of the PMSG rotor angular velocity from the synchronous speed
is made to track a reference setpoint, b associated control input E f

13.2 Differential Flatness and Optimal Control

Using differential flatness theory one can also reformulate optimal control problems.
Consider the classical optimal control problem
T
minu J (u) = 0 L(x(t), u(t))dt (13.42)
13.2 Differential Flatness and Optimal Control 687

with ẋ = f (x, u), x(0) = a, x(T ) = b where a, b and T > 0 are given. Suppose
that ẋ = f (x, u) is a differentially flat system with y = h(x, u, . . . , u (r ) ) being the
flat output. It holds that all state variables and the control inputs can be written as
functions of the flat output and of its derivatives

ẋ = φ(y, . . . , y (r ) ) u = α(y, . . . , y (q) ) (13.43)

The numerical solution of minu J (u) requires prior discretization of the state-space
which is an approximation of finite dimension. A different approach is based
Ndiscretization of the flat output, according to a relation of the form yi (t) =
on
j=1 Aij λ j (t). The initial and final conditions on x signify that Aij belong to a sub-
space, noted as V . Thus, one is led to the following formulation of the minimization
problem in optimal control
T
min A∈V J (A) = 0 L(φ(y, . . . , y (r ) ), α(y, . . . , y (q) ))dt (13.44)
N
where yi have been replaced by i=1 Aij λ j (t). This representation can be used for
generating optimal trajectories.
In conclusion, and through the analysis provided in this section, it can be stated
that differential flatness theory is not only used for the solution of nonlinear control
and filtering problems but is also met in the background of other control approaches,
such as backstepping control and optimal control.

13.3 Boundary Control of Nonlinear PDE Dynamics Using

Differential Flatness Theory

13.3.1 Overview

In this section, a differential flatness theory-based control method is developed for

distributed parameter systems, by using cascading loops and by applying as control
inputs the PDE’s boundary conditions. Boundary control of distributed parameter
systems described by nonlinear partial differential equations is important for several
applications [122, 150, 151, 169, 232, 242, 427, 453, 465]. As mentioned, parabolic
differential equations are met in diffusion phenomena that appear in heat transfer,
chemical processes as well as in quantum mechanics [320, 557, 619]. Moreover,
nonlinear hyperbolic differential equations mostly describe wave-type phenomena
[33, 135, 175, 253, 254, 371, 466, 516]. Additionally, elliptic partial differential
equations are met in problems of vibrating structures, acoustic cavities, and radiation
waves. The section treats the problem of boundary control of nonlinear distributed
parameter systems, which means that the boundary conditions are used as control
688 13 Differential Flatness Theory in the Background of Other …

inputs to modify the PDE dynamics. Boundary control or lumped input control
of nonlinear distributed parameter systems are problems of elevated difficulty and
many remarkable results on this topic exist [59, 154, 363, 560]. If the PDE system
is decomposed into an equivalent set of ordinary nonlinear differential equations,
then controlling such a set of dynamical subsystems by the boundary conditions
implies underactuation (actually one has to control the different degrees of freedom
associated with the set of nonlinear ODEs by using only a couple of control inputs).
The control approach followed in this section is as follows. First, by implementing
a procedure as in the numerical solution of PDEs, a set of equivalent nonlinear
ordinary differential equations is obtained [133, 175, 186, 255, 388, 558]. Next, it
is shown that the system of the nonlinear ODEs is a differentially flat one. This
means that all its state variables and the control inputs can be written as differential
functions of one single algebraic variable which is the flat output [55, 152, 286, 284,
495]. Moreover, by examining independently each nonlinear ODE it is shown that
this stands again for a differentially flat system, for which a virtual control input can
be computed as in the case of flatness-based control for the trivial system. The virtual
control input is chosen such that the ODE subsystem dynamics is linearized and the
tracking error is eliminated. The boundary condition which appears in the nonlinear
ODE subsystem that comprises the last rows of the state-space description, stands
for the aggregate control input. The computation of the boundary control input uses
recursively all virtual control inputs mentioned above, moving from the last ODE
system to the first one. Thus, by tracing the rows of the state-space model backwards,
at each iteration of the control algorithm, one can finally obtain the control input that
should be applied to the nonlinear PDE system so as to assure that all its state vector
elements will converge to the desirable setpoints. By analyzing the dynamics of the
closed-loop system that results from the application of the aforementioned control
method, asymptotic stability is confirmed. The efficiency of the proposed boundary
control scheme for nonlinear PDEs is tested through simulation experiments.

13.3.2 Transformation of the PDE Model into a Set of

Nonlinear ODEs

13.3.2.1 Decomposition of the PDE Model into a Set of Equivalent ODEs

The transformation of the nonlinear PDE model into an equivalent description which
consists of a set of nonlinear ODEs follows the stages described in Chap. 12. The
previously considered nonlinear wave equation is used again

∂2φ
= k ∂∂ xφ2 + f (φ)
2
(13.45)
∂t 2
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 689

where u(x, t) is assumed to be the external control input. Using the approximation
for the partial derivative
∂2φ φi+1 −2φi +φi−1
2 =
(13.46)
∂x 2 Δx

and by considering spatial measurements of variable φ along axis x at points x0 +

iΔx, i = 1, 2, . . . , N one has

∂ 2 φi
∂t 2
= K
φ
Δx 2 i+1
− 2K
φ
Δx 2 i
+ K
φ
Δx 2 i−1
+ f (φi ) (13.47)

By considering the associated samples of φ given by φ0 , φ1 , . . . , φ N , φ N +1 one has

∂ 2 φ1
∂t 2
= K
φ
Δx 2 2
− 2K
φ
Δx 2 1
+ K
φ
Δx 2 0
+ f (φ1 )
∂ 2 φ2
∂t 2
= K
φ
Δx 2 3
− 2K
φ
Δx 2 2
+ K
φ
Δx 2 1
+ f (φ2 )
∂ 2 φ3
∂t 2
= K
φ
Δx 2 4
− 2K
φ
Δx 2 3
+ K
φ
Δx 2 2
+ f (φ3 )
(13.48)
···
∂ 2 φ N −1
∂t 2
= K
φ − Δx
Δx 2 N
2K
2 φ N −1 + K
φ
Δx 2 N −2
+ f (φ N −1 )
∂ 2 φN
∂t 2
= K
φ
Δx 2 N +1
− Δx
2K
2 φN + K
φ
Δx 2 N −1
+ f (φ N )

By defining the following state vector


x T = φ1 , φ2 , . . . , φ N (13.49)

one obtains the state-space description

ẍ1 = K
x
Δx 2 2
− 2K
x
Δx 2 1
+ K
φ
Δx 2 0
+ f (x1 )
ẍ2 = K
x
Δx 2 3
− 2K
x
Δx 2 2
+ K
x
Δx 2 1
+ f (x2 )
ẍ3 = K
x
Δx 2 4
− 2K
x
Δx 2 3
+ K
x
Δx 2 2
+ f (x3 )
(13.50)
···
ẍ N −1 = K
x − Δx
Δx 2 N
2K
2 x N −1 + K
x
Δx 2 N −2
+ f (x N −1 )
ẍ N = K
φ
Δx 2 N +1
− Δx
2K
2 xN + K
x
Δx 2 N −1
+ f (x N )

Next, the following state variables are defined

y1,i = xi
(13.51)
y2,i = ẋi

and the state-space description of the system becomes as follows

690 13 Differential Flatness Theory in the Background of Other …

ẏ1,1 = y2,1
ẏ2,1 = K
y
Δx 2 1,2
− 2K
y + Δx
Δx 2 1,1
K
2 φ0 + f (y1,1 )
ẏ1,2 = y2,2
ẏ2,2 = Δx K
2 y1,3 − Δx 2 y1,2 + Δx 2 y1,1 + f (y1,2 )
2K K

ẏ1,3 = y2,3
ẏ2,3 = Δx K
2 y1,4 − 2K
y + Δx
Δx 2 1,3
K
2 y1,2 + f (y1,3 )
(13.52)
···
···
ẏ1,N −1 = y2,N −1
ẏ2,N −1 = Δx K
2 y1,N − Δx 2 y1,N −1 + Δx 2 y1,N −2 + f (y1,N −1 )
2K K

ẏ1,N = y2,N
ẏ2,N = Δx K
2 φ N +1 − Δx 2 y1,N + Δx y1,N −1 + f (y1,N )
2K K

As shown in Chap. 12, the dynamical system described in Eq. (13.52) is a differen-
tially flat one with flat output defined as the vector ỹ = [y1,1 , y1,2 , . . . , y1,N ]. Indeed
all state variables can be written as functions of this flat output and its derivatives.

13.3.2.2 State-Space Description of the PDE Model in a Canonical Form

Finally, the problem of state-space description of the PDE dynamics is studied again.
Moreover, in the state-space model of Eq. (13.52) by defining the new control inputs

v1 = K
y − Δx
Δx 2 1,2
2K
2 y1,1 + Δx 2 φ0 + f (y1,1 )
K

v2 = K
y − Δx
Δx 2 1,3
2K
2 y1,2 + Δx 2 y1,1 + f (y1,2 )
K

v3 = K
y − Δx
Δx 2 1,4
2K
2 y1,3 + Δx 2 y1,2 + f (y1,3 )
K
(13.53)
···
v N −1 = K
y
Δx 2 1,N
− Δx
2K
2 y1,N −1 + Δx 2 y1,N −2 + f (y1,N −1 )
K

v N = Δx
K
2 φ N +1 − Δx 2 y1,N + Δx 2 y1,N −1 + f (y1,N )
2K K

the following state-space description is obtained

⎛ ⎞ ⎛ ⎞
0 1 0 0 ··· 0 0 0 0 0 0 0 ··· 0 0
⎛ ⎞ ⎛ ⎞
ẏ1,1 ⎜ 0 0 0 0 ··· 0 0 0 0 ⎟ y1,1 ⎜ 1 0 0 ··· 0 0 ⎟
⎜ ⎟ ⎜ ⎟
⎜ ẏ2,1 ⎟ ⎜ 0 0 0 1 · · · 0 0 0 0 ⎟ ⎜ y2,1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎛ ⎞
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ v1
⎜ ẏ1,2 ⎟ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎜ y1,2 ⎟ ⎜ 0 1 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ2,2 ⎟ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎜ y2,2 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ v2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ v3 ⎟
⎜ ··· ⎟ = ⎜ 0 0 0 0 ··· 0 0 0 0 ⎟⎜ ··· ⎟ + ⎜ 0 0 1 ··· 0 0 ⎟⎜ ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ ẏ1,N −1 ⎟ ⎜· · · · · · · · · · · · · · · · · ·⎟ ⎜ y1,N −1 ⎟ ⎜· · · · · · · · · · · ·⎟ ⎜ · · · ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎝v N −1 ⎠
⎜ ẏ2,N −1 ⎟ ⎜ 0 0 0 0 · · · 0 1 0 0 ⎟ ⎜ y2,N −1 ⎟ ⎜ 0 0 0 · · · 0 0 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ vN
⎝ ẏ1,N ⎠ ⎜ 0 0 0 0 · · · 0 0 0 0 ⎟ ⎝ y1,N ⎠ ⎜ 0 0 0 · · · 1 0 ⎟
⎜ ⎟ ⎜ ⎟
ẏ2,N ⎝ 0 0 0 0 ··· 0 0 0 1 ⎠ y2,N ⎝ 0 0 0 ··· 0 0 ⎠
0 0 0 0 ··· 0 0 0 0 0 0 0 ··· 0 1
(13.54)
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 691

The system of Eqs. (13.54) and (13.55) is in the controller and observer canonical
form. By selecting measurements from a subset of points x j j∈[1, 2, . . . , m], the
associated observation (measurement) equation becomes
⎞⎛
⎛ y1,1
⎞
⎛ ⎞ 1 0 0 ··· 0 0 ⎜ ⎟
⎜ y2,1 ⎟
z1 ⎜ 0 0 0 · · · 0 0 ⎟ ⎜ y1,2 ⎟
⎜ z2 ⎟ ⎜ ⎟⎜ ⎟
⎜ ⎟ = ⎜· · · · · · · · · · · ·⎟ ⎜ y2,2 ⎟ (13.55)
⎝· · ·⎠ ⎜ ⎟ ⎜ ⎟
⎝ 0 0 0 ··· 1 0 ⎠⎜ ··· ⎟
zm ⎜ ⎟
0 0 0 · · · 0 0 ⎝ y1,N ⎠
y2,N

Finally, in matrix form one has the following state-space description of the system

ỹ˙ = A ỹ + Bv
(13.56)
z̃ = C ỹ

13.3.3 Differential Flatness of the Nonlinear PDE Model

At the present section, a different approach will be followed to show that the state-
space description of the nonlinear PDE dynamics, which has as control input only
the boundary condition φ N +1 is a differentially flat one. One has

ẏ1,1 = y2,1
(13.57)
ẏ2,1 = − Δx
2K
2 y1,1 + K
φ + f (y1,1 ) +
Δx 2 0
K
y
Δx 2 1,2

ẏ1,2 = y2,2
(13.58)
ẏ2,2 = − Δx
2K
2 y1,2 + ΔxK
2 y1,1 + f (y1,2 ) +
K
y
Δx 2 1,3

ẏ1,3 = y2,3
ẏ2,3 = − Δx
2K
y
2 1,3 + K
y + f (y1,3 ) +
Δx 2 1,2
K
y
Δx 2 1,4
(13.59)
··· ···

ẏ1,i = y2,i
ẏ2,i = − Δx
2K
2 y1,i +
K
y
Δx 2 1,i−1
+ f (y1,i ) + K
y
Δx 2 1,i+1
(13.60)
··· ···

ẏ1,N −1 = y2,N −1
(13.61)
ẏ2,N −1 = − Δx
2K
2 1,N −1 + Δx 2 y1,N −2 + f (y1,N −1 ) +
y K K
y
Δx 2 1,N

ẏ1,N = y2,N
(13.62)
ẏ2,N = − Δx
2K
2 y1,N +
K
y
Δx 2 1,N −1
+ f (y1,N ) + K
φ
Δx 2 N +1
692 13 Differential Flatness Theory in the Background of Other …

The flat output is considered to be the state variable y1,1 , which is denoted as y = y1,1 .
Next, it is shown that all state variables which stand also for virtual control inputs of
the system αi = y1,i+1 , can be written as functions of the flat output y = y1,1 .
From Eq. (13.57) and considering that φ0 is constant one obtains

yi,2 = α1 = 1
[ ÿ
K /Δx 2 1,1
+ Δx2K
2 y1,1 − Δx 2 φ0 − f (y1,1 )]
K
(13.63)
⇒y1,2 = h 1 (y, ẏ, ÿ, . . .)

From Eq. (13.58) one obtains

y1,3 = α2 = 1
[ ÿ
K /Δx 2 1,2
+ Δx2K
2 y1,2 − Δx 2 y1,1 − f (y1,2 )]
K
(13.64)
⇒y1,3 = h 2 (y, ẏ, ÿ, . . .)

From Eq. (13.59) one obtains

y1,4 = α3 = 1
[ ÿ
K /Δx 2 1,3
+ Δx2K
2 y1,3 − Δx 2 y1,2 − f (y1,3 )]
K
(13.65)
⇒y1,4 = h 3 (y, ẏ, ÿ, . . .)

Continuing in a similar manner, from Eq. (13.60) one obtains

y1,i+1 = αi = 1
[ ÿ
K /Δx 2 1,i
+ Δx
2K
2 y1,i − Δx 2 y1,i−1 − f (y1,i )]
K
(13.66)
⇒y1,i+1 = h i (y, ẏ, ÿ, . . .)

and following a similar procedure, from Eq. (13.61) one gets

y1,N = α N −1 = 1
[ ÿ
K /Δx 2 1,N −1
+ Δx
2K
2 y1,N −1 − Δx 2 y1,N −2 − f (y1,N −1 )]
K

⇒y1,N = h N −1 (y, ẏ, ÿ, . . .)

(13.67)
Finally, from Eq. (13.62) one gets

φ N +1 = α N −1 = 1
[ ÿ
K /Δx 2 1,N
+ Δx
2K
2 y1,N − Δx 2 y1,N −1 − f (y1,N )]
K
(13.68)
⇒φ1,N +1 = h N (y, ẏ, ÿ, . . .)

The above procedure confirms that all state variables of the model

y1,1 , y2,1 y1,2 , y2,2 y1,3 , y2,3 ···

(13.69)
y1,i , y2,i ··· y1,N −1 , y2,N −1 y1,N , y2,N

and the control input which is the boundary condition φ N +1 can be written as func-
tions of the flat output y = y1,1 and of the flat output’s derivatives. Consequently,
the state-space model of the PDE dynamics is differentially flat.
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 693

Additionally, one can consider decomposition of the PDE state-space equation

into submodels, where at each submodel the virtual control input is αi = y1,i+1

ẏ1,i = y2,i
ẏ2,i = − Δx
2K
2 y1,i +
K
y
Δx 2 1,i−1
+ f (y1,i ) + K
y
Δx 2 1,i+1
(13.70)
··· ···

and with local flat output y1,i . Thus one can confirm that all such subsystems i =
1, 2, . . . , N are differentially flat. Next, one can compute the virtual inputs which
are applied to each subsystem.

13.3.4 Computation of a Boundary Conditions-Based

Feedback Control Law

To implement boundary feedback control in the system of Eqs. (13.57)–(13.62), the

nonlinear wave PDE model is rewritten as follows:

ÿ1,1 = − Δx
2K
2 y1,1 +
K
φ
Δx 2 0
+ f (y1,1 ) + K
y
Δx 2 1,2
(13.71)

ÿ1,2 = − Δx
2K
2 y1,2 +
K
y
Δx 2 1,1
+ f (y1,2 ) + K
y
Δx 2 1,3
(13.72)

ÿ1,3 = − Δx
2K
2 y1,3 +
K
y
Δx 2 1,2
+ f (y1,3 ) + K
y
Δx 2 1,4 (13.73)
··· ···

ÿ1,i = − Δx
2K
2 y1,i +
K
y
Δx 2 1,i−1
+ f (y1,i ) + K
y
Δx 2 1,i+1 (13.74)
··· ···

ÿ1,N −1 = − Δx
2K
2 y1,N −1 +
K
y
Δx 2 1,N −2
+ f (y1,N −1 ) + K
y
Δx 2 1,N
(13.75)

ÿ1,N = − Δx
2K
2 y1,N +
K
y
Δx 2 1,N −1
+ f (y1,N ) + K
φ
Δx 2 N +1
(13.76)

The boundary condition φ0 is assumed to be known and remains steady. The boundary
condition φ N +1 stands for the control input. The feedback control law is designed
as follows:
∗ =
α1 = y1,2 [ ÿ d − kd,1 ( ẏ1,1 − ẏ1,1
d )−k
p,1 (y1,1 − y1,1
1 d )+
(K /Δx 2 ) 1,1
(13.77)
+ Δx2K
2 y1,1 − Δx 2 φ0 − f (y1,1 )]
K
694 13 Differential Flatness Theory in the Background of Other …

∗ =
α2 = y1,3 [ ÿ d − kd,2 ( ẏ1,2 − ẏ1,2
d )−k
p,2 (y1,2 − y1,2 )+
1 d
(K /Δx 2 ) 1,2
+ Δx2K
2 y1,2 − Δx 2 y1,1 − f (y1,2 )]⇒
K
(13.78)
α2 = 1
[α̈ − kd,2 ( ẏ1,2 − α̇1 ) − k p,2 (y1,2 − α1 )+
(K /Δx 2 ) 1
+ Δx
2K
2 y1,2 −
K
y
Δx 2 1,1
− f (y1,2 )]

∗ =
α3 = y1,4 [ ÿ d − kd,3 ( ẏ1,3 − ẏ1,3
d )−k
p,3 (y1,3 − y1,3
1 d )+
(K /Δx 2 ) 1,3
+ Δx 2 y1,3 − Δx 2 y1,2 − f (y1,3 )]⇒
2K K
(13.79)
α3 = 1
[α̈ − kd,3 ( ẏ1,3 − α̇2 ) − k p,3 (y1,3
(K /Δx 2 ) 2
− α2 )+
+ Δx 2K
2 y1,3 − Δx 2 y1,2 − f (y1,3 )]
K

and continuing in a similar manner

∗
αi = y1,i+1 = 1
[ ÿ d − kd,i ( ẏ1,i − ẏ1,i
d ) − k (y − y1,i
d )+
(K /Δx 2 ) 1,i p,i 1,i

+ Δx 2 y1,i − Δx 2 y1,i−1 − f (y1,i )]⇒

2K K
(13.80)
αi = 1
[α̈
(K /Δx 2 ) i−1
− kd,i ( ẏ1,i − α̇i−1 ) − k p,i (y1,i − αi−1 )+
+ Δx2K
2 y1,i − Δx 2 y1,i−1 − f (y1,i )]
K

Following this procedure one arrives to compute the control inputs which are asso-
ciated with the last two rows of the state-space model
∗
α N −1 = y1,N = [ ÿ d − kd,N −1 ( ẏ1,N −1 − ẏ1,N −1 ) − k p,N −1 (y1,N −1 − y1,N −1 )+
1 d d
(K /Δx 2 ) 1,N −1

+ Δx 2 y1,N −1 − Δx 2 y1,N −2 − f (y1,N −1 )]⇒

2K K

α N −1 = 1
[α̈
(K /Δx 2 ) N −2
− kd,N −1 ( ẏ1,N −1 − α̇ N −2 ) − k p,N −1 (y1,N −1 − α N −2 )+
+ Δx 2 y1,N −1 − Δx
2K K
2 y1,N −2 − f (y1,N −1 )]
(13.81)

φ N +1 = 1
[ ÿ d − kd,N ( ẏ1,N − ẏ1,N
(K /Δx 2 ) 1,N
d )−k
p,N (y1,N − y1,N )+
d

+ Δx2K
2 y1,N − Δx 2 y1,N −1 − f (y1,N )]⇒
K
(13.82)
αN = 1
[α̈
(K /Δx 2 ) N −1
− kd,N ( ẏ1,N − α̇ N −1 ) − k p,N (y1,N − α N −1 )+
+ Δx 2 y1,N − Δx
2K K
2 y1,N −1 − f (y1,N )]

Consequently, the computation of the aggregate control input φ N +1 which is exerted

on the PDE model is performed backwards, by substituting recursively into φ N +1
the virtual control inputs a N −1 , a N −2 ,. . ., ai ,. . ., a2 , a1 . It is noted that the concept
of flatness-based control implementation in cascading loops has been already used
in lumped parameter systems and the above results show that it can also applied
effectively to distributed parameter systems [117, 440, 442].
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 695

13.3.5 Closed-Loop Dynamics

By substituting Eq. (13.82) into Eq. (13.76) of the state-space model of the PDE
dynamics, and using the definition y1,N − α N −1 = z N one has

ÿ1,N = α̈ N −1 − kd,N ( ẏ N − α̇ N −1 ) − k p,N (y1,N − α N −1 )⇒

( ÿ1,N − α̈ N −1 ) + kd,N ( ẏ1,N − α̇ N −1 ) + k p,N (y1,N − α N −1 ) = 0⇒ (13.83)
z̈ N + kd,N ż N + k p,N z N = 0

By substituting Eq. (13.81) into Eq. (13.75) of the state-space model of the PDE
dynamics, and using the definition y1,N −1 − α N −2 = z N −1 one has

ÿ1,N −1 = α̈ N −2 − kd,N −1 ( ẏ N −1 − α̇ N −2 ) − k p,N −1 (y1,N −1 − α N −2 )⇒

( ÿ1,N −1 − α̈ N −2 ) + kd,N −1 ( ẏ1,N −1 − α̇ N −2 ) + k p,N −1 (y1,N −1 − α N −2 ) = 0⇒
z̈ N −1 + kd,N −1 ż N −1 + k p,N −1 z N −1 = 0
(13.84)
Moving backward, and by substituting Eq. (13.80) into Eq. (13.74) of the state-space
model of the PDE dynamics, and using the definition y1,i − αi−1 = z i one has

ÿ1,i = α̈i−1 − kd,i ( ẏi − α̇i−1 ) − k p,i (y1,i − αi−1 )⇒

( ÿ1,i − α̈i−1 ) + kd,i ( ẏ1,i − α̇i−1 ) + k p,i (y1,i − αi−1 ) = 0⇒ (13.85)
z̈ i + kd,i ż i + k p,i z i = 0

Similarly, continuing with the rest of the equations of the state-space model and by
substituting Eq. (13.79) into Eq. (13.73), while also using the definition y1,3 −α2 = z 3
one has

ÿ1,3 = α̈2 − kd,3 ( ẏ3 − α̇2 ) − k p,3 (y1,3 − α2 )⇒

( ÿ1,3 − α̈2 ) + kd,3 ( ẏ3 − α̇2 ) + k p,3 (y1,3 − α2 ) = 0⇒ (13.86)
z̈ 3 + kd,3 ż 3 + k p,3 z 3 = 0

Equivalently, by substituting Eq. (13.78) into Eq. (13.72), and using the definition
y1,2 − α1 = z 2 one has

ÿ1,2 = α̈1 − kd,2 ( ẏ2 − α̇1 ) − k p,2 (y1,2 − α1 )⇒

( ÿ1,2 − α̈1 ) + kd,2 ( ẏ2 − α̇1 ) + k p,2 (y1,2 − α1 ) = 0⇒ (13.87)
z̈ 2 + kd,2 ż 2 + k p,2 z 2 = 0

Finally, by substituting Eq. (13.77) into Eq. (13.71), and using the definition y1,1 −
d = z one obtains
y1,1 1

ÿ1,1 = ÿ1,1
d − k ( ẏ
d,1 1,1 − ẏ1,1 ) − k p,1 (y1,1 − y1,1 )⇒
d d

( ÿ1,1 − ÿ1,1 ) + kd,1 ( ẏ1,1 − ẏ1,1 ) + k p,1 (y1,1 − y1,1 ) = 0⇒

d d (13.88)
z̈ 1 + kd,1 ż 1 + k p,1 z 1 = 0
696 13 Differential Flatness Theory in the Background of Other …

Thus, the dynamics of the closed-loop system becomes

z̈ 1 + kd,1 ż 1 + k p,1 z 1 = 0
z̈ 2 + kd,2 ż 2 + k p,2 z 2 = 0
z̈ 3 + kd,3 ż 3 + k p,3 z 3 = 0
··· ···
(13.89)
z̈ i + kd,i ż i + k p,i z i = 0
··· ···
z̈ N −1 + kd,N −1 ż N −1 + k p,N −1 z N −1 = 0
z̈ N + kd,N ż N + k p,N z N = 0

The dynamics of the closed-loop system can be also written in matrix form

Z̃¨ + K D Z̃˙ + K P Z̃ = 0 (13.90)

where
Z̃ = [z 1 , z 2 , z 3 , . . . , z i , . . . , z N −1 , z N ]T
K D = diag[kd,1 , kd,2 , kd,3 , . . . , kd,i , . . . , kd,N −1 , kd,N ]
K P = diag[k p,1 , k p,2 , k p,3 , . . . , k p,i , . . . , k p,N −1 , k p,N ]

After suitable selection of the coefficients kd,i , k P,i , i = 1, 2, . . . , N such that the
polynomials p(s) = s 2 + kd,i s + k p,i to be Hurwitz ones, it can be assured that
limt→∞ z i (t) = 0 ∀ i = 1, . . . , N and that the closed-loop system is asymptotically
stable.
Moreover, to prove asymptotic stability for the closed-loop system, the following
Lyapunov function can be used
N kd,i 2 N
V = i=1 2 z i + 1 2
i=1 2 ż i (13.91)

The derivative of this Lyapunov function with respect to time is

N N
V̇ = i=1 k p,i z i ż i + i=1 ż i z̈ i ⇒
N N
V̇ = i=1 k p,i z i ż i + i=1 ż i (−kd,i ż i − k pi z i )⇒
N N N (13.92)
V̇ = i=1 k p,i z i ż i − i=1 kd,i ż i2 − i=1 k p,i z i ż i ⇒
N
V̇ = − i=1 kd,i ż i < 0 2

Thus, it is proven again that the closed-loop system is globally asymptotically stable
(Fig. 13.13).
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 697

φ
Fig. 13.13 a Plot of the 2D-wave function after applying boundary control, b plot of the reference
2D-wave function

13.3.6 Simulation Tests

Simulation examples about the proposed control method for distributed parame-
ter systems are provided for the case of the nonlinear wave PDE that was given in
Eq. (13.45). The obtained results are depicted in Figs. 13.14, 13.15, 13.16, and 13.17.
The spatial discretization of the PDE model consisted of N = 25 points. The sam-
pling period was chosen to be Ts = 0.01 s. The boundary condition φ0 of the PDE

(a) 5 (b) 500

4 400

3 300
boundary control φN

2 200

1 100

0 0
1

−1 −100

−2 −200

−3 −300

−4 −400

−5 −500
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
t t

Fig. 13.14 a Variation in time of the solution of the PDE at grid point x1 (blue line) and associated
reference setpoint (red line), b control input exerted to the PDE system through the boundary
condition φ N +1
698 13 Differential Flatness Theory in the Background of Other …

(a) 5 10 (b) 20 20

5 10 10

0 0 0 0
2

7
−5 −10 −10

−5 −10 −20 −20

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40

time time time time

10 20 20 20

5 10 10
0
0 0 0
5

9
4

−20
−5 −10 −10

−10 −20 −20 −40

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 13.15 a Variation in time of the solution of the PDE at grid points x2 to x5 (blue line) and
associated reference setpoint (red line), b variation in time of the solution of the PDE at grid points
x6 to x9 (blue line) and associated reference setpoint (red line)

(a) 40 40 (b) 40 40

20 20 20 20
11

14

15
10

0 0 0 0

−20 −20 −20 −20

−40 −40 −40 −40

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
40 40 40 50

20 20 20
13

16

17
12

0 0 0 0

−20 −20 −20

−40 −40 −40 −50

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 13.16 a Variation in time of the solution of the PDE at grid points x10 to x13 (blue line) and
associated reference setpoint (red line), b variation in time of the solution of the PDE at grid points
x14 to x17 (blue line) and associated reference setpoint (red line)

was taken to be known and constant. The boundary condition φ N +1 served as the
control input and was computed at each iteration of the control algorithm according
to the procedure described in Sect. 13.3.4.
The numerical simulation experiments have confirmed the theoretical findings
of this section. It has been shown that by applying the proposed control method the
nonlinear PDE dynamics can be modified so as to converge to the desirable reference
profile. The control input that succeeds this, changes smoothly and has a moderate
range of variation. The accuracy of tracking of the reference setpoints was indeed
13.3 Boundary Control of Nonlinear PDE Dynamics Using … 699

(a) 50 50 (b) 100 100

50 50
18

19

22

23
0 0 0 0

−50 −50

−50 −50 −100 −100

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time
50 50 100 400

200
0 0
20

21

24

25
0 0
−50 −100
−200

−50 −100 −200 −400

0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
time time time time

Fig. 13.17 a Variation in time of the solution of the PDE at grid points x18 to x21 (blue line) and
associated reference setpoint (red line), b variation in time of the solution of the PDE at grid points
x22 to x25 (blue line) and associated reference setpoint (red line)

satisfactory. It is noted that the application of the proposed nonlinear PDE feedback
control method is not limited to the case of nonlinear wave dynamics but can be also
extended to several other types of PDEs of the parabolic, hyperbolic, and elliptic
type and to various types of nonlinear PDEs.
References

1. Aguglia, P., Voiarouge, P., Wamkeue, R., Cros, J.: Determination of fault operation dynamical
constraints for the design of wind turbine DFIG drives. Math. Comput. Simul., Elsevier 81,
252–262 (2010)
2. Aguiar, A.P., Hespanha, J.P.: Trajectory-tracking and path-following of underactuated au-
tonomous vehicles with parametric modelling uncertainty. IEEE Trans. Autom. Control
52(3), 1362–1379 (2007)
3. Aguilar-Ibánez, C., Sira-Ramirez, H., Suárez-Castañón, M.S., Martínez-Navarro, E.,
Moreno-Armendariz, M.A.: The trajectory tracking problem for an unmanned fourrotor
system: flatness-based approach. Int. J. Control, Taylor and Francis 85(1), 69–77 (2012)
4. Akin, B., Orguner, U., Ersak, A.: State estimation of induction motor using unscented Kalman
filter. In: Control Applications, Proceedings of the IEEE CCA 2003 Conference, Instabul,
Turkey (2003)
5. Akin, B., Orguner, U., Ersak, A.: A comparative study on Kalman filtering techniques de-
signed for state estimation of industrial AC drive systems. In: Proceedings of the IEEE
International Conference on Mechatronics 2004, ICM’04, Istanbul, June 2004
6. Akin, B., Orguner, U., Ersak, A.: Simple derivative-free nonlinear state observer for sensor-
less AC drives. IEEE/ASME Trans. Mechatron. 11(5), 634–643 (2006)
7. Alanis, A.Y., Sanchez, E.N., Ricalde, R.J., Perez-Cisneros, M.A.: Discrete-time reduced
order neural observers for uncertain nonlinear systems. Int. J. Neural Syst., World Scientific
20(1), 29–38 (2010)
8. Ali, A., Blath, J.P.: Nonlinear torque control of a Spark-Ignited engine. In: Proceedings of
the American Control Conference, Minneapolis, Minnesota, June 2006
9. Allag, A., Hammoudi, M., Mimoune, S.M., Ayad, M.Y., Becherif, M., Miraoui A.: Tracking
control via adaptive backstepping approach for a three-phase PWM AC-Dc converter. In:
IEEE ISIE 2007 International Symposium on Industrial Electronics (2007)
10. Alipi, C., de Russion, C., Piuri, V.: A neural network-based control solution to air-fuel ratio
control for automotive fuel injection systems. IEEE Trans. Syst. Man Cybern. C Appl. Rev.
33(2), 259–268 (2003)
11. Altifai, M.T., Hassan, M.F., Zribi, M.: Decentralized load frequency controller for a multi-
area interconnected power system. Electr. Power Energy Syst., Elsevier 33, 198–209 (2011)
12. Alonge, F., D’Ippolito, A.: Robustness analysis of an extended Kalman filter for sensorless
control of induction motors. In: IEEE International Symposium on Industrial Electronics,
ISIE 2010, Bari, Italy, July 2010

© Springer International Publishing Switzerland 2015 701

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5
702 References

13. An, L., Sepehri, L.: Leakage fault identification in a hydraulic positioning system using
extended Kalman filter. In: Proceedings of the 2004 American Control Conference Boston,
Massachusetts, June–July 2004
14. Andone, D., Dobrescu, R., Hossu, A., Dobrescu, M.: Application of fuzzy model prediction
control to a drum boiler. Integr. Comput. Aided Eng., IOP Press 13(4), 347–359 (2006)
15. Anene, E.C., Aliya, U.O., Lévine, J., Vemayagamoorthy, G.K.: Flatness-based feedback
linearization of a synchronous machine model with static excitation and fast turbine valving.
In: IEEE 2007 Power Engineering, General Meeting, Tampa, Florida, USA, June 2007
16. Aoustin, Y., Fliess, M., Mounier, H., Rouchon P., Rudoplh J.: Theory and practice in the
motion planning control of a flexible robot arm using Mikusiǹski operators. In: Proceedings
of 4th Symposium on Robotics and Control, pp. 287–293, Nantes (1997)
17. Ariaei, F., Jonckheere, E.: LDV approach to circular trajectory tracking of the underactuated
hovercrafts with modelling parameter uncertainties. In: Proceedings of the 2006 American
Control Conference, Minneapolis, Minnesota USA, June 2006
18. Arsie, I., Di Iorio, S., Pianese, C., Rizzo, G., Sorrentino, M.: Recurrent neural networks for
air-fuel ratio estimation and control in spark-ignited engines. In: Proceedings of the 17th
IFAC World Congress, Seoul, Korea, July 2008
19. Arsie, I., Pianese, C., Sorrentino, M.: A procedure to enhance identification of recurrent
neural networks for simulating air-fuel ratio dynamics in SI engines. Eng. Appl. Artif. Intell.
19(1), 65–77 (2006)
20. Arsie, I., Pianese, C., Rizzo, G., Cioffi, V.: An Adaptive estimator of fuel film dynamics
in the intake port of a spark ignition engine. Control Eng. Pract., Elsevier 11(3), 303–309
(2003)
21. Arteaga, M.A., Siciliano, B.: On Tracking control of flexible robot arms. IEEE Trans. Autom.
Control 45(3), 520–527 (2000)
22. Astrom, K.J., Wittenmark, B.: Adaptive Control. Addison Wesley, Boston (1995)
23. Auat, C.F., Steiner, G., Perez, P.G., Carelli, R.: Optimized EIF-SLAM algorithm for precision
agriculture mapping based on stem detection. Comput. Electron. Agric. (Elsevier). 78, 195–
207 (2011)
24. Azou, S., Burel, G.: Design of a Chaos-Based Spread-Spectrum Communication System
Using Dual Unscented Kalman Filters. IEEE Communications 2002, Bucharest, Romania
(2002)
25. Azou, S., Luca, M.B., Burel, G.: Automatic Gain Control in a Kalman Filter-Based Synchro-
nization Chaotic Receiver. IEEE Communications 2004, Bucharest, Romania, June 2004
26. Bagordo, G., Cazzulani, G., Resta, F., Ripamonti, F.: A modal disturbance estimator for
vibration suppression in nonlinear flexible structures. J. Sound Vib., Elsevier 330, 6061–
6069 (2011)
27. Ballachi, A., Benvenuti, L., di Benedetto, M.D., Sangiovanni-Vincentelli, A.: The design of
dynamical observers for hybrid systems: theory and application of an automotive control
problem. Automatica, Elsevier 49(4), 915–925 (2013)
28. Bao, X., Zhuo, F., Tian, Y., Tan, P.: Simplified feedback linearization control of three-phase
photovoltaic inverter with an LCL filter. IEEE Trans. Power Electron. 28(6), 2739–2752
(2013)
29. Barrera-Cardiel, E., Pastor-Gomez, N.: Microcontroller-based power-angle instrument for
a power system laboratory. In: IEEE Powering Engineering Society Summer Meeting, pp.
1008–1012 (1999)
30. Bar-Shalom, Y.: Update with out-of-sequence measurements in tracking: exact solution.
IEEE Trans. Aerosp. Electron. Syst. 38, 769–778 (2007)
31. Basseville, M., Nikoforov, I.: Detection of Abrupt Changes: Theory and Applications,
Prentice-Hall, Englewood Cliffs (1993)
32. Basseville, M., Benveniste, A., Zhang, Q.: Surveilliance d’ installations industrielles:
démarche générale et conception de l’ algorithmique. IRISA Publication Interne No 1010
(1996)
References 703

33. Basseville, M., Benveniste, A., Gach-Devauchelle, B., Goursat, M., Bonnecase, D., Doray,
P., Prevosto, M., Olagnon, M.: In situ damage monitoring in vibration mechanics: diagnostics
and predictive maintenance. Mech. Syst. Sig. Process., Elsevier 7(5), 401–423 (1993)
34. Basturk, H.I., Kristic, M.: Adaptive wave cancellation by acceleration feedback for ramp-
connected air-cushion-actuated surface effect ships. Automatica, Elsevier 49, 2591–2602
(2010)
35. Becker, J., Meurer, T.: Feedforward tracking control for non-uniform Timoshenko beam
models: combining differential flatness, modal analysis, and FEM. ZAMM J. Appl. Math.
Mech., Wiley 87(1), 37–58 (2005)
36. Benveniste, A., Basseville, M., Moustakides, G.: The asymptotic local approach to change
detection and model validation. IEEE Trans. Autom. Control 32(7), 583–592 (1987)
37. Bertoglio, C., Chapelle, D., Fernandez, M.A., Gerbeau, J.F., Moireau, P.: State observers
of a vascular fluid-structure interaction model through measurements in the solid, INRIA
research report no 8177, Dec 2012
38. Besançon, G., Zhang, Q., Hammouri, H.: High-gain observer based state and parameter esti-
mation in nonlinear systems, IFAC Symposium on Nonlinear Control Systems (NOLCOS),
Stuttgart, Germany (2004)
39. Berman, S., Edan, Y.: Intelligent automatic guided vehicles. In: Rigatos, G. (ed.) Intelligent
Industrial Systems: Modelling, Automation and Adaptive Behavior. IGI Publications (2010)
40. Beltran-Carbajal, F., Chavez-Conde, E., Silva Navarro, G., Vazquez Gonzalez, B., Favela
Contreras, A.: Control of nonlinear active vehicle suspension systems using disturbance
observers. In: Vibration Analysis and Control—New Trend and Developments. InTech
Publications (2011)
41. Bertini, D., Bittanti, S., Savaresi, S.M.: Decoupled cushion control in ride control systems
for air cushion catamarans. Control Eng. Pract., Elsevier 8, 191–203 (2000)
42. Bertoglio, C., Chapelle, D., Fernandez, M.A., Gerbeau, J.F., Moireau, P.: State observers
of a vascular fluid-structure interaction model through measurements in the solid, INRIA
research report no 8177, Dec 2012
43. Besnard, L., Shtessel, Y.B., Landrum, B.: Quadrotor vehicle control via sliding mode con-
troller driven by sliding mode disturbance observer. J. Franklin Inst., Elsevier 349, 658–684
(2012)
44. Bevly, D.M., Ryu, J., Gerdes, J.C.: Integrating INS sensors With GPS measurements for
continuous estimation of vehicle sideslip, roll, and tire cornering stiffness. IEEE Trans.
Intell. Transp. Syst. 7(4), 483–493 (2006)
45. Bidram, A., Davoudi, A., Lewis, F.L., Guerrero, J.M.: Distributed cooperative secondary
control of microgrids using feedback linearization. IEEE Trans. Power Syst. 28(3), 3462–
3470 (2013)
46. Bidram, A., Lewis, F.L., Davoudi, A.: Synchronization of nonlinear heterogeneous coopera-
tive systems using input-output feedback linearization. Automatica, Elsevier 50, 2578–2585
(2014)
47. Blaabjerg, F., Teodorescu, R., Liserre, M., Timbus, A.V.: Overview of control and grid
synchronization for distributed power generation systems. IEEE Trans. Industr. Electron.
53(5), 1398–1409 (2006)
48. Blake Vance, J., Kaul, B.C., Jagannathan, S., Drallmeier, J.A.: Output feedback controller
for operation of spark ingition engines at lean conditions using neural networks. IEEE Trans.
Control Syst. Technol. 16(2), 214–227 (2008)
49. Bodson, M., Chiasson, J., Novotnak, R.: High-performance induction motor control via
input-output linearization. IEEE Control Syst. Mag. 14(4), 24–33 (1994)
50. Boizot, N., Busvelle, E., Gauthier, J.-P.: An adaptive high-gain observer for nonlinear sys-
tems. Automatica, Elsevier 46, 1483–1488 (2010)
51. Borsje, P., Chan, T.F., Wong, Y.K., Ho, S.L.: A comparative study of Kalman filtering for
sensorless control of a permanent-magnet synchronous motor drive. In: 2005 IEEE Inter-
national Conference on Electric Machines and Drives, art. no. 1531434, pp. 815–822, San
Antonio, Texas (2005)
704 References

52. Bosnak, M., Matko, D., Blasic, S.: Quadrocopter control using an on-board video system
with off-board processing. Robot. Auton. Syst., Elsevier 60, 657667 (2012)
53. Boudaond, M., Haddad, Y., Le Gorrec, Y.: Modelling and optimal force control of a nonlinar
electrostatic microgripper. IEEE/ASME Trans. Mechatron. 18, 1130–1139 (2012)
54. Boukhezzar, B., Siguerdidjane, H.: Nonlinear control with wind estimation of a DFIG vari-
able speed wind turbine for power capture optimization. Energy Convers. Manage., Elsevier
50, 885–892 (2009)
55. Bououden, S., Boutat, D., Zheng, G., Barbot, J.P., Kratz, F.: A triangular canonical form for
a class of 0-flat nonlinear systems. Int. J. Control, Taylor and Francis 84(2), 261–269 (2011)
56. Blackmore, S., Stout, W., Wang, M., Runov, B.: Robotic agriculture—the future of agri-
cultural mechanisation? In: Stafford, J.V. (ed.) Fifth European Conference on Precision
Agriculture, pp. 621–628. Wageningen Academic Publishers, The Netherlands (2005)
57. Blake Vance, J., Kaul, B.C., Jagannathan, S., Drallmeier, J.A.: Output feedback controller
for operation of spark ingition engines at lean conditions using neural networks. IEEE Trans.
Control Syst. Technol. 16(2), 214–227 (2008)
58. Blasco, V., Kaura, V.: A new mathematical model and control of a three-phase AC-DC
voltage source converter. IEEE Trans. Power Electron. 12(1), 116–123 (1997)
59. Boussaada, I., Cela, A., Mounier, H., Niculescu, S.I.: Control of drilling vibrations: a time-
delay system-based approach. In: 11th Workshop on Time Delay Systems, France, Feb 2013
60. Brahmi, J., Krichen, L., Ouali, A.: A comparative study between three sensorless control
strategies for PMSG in wind energy conversion systems. Appl. Energy, Elsevier 86, 1565–
1573 (2009)
61. Bresch-Pietri, D., Chauvin, J., Petit, N., Controller, adaptive backstepping, for uncertain
systems with unknown input time-delay. In: Application to SI Engines, 49th IEEE Conference
on Decision and Control, Atlanta, GA, Dec 2010
62. Brdys, M.A., Kulowski, G.J.: Dynamic neural controllers for induction motor. IEEE Trans.
Neural Networks 10(2), 340–355 (1999)
63. Calderaro, V., Galdi, V., Piccolo, A., Siano, P.: A fuzzy controller for maximum energy
extraction from variable speed wind power generation systems. Electr. Power Syst. Res.,
Elsevier 78, 1109–1118 (2008)
64. Campillo, F., Mevel, L.: Recursive maximum likelihood estimation for structural health
monitoring: tangent filter implementations. In: CDC-ECC ’05, 44th IEEE Conference on
Decision and Control 2005 and 2005 European Control Conference
65. Campos-Gaona, D., Moreno-Goytia, E.L., Anaya-Lara, O.: Fault ride-through improvement
of DFIG-NT by integrating a two-degrees-of-freedom internal model control. IEEE Trans.
Industr. Electron. 60(3), 1133–1145 (2013)
66. Capecce, S.L., Cecati, C., Rotondale, N.: A sensorless control technique for low cost AC/DC
converters. In: IEEE Industry Applications Conference 2003, 38th IAS Annual Meeting, vol.
3, pp. 1546–1551, 2003
67. Capecce, S.L., Cecati, C., Rotondale, N.: A new three-phase active rectifier based on power
matching modulation. In: IEEE IECON 2003, 29th Annual Conference of the Industrial
Electronics Society, vol. 1, pp. 202–207 (2003)
68. Caron, F., Duflos, E., Pomorski, D., Vanheeghe, P.: GPS/IMU data fusion using multi-sensor
Kalman Filtering: introduction of conceptual aspects. Inf. Fusion, Elsevier 7, 221–230 (2006)
69. Cariou, C., Lemain, R., Thuilot, B., Verducat, M.: Automatic guidance of a four-wheel-
steering mobile robot for accurate field operations. J. Field Robot., Wiley 26, 504–516
(2009)
70. Castelli-Dezza, F., Iocchetti, M.F., Perini, P.: An observer for sensorless DFIM drives based
on the natural fifth harmonic of the line voltage, without stator current measurements. IEEE
Trans. Industr. Electron. 60(10), 4301–4309 (2013)
71. Castilla, M., Miret, J., Camacho, A., Matas, J., García de Vicuña, L.: Reduction of current
harmonic distortion in three-phase grid-connected photovoltaic inverters via resonant current
control. IEEE Trans. Industr. Electron. 60(4), 1464–1472 (2013)
References 705

72. Cecati, C., Dell’ Aquila, A., Lecci, A., Liserre, M.: Implementation issues of a fuzzy-logic-
based three-phase active rectifier employing only voltage sensors. IEEE Trans. Industr. Elec-
tron. 52(2), 378–385 (2005)
73. Cecati, C., Ciancetta, F., Siano, P.: A multilevel inverter for photovoltaic systems with fuzzy
logic control. IEEE Trans. Industr. Electron. 57(12), 4115–4125 (2010)
74. Cecati, C., Ciancetta, F., Siano, P.: A FPGA/fuzzy logic—based multilevel inverter. In:
Proceedings of the ISIE 2009, 18th IEEE International Symposium on Industrial Electronics,
pp. 706–711 (2009)
75. Cecati, C., Dell’ Aquila, A., Liserre, M., Ometto, A.: A fuzzy-logic-based controller for
active rectifier, IEEE Trans. Ind. Appl. 39(1), 105–112 (2003)
76. Chladny, R.R., Koch, C.R.: Flatness-based tracking of an electromechanical variable valve
timing actuator with disturbance observer feedforward compensation. IEEE Trans. Control
Syst. Technol. 16(4), 652–663 (2008)
77. Chamseddine, A., Li, T., Zhang, Y., Rabbath, C.A., Theilliol, D.: Flatness-based trajectory
planning for a quadrotor unmanned aerial vehicle test-bed considering actuator and system
constraints. In: 2012 American Control Conference Fairmont Queen Elizabeth, Montréal,
Canada, 27–29 June 2012
78. Chang, C.H., Lin, Y.H., Chen, Y.M., Chang, Y.R.: Simplified reactive power control for
single-phase grid-connected photovoltaic inverters. IEEE Trans. Industr. Electron. 61(5),
2286–2296 (2014)
79. Chang, Y.C.: A robust tracking control for chaotic Chua’s circuits via fuzzy approach. IEEE
Trans. Circ. Syst. I 48(7), 889–895 (2001)
80. Chang, Y.D.: Digital secure communication via chaotic systems. Digital Sig. Process.,
Elsevier 19, 693–699 (2009)
81. Chauvin, J.: Observer design for a class of wave equations driven by an unknown periodic
input. In: 18th World Congress. Milano, Italy (2011)
82. Chauvin, J., Corde, G., Moulin, P., Castagné, M., Petit, N., Rouchon, P.: Real-time combus-
tion torque estimation on a diesel engine test bench using time-varying Kalman filtering. In:
43rd IEEE Conference on Decision and Control, Bahamas, Dec 2004
83. Chauvin, J., Moulin, P., Corde, G., Petit, N., Rouchon, P.: Filtering, Kalman, for real-time
individual cylinder air fuel ratio observer on a diesel engine test bench. In: Proceedings:
American Control Conference. Minneapolis, Minnesota (2006)
84. Chavarría, J., Biel, D., Guinjoan, F., Meza, C., Negroni, J.J.: Energy-balance control of PV
cascaded multilevel grid-connected inverters under level-shifted and phase-shifted pwms.
IEEE Trans. Industr. Electron. 60(1), 98–111 (2013)
85. Chen, W.H., Ballance, D.J., Gawthrop, P.J., Reilly, J.O.: A nonlinear disturbance observer
for robotic manipulators. IEEE Trans. Industr. Electron. 47(4), 932–938 (2000)
86. Chen, L., Mercorelli, P., Liu, S.: A Kalman estimator for detecting repetitive disturbances.
In: ACC 2005, American Control Conference, Portland, Oregon (2005)
87. Chen, B.S., Lee, C.H., Chang, Y.C.: H∞ tracking design of uncertain nonlinear SISO sys-
tems: adaptive fuzzy approach. IEEE Trans. Fuzzy Syst. 4, 32–43 (1996)
88. Chen, B., Liu, X., Tong, S.: Adaptive fuzzy output tracking control of MIMO nonlinear
uncertain systems. IEEE Trans. Fuzzy Syst. 15(2), 287–300 (2007)
89. Chen, C.H., Lin, C.M., Chen, T.Y.: Intelligent adaptive control for MIMO uncertain nonlinear
systems. Expert Syst. Appl., Elsevier 35, 865–877 (2008)
90. Chen, C.S.: Robust self-organising neural-fuzzy control with uncertainty observer for MIMO
nonlinear systems. IEEE Trans. Fuzzy Syst. 19(4), 694–706 (2011)
91. Chen, C.S.: Dynamic structure adaptive neural fuzzy control for MIMO uncertain nonlinear
systems. Inf. Sci., Elsevier 179, 2676–2688 (2009)
92. Chen, C.S.: Robust self-organising neural-fuzzy control with uncertainty observer for MIMO
nonlinear systems. IEEE Trans. Fuzzy Syst. 19(4), 694–706 (2011)
93. Chen, X.B., Siljiak, D.D., Stankovic, S.S.: Decentralized H∞ design of automatic generation
control. In: IFAC 15th World Congress, Barcelona, Spain (2002)
706 References

94. Chen, G., Moiola, J.L., Wang, H.O.: Bifurcation control: theories, methods and applications.
Int. J. Bifurcat. Chaos 10(3), 511–548 (2000)
95. Chen, G., Zhang, L., Cai, X., Zhang, W., Yin, C.: Nonlinear control of the doubly fed
induction generator by input-output linearizing strategy. In: Hu, W. (ed.) Electronics and
Signal Processing, Lecture Notes in Electrical Engineering, vol. 97, pp. 601–608. Springer
(2011)
96. Chen, B., Liu, X., Tong, S.: Adaptive fuzzy approach to control unified chaotic systems.
Chaos, Solitons and Fractals, Elsevier 34, 1180–1187 (2007)
97. Chen, D., Zhao, W., Sprott, J.C., Ma, X.: Application of Takagi-Sugeno fuzzy model to
a class of chaotic synchronization and anti-synchronization. Nonlinear Dyn., Springer 73,
1495–1505 (2013)
98. Chengzhi, Y., Cong, W.: Persistency of excitation and performance of deterministic learning.
In: Proceedings of the 30th Chinese Control Conference, Yantai, China, July 2011
99. Cho, Y.W., Park, C.W., kim, J.H., Park, M.: Indirect model reference adaptive fuzzy control
of dynamic fuzzy-state space model. IET Proc. Control Theor. Appl. 148(4), 273–282 (2005)
100. Chiasson, J.: A new approach to dynamic feedback linearization control of an induction
motor. IEEE Trans. Autom. Control 43(3), 391–397 (1998)
101. Chiu, C.S.: Mixed feedforward/feedback adaptive fuzzy control for a class of MIMO non-
linear systems. IEEE Trans. Fuzzy Syst. 14(6), 716–727 (2006)
102. Chinniah, Y., Burton, R., Habibi, S.: EKF applications to actuators state estimation-based
control. Mechatronics, Elsevier 16, 643–653 (2006)
103. Chladny, R.R., Koch, C.R.: Flatness-based tracking of an electromechanical variable valve
timing actuator with disturbance observer feedforward compensation. IEEE Trans. Control
Syst. Technol. 16(4), 652–663 (2008)
104. Cho, Y.W., Park, C.W., kim, J.H., Park, M.: Indirect model reference adaptive fuzzy control
of dynamic fuzzy-state space model. IET Proc. Control Theor. Appl. 148(4), 273–282 (2005)
105. Choi, S.B., Hedrick, J.K.: An observer-based controller design method for improving air/fuel
characteristics of spark ignition engines. IEEE Trans. Control Syst. Technol. 6(3), 325–334
(1998)
106. Colin, G., Chamuillard, Y., Bloch, G., Corde, G.: Neural control of fast nonlinear systems:
application to a turbocharged SI engine with VCT. IEEE Trans. Neural Networks 18(4),
1101–1114 (2007)
107. Cortesao, R., Park, J., Khatib, O.: Real-time adaptive control for haptic tele-manipulation
with Kalman active observers. IEEE Trans. Robot. 22(5), 987–999 (2006)
108. Cortesao, R.: On Kalman active observers. J. Intell. Robot. Syst., Springer 48(2), 131–155
(2006)
109. Cot, L.D., Bes, C.: Study of the robustness of an enhanced CSK system by using the
extended Kalman filter. In: 2011 International Conference for Internet Technology and Se-
cured Transactions (ICITST), Abu Dhabi (2011)
110. Craven, P.J., Sutton, R., Burns, R.S.: Multivariable neurofuzzy control of an autonomous
underwater vehicle. Integr. Comput. Aided Eng., IOS Press 6(4), 275–288 (1999)
111. Crespo, L., Agrawal, S.: Differential flatness and cooperative tracking in the Lorenz System.
In: Proceedings of the American Control Conference Denver, Colorado, June 2003
112. Cui, Y., Ge, S.S.: Autonomous vehicle positioning with GPS in urban canyon environments.
IEEE Trans. Robot. Autom. 19(1), 15–15 (2003)
113. Dabo, M., Langlois, N., Chafouk, H.: Asymptotic tracking applied to the control of a tur-
bocharged diesel engine. In: UKACC International Conference on Control, University of
Manchester, 2008
114. Dacic, D.B., Subbotin, M.V., Kootovic, P.V.: Control effort reduction in tracking feedback
laws. IEEE Trans. Autom. Control 51(11), 1831–1837 (2006)
115. Damm, G., Marino, R., Lamnabhi-Lagarrigue, R.: Adaptive nonlinear output feedback for
stabilization and voltage regulation of power generators with unknown parameters. Int. J.
Robust Nonlinear Control, Wiley 14, 833–855 (2004)
References 707

116. Dasgupta, S., Mohan, S.N., Sahoo, S.K., Panda, S.K.: Lyapunov function-based current
controller to control active and reactive power flow from a renewable energy source to a
generalized three-phase microgrid system. IEEE Trans. Industr. Electron. 60(2), 799–813
(2013)
117. Dannehl, J., Fuchs, F.W.: Flatness-based control of an induction machine fed via voltage
source inverter—concept, control design and performance analysis. In: IECON 2006, IEEE
Conference on Industrial Electronics, pp. 5125–5130 (2006)
118. Delaleau, E., Louis, J.P., Ortega, R.: Modelling and control of induction motors. Int. J. Appl.
Math. Comput. Sci. 11(1), 105–129 (2001)
119. De Leon Moralers, J., Busawon, K., Acha-Duza, S.: A robust observer-based controller for
synchronous generators. Electr. Power Energy Syst., Elsevier, 23, 195–211 (2001)
120. Del Angel, A., Geurts, P., Ernst, D., Glavic, M., Wehenkel, L.: Estimation of rotor angles
of synchronous machines using artificial neural networks and local PMU-based quantities.
Neurocomputing, Elsevier 70, 2668–2678 (2007)
121. Delvecchio, D., Spelta, C., Perico, G., Savaresi, S.M.: Accelerometer-based estimation of
the elongation speed in a motorcycle suspension via Kalman-filter techniques. In: 49th IEEE
Conference on Decision and Control, Atlanta (2010)
122. De Luca, A., Siciliano, B.: Regulation of flexible arms under gravity. IEEE Trans. Robot.
Autom. 9, 463–467 (1993)
123. Delvecchio, D., Spelta, C., Savaresi, S.M.: Estimation of the Tire vertical deflection in a
motorcycle suspension via Kalman-Filtering techniques. In: 2011 IEEE International Con-
ference on Control Applications (CCA) Part of 2011 IEEE Multi-Conference on Systems
and Control Denver (2011)
124. Demetriou, M.A.: Design of consensus and adaptive consensus filters for distributed para-
meter systems. Automatica, Elsevier 46, 300–311 (2010)
125. Dendouga, A., Abdessemed, R., Bendaas, M.L., Chaiba, A.: Decoupled active and reactive
power control of a doubly-fed induction generator (DFIG). In: Proceedings of the 15th
Mediterranean Conference on Control and Automation, Athens, Greece, July 2007
126. DeSouza, G.N., Kak, A.C.: A subsumptive, hierarchical and distributed vision-based archi-
tecture for smart robotics. IEEE Trans. Syst. Man Cybern. B 34(5), 1988–2002 (2004)
127. Dimassi, H., Loria, A., Belghith, S.: A new secured transmission scheme based on chaotic
synchronisation via smooth adaptive unknown-input observers. Commun. Nonlinear Sci.
Numer. Simul., Elsevier 17, 3727–3739 (2012)
128. Do, K.D., Pan, J.: Global tracking control of underactuated ships with nonzero off-diagonal
terms in their system matrices. Automatica, Elsevier 41, 87–95 (2005)
129. Do, K.D.: Practical formation control of multiple underactuated ships with limited sensing
ranges. Robot. Auton. Syst., Elsevier 59(6), 457–471 (2011)
130. Do, T.D., Leu, V.Q., Choi, Y.S., Choi, H.H., Jung, J.W.: An adaptive voltage control strategy
of three-phase inverter for stand-alone distributed generation systems. IEEE Trans. Industr.
Electron. 60(12), 5660–5672 (2013)
131. Dong, L., Zhang, Y., Gao, Z.: A robust decentralized load frequency controlller for intercon-
nected power systems. ISA Trans., Elsevier 51, 410–419 (2012)
132. Doyle, J.C., Glover, K., Khargonekar, P.P., Francis, B.A.: State-space solutions to standard
H2 and H∞ control problems. IEEE Trans. Autom. Control 34, 831–847 (1989)
133. Dragonescu, A., Soane, A.M.: Multigrid solution of a distibuted optimal control problem
constrained by the Stokes equations. Appl. Math. Comput., Elsevier 219, 5622–5634 (2013)
134. Du, H., Zhang, N.: H∞ control of active vehicle suspensions with actuator time delay. J.
Sound Vib., Elsevier 301, 236–252 (2007)
135. Dubljevic, S.: Optimal boundary control of Kuramoto-Sivasjinsky equation. In: 2009 Amer-
ican Control Conference, St. Luis, MO, USA, June 2009
136. Dydek, Z.T., Annaswamy, A.M., Lavretsky, E.: Adaptive control of quadrotor UAVs: a design
trade study with flight evaluations. IEEE Trans. Control Syst. Technol. 21(4), 1400–1406
(2013)
708 References

137. Efimov, D.V., Fradkov, A.L.: Adaptive tuning to bifurcation for time-varying nonlinear sys-
tems. Automatica, Elsevier 42, 417–425 (2006)
138. Efimov, D., Zolghadri, A., Simon, P.: Improving fault detection abilities of extended Kalman
filters by covariance matrices adjustment. In: 2010 Conference on Control and Fault Tolerant
Systems, Nice, France, 6–8 Oct 2010
139. El Magri, A., Giri, F., Bensançon, G., El Fadili, A., Dagard, L., Chaoui, F.Z.: Sensorless
adaoptive output feedback control of wind energy systems with PMS generators. Control
Eng. Pract., Elsevier, 21, 530–543 (2013)
140. Elkington, K., Knazkins, V., Ghanhari, M.: On the stability of power systems containing
doubly-fed induction generator-based generation. Elect. Power Syst. Res., Elsevier 78, 1477–
1484 (2008)
141. Escobar, G., Chevreau, D., Ortega, R., Mendes, E.: An adaptive passivity-based controller
for a unity power factor rectifier. IEEE Trans. Control Syst. Technol. 9(4), 637–644 (2001)
142. Fallahi, K., Leung, H.: A chaos secure communication scheme based on multiplication
modulation. Commun. Nonlinear Sci. Numer. Simul., Elsevier 15, 368–383 (2010)
143. Fallahi, K., Raoufi, R., Khoshbin, H.: An application of Chen system for secure chaotic
communication based on Extended Kalman Filter and multi-shift cipher algorithm. Commun.
Nonlinear Sci. Numer. Simul., Elsevier 13, 763–781 (2008)
144. Fan, L., Wehbe, Y.: Extended Kalman filtering based real-time dynamic state and parameter
estimation using PMU data. Electr. Power Syst. Res., Elsevier, 103, 168–177 (2013)
145. Farinwata, S., Filev, D., Langari, R.: Fuzzy Control: Synthesis and Analysis. Wiley, New
York (2000)
146. Fei, J., Zhou, J.: Robust adaptive control of MEMS triaxial gyroscope using fuzzy compen-
sator. IEEE Trans. Syst. Man Cybern. Part B Cybern. 42(6), 1599–1607 (2012)
147. Feki, M., Robert, B., Gelle, G., Colas, M.: Secure digital communication using discrete-time
chaos synchronization. Chaos Solitons and Fractals 18, 881–890 (2003)
148. Fiengo, G., Grizzle, J.W., Cook, J.A., Karnik, A.Y.: Dual-UEGO active catalyst control for
emissions reduction: design and experimental validation. IEEE Trans. Control Syst. Technol.
13(5), 722–735 (2005)
149. Flardh, O., Ericsson, G., Klingborg, E., Martensson, J.: Optimal air-path control during load
transients on a spark ignited engine with variable geometry turbine and variable valve timing.
IEEE Trans. Control Syst. Technol. 22(1), 82–92 (2013)
150. Fliess, M., Mounier, H., Rouchon, P., Rudolph, J.: Systèmes lineaires sur les operateurs
de Mikusiński et commande d’ une poutre flexible. In: ESAIM Proceedinds Élasticité, vis-
coélasticité et contrôle optimal, huitièmes entretiens du Centre Jacques Cartier, Lyon (1996)
151. Fliess, M., Mounier, H., Rouchon, P., Rudolph, J.: A distributed parameter approach to the
control of a tubular reactor: a multi-variable case. In: Proceedings of 37th Conference on
Decision and Control, pp. 439–442, Tampa (1998)
152. Fliess, M., Mounier, H.: Tracking control and π -freeness of infinite dimensional linear sys-
tems. In: Picci, G., Gilliam, D.S. (eds.) DynAmical Systems, Control, Coding and Computer
Vision, vol. 258, pp. 41–68. Birkhaüser, Basel (1999)
153. Fliess, M., Lévine, J., Martin, P., Rouchon, P.: A Lie-Backlünd approach to equivalence and
flatness of nonlinear systems. IEEE Trans. Autom. Control 44(5), 922–937 (1999)
154. Fliess, M., Mounier, H.: An algebraic framework for infinite-dimensional linear systems. In:
Proceedings of International School on Automatic Control of Lille, Control of Distributed
Parameter Systems: Theory and Applications, Grenoble, France (2002)
155. Forchetti, D.G., Solsona, J.A., Garcia, G.O., Valla, M.I.: A control strategy for stand-alone
wound rotor induction machine. Electr. Power Syst. Res., Elsevier 77, 163–169 (2007)
156. Forchetti, D.G., Garcia, G.O., Valla, M.I.: Adaptive observer for sensorless control of stand-
alone doubly-fed induction generator. IEEE Trans. Industr. Electron. 56(10), 4174–4180
(2009)
157. Fouladirad, M., Nikiforov, I.: Optimal statistical fault detection with nuisance parameters.
Automatica, Elsevier 41, 1157–1171 (2005)
References 709

158. Fradkov, A.I., Andrievsky, B., Andrievsky, A.: Practically stable observer-based synchro-
nization of discrete-time chaotic systems over the limited-band communication channel. In:
3rd International Conference on Physics and Control PhysCon 2007, Potsdam, Germany,
Sept 2007
159. Franch, J., Agrawal, S.K., Sangwan, V.: Differential flatness of a class of n-DOF planar
manipulators driven by 1 or 2 actuators. IEEE Trans. Autom. Control 55(2), 548–554 (2010)
160. Fu, K.S., Gonzalez, R.C., Lee, G.S.G.: Robotics: Control, Sensing, Vision and Intelligence,
McGraw-Hill, New York (1987)
161. Ghahremani, E., Kamwa, I.: Dynamic state estimation in power system by applying the
extended Kalman filter with unknown inputs to phasor measurements. IEEE Trans. Power
Syst. 26(4), 2556–2566 (2011)
162. Galdi, V., Piccolo, A., Siano, P.: Designing an adaptive fuzzy controller for maximum wind
energy extraction. IEEE Trans. Energy Convers. 23(2), 559569 (2008)
163. Gan, Q., Harris, C.J.: Comparison of two measurement fusion methods for Kalman-filter-
based multisensor data fusion. IEEE Trans. Aerosp. Electron. Syst. 37(1), 273–280 (2001)
164. Gao, Y., Er, M.J.: Online adaptive fuzzy neural identification and control of a class of MIMO
nonlinear systems. IEEE Trans. Fuzzy Syst. 11(4), 462–477 (2003)
165. Gao, H., Sun, W., Shi, P.: Robust sampled-data H∞ control for vehicle active suspension
systems. IEEE Trans. Control Syst. Technol. 18(1), 238–245 (2010)
166. Ge, S.S., Hang, C.C., Zhang, T.: Nonlinear adaptive control using neural networks and its
application to CSTR systems. J. Process Control, Elsevier 9, 313–323 (1998)
167. Ge, S.S., Wang, C.: Direct adaptive NN control of a class of nonlinear systems. IEEE Trans.
Neural Networks 13(1), 214–221 (2002)
168. Ge, S.S., Hang, C.C., Zhang, T.: Adaptive neural network control of nonlinear systems by
state and output feedback. IEEE Trans. Syst. Man Cybern. Part B Cybern. 29, 818–828
(1999)
169. Ge, S.S., Lee, T.H., Zhu, G.: Energy-based robust controller design for multi-link flexible
robots. Mechatronics, Elsevier, 6(7), 779–798 (1996)
170. Georges, D., De Wit, C., Ramirez, J.: Nonlinear H2 and H∞ optimal controllers for current-
fed induction motors. IEEE Trans. Autom. Control 44(7), 1430–1435 (1999)
171. Gensior, A., Nguyen, T.M.P., Rudolph, J., Guldner, H.: Flatness-based loss optimization
and control of a doubly-fed induction generator system. IEEE Trans. Control Syst. Technol.
19(6), 1457–1466 (2011)
172. Gensior, A., Rudolph, J., Güldner, H.: Flatness-based control of three-phase boost rectifiers.
In: 11th European Conference on Power Electronics and Applications, EPE 2005, Dresden,
Germany, Sept 2005
173. Gensior, A., Sira-Ramirez, H., Rudolph, J., Guldner, H.: On some nonlinear current con-
trollers for three-phase boost rectifiers. IEEE Trans. Industr. Electron. 56(2), 360–370 (2009)
174. Gerasimov, D.N., Javaherian, H., Nikiforov, V.O.: Data driven inverse-model control of SI
engines. In: 2011 American Control Conference, San Francisco, CA, USA, June–July 2011
175. Gerdes, M., Greif, G., Peich, H.J.: Numerical optimal control of the wave equation: optimal
boundary control of a string to rest in finite time. In: Proceedings of the 5th Mathmod
Conference, Vienna, Austria, Feb 2006
176. Ghommam, J., Mnif, F.: Coordinated path-following control for a group of underactuated
surface vessels. IEEE Trans. Industr. Electron. 56(10), 3951–3963 (2009)
177. Gibbs, B.P.: Advanced Kalman Filtering, Least Squares and Modelling: A Practical Hand-
book. Wiley, New Jersey (2011)
178. Golapalakhrishnan, A., Kaisare, N., Narasimhan, S.: Incorporating delayed and infrequent
measurements in extended Kalman filter-based nonlinear state estimation. J. Process Control,
Elsevier 21, 119–129 (2011)
179. Gorinevsky, D.: On the persistency of excitation in radial basis function network identification
of nonlinear systems. IEEE Trans. Neural Networks 6(5), 1237–1244 (1995)
180. Granle, M., Zhu, G., Saydy, L.: Sliding-mode tracking control of an electrostatic parllel-plate
MEMs. In: IEEE/ASME International Conference on Advanced Intelligent Mechatronics,
Montréal, Canada, July 2010
710 References

181. Guardiola, C., Pla, B., Blanco-Rodriguez, D., Erikson, L.: A computationally efficient
Kalman Filter-based estimator for updating look-up tables applied to NOx estimation in
diesel engines. Control Eng. Pract., Elsevier 21(11), 1455–1468 (2013)
182. Guerrero, J.M., Chandorkar, M., Lee, T.L., Loh, P.C.: Advanced control architectures for
intelligent microgrids-part I: decentralized and hierarchical control. IEEE Trans. Industr.
Electron. 60(4), 1260–1262 (2013)
183. Gupta, A. Malley, M.K.O.: Disturbance-observer-based force estimation for haptic feedback.
ASME J. Dyn. Syst. Meas. Control 133(1), Article no. 014505 (2011)
184. Guo, S., Xie, W., Ma, B.: Robust position stabilization of underactuated hovercrafts with
modelling parameter uncertainties. In: Proceedings of the 32nd Chinese Control Conference,
Xian China, July 2013
185. Guo, B.Z., Xu, C.Z., Hammouri, H.: Output feedback stabilization of a one-dimensional wave
equation with an arbitrary time-delay in boundary observation. ESAIM Control Optim. Calc.
Var. 18, 22–25 (2012)
186. Guo, L., Billings, S.A.: State-space reconstruction and spatio-temporal prediction of lattice
dynamical systems. IEEE Trans. Autom. Control 52(4), 622–632 (2007)
187. Gugat, M., Keiner, A., Leugering, G.: Optimal distributed control of the wave equation
subject to state constraints. ZAMM J. Appl. Math. Mech. 89(6), 420–444 (2009)
188. Guzzella, L., Simons, M., Geering, H.P.: Feedback-linearizing air-fuel ratio control. Control
Eng. Pract., Elsevier 5(8), 1101–1105 (1997)
189. Hagenmeyer, V., Delaleau, E.: Robustness analysis of exact feedforward linearization based
on differential flatness. Automatica, Elsevier 39, 1941–1946 (2003)
190. Hagenmeyer, V., Delaleau, E.: Robustness analysis with respect to exogenous perturbations
for flatness-based exact feedforward linearization. IEEE Trans. Autom. Control 55(3), 727–
731 (2010)
191. Han, T.T., Ge, S.S.: Cooperative control design for circular flocking of underactuated hov-
ercrafts, 2011 IEEE Conference on Decision and Control and European Control Conference
(CDC-ECC), Orlando Florida, Dec 2011
192. Harris, C., Hong, X., Gan, Q.: Adaptive Modelling, Estimation and Fusion from Data.
Springer, Berlin (2002)
193. Harrison, R.F.: Optimal LQG regulation of multi-axled vehicle suspension dynamics. J.
Sound Vib., Elsevier 174, 239–247 (1994)
194. Haschke, R., Steil, J.J.: Input-space bifurcation manifolds of recurrent neural networks.
Neurocomputing, Elsevier 64, 25–38 (2005)
195. He, B., Shen, T., Kako, J., Ouyang, M.: Input observer-based individual cylinder air-fuel
ratio control: modelling, design and validation. IEEE Trans. Control Syst. Technol. 16(5),
1057–1065 (2008)
196. He, Y., McPhee, J.: Multidisciplinary design optimization of mechatronic vehicles with active
suspensions. J. Sound Vib., Elsevier 283, 217–241 (2005)
197. He, G., Gang, Z.: Finite-time stabilization of a comb-drive electrostatic microactuator.
IEEE/ASME Trans. Mechatron., Elsevier 17(1), 107–115 (2012)
198. Hidayat, Z., Babuska, R., de Schutter, B., Nunez, A.: Decentralized Kalman Filter com-
parison for distributed parameter systems: a case study for a 1D heat conduction process.
In: Proceedings of the 16th IEEE International Conference on Emerging Technologies and
Factory Automatio, ETFA 2011, Toulouse, France, Sept 2011
199. Hilairet, M., Augerb, E., Berthelot, F.: Speed and rotor flux estimation of induction machines
using a two-stage extended Kalman filter. Automatica, Elsevier 45, 1819–1827 (2009)
200. Holtz, J.: Sensorless control of induction motor drives. Proc. IEEE 90(8), 1359–1394 (2002)
201. Hong, K.S.: An open-loop control for underactuated manipulators using oscillatory inputs:
steering capability of an unactuated joint. IEEE Trans. Control Syst. Technol. 10(3), 469–479
(2002)
202. Horng, J.H.: Neural adaptive tracking control of a DC motor. Inf. Sci., Elsevier 118, 1–13
(1999)
References 711

203. Hosseini, M., Zhu, G., Peter, Y.A.: A new formulation of fringing capacitance and its appli-
cation to the control of parallel-plate electrostatic micro-actuators. Analog Integr. Circ. Sig.
Process., Springer 53, 119–128 (2007)
204. Hou, J., Jankowski, L., Qu, J.: An online substructure identification method for local struc-
tural health monitoring. Smart Mater. Struct., Inst. Phys. Publishing 22, 95–117 (2013)
205. Hovakimyan, N., Nardi, F., Calise, A.J.: A novel error observer-based adaptive output feed-
back approach for control of uncertain systems. IEEE Trans. Autom. Control 47(8), 1310–
1314 (2002)
206. Hrovat, D.: Optimal active suspension structures for quarter-car vehicle models. Automatica,
Elsevier 26(5), 845–860 (1998)
207. Hsiao, W.Y., Chiang, H.H., Lee, T.T.: Sliding-mode-based filtered feedback control design
for active suspension system. In: IEEE SMC 2011 International Conference on Systems,
Man and Cybernetics. Anchorage, Alaska (2011)
208. Hu, Z., Feng, J.: Blind channel equalization algorithm based on dual Unscented Kalman
Filter for chaotic multi-input multi-output communication systems. Trans. Tianjin Univ.,
Springer 18, 33–37 (2012)
209. Hua, J.S., Yin, D., Hori, Y.: Fault-tolerant traction control of electric vehicles. Control Eng.
Pract., Elsevier 19, 204–213 (2011)
210. Huang, Y.S., Zhou, D.Q., Xiao, S.P., Ling, D.: Coordinated deventralized hybrid adaptive
output feedback fuzzy control for a class of large-scale nonlinear systems with strong inter-
connections. IET Control Theor. Appl. 3(9), 1261–1274 (2009)
211. Huang, Z., Schneider, K., Nieplocha, J.: Feasibility studies of applying Kalman Filter tech-
niques to power system dynamic state estimation. In: IPEC 2007 Power Engineering Inter-
national Conference, pp. 376–382
212. Huerta, F., Pizzaro, D., Cobreces, S., Rodriguez, F., Giron, C., Rodriguez, A.: LQG Servo
controller for the current control of LCL grid-connected voltage-source converters. IEEE
Trans. Industr. Electron. 59(11), 4272–4284 (2012)
213. Hugues-Salas, O., Shove, K.A.: An extended Kalman filtering approach to nonlinear time-
delay systems: application to chaotic secure communications. IEEE Trans. Circ. Syst. I:
Regul. Pap. 57, 2520–2530 (2010)
214. Houari, A., Renaudineau, H., Martin, J.P., Pierfederici, S., Meibody-Tabar, F.: Flatness-based
control of three-phase inverter with output LC filter. IEEE Trans. Industr. Electron. 59(7),
2890–2897 (2012)
215. Hwang, C.L., Shih, Y.Y.: A distributed active-vision network-space approach for the navi-
gation of a car-like wheeled robot. IEEE Trans. Industr. Electron. 56(3), 846–855 (2002)
216. Isidori, A.: Nonlinear Control Systems, 3rd edn. Springer, New York (1995)
217. Jankovic, M., Magner, S.: Control of engines with fully variable valvetrain. In: American
Control Conference, Portland, June 2005
218. Jaward, M.H., Kadirkamanathan, V.: Interacting multiple models for single-user channel
estimation and equalization. In: ICASSP 2001, IEEE International Conference on Acoustics,
Speech, and Signal Processing, Salt Lake City, Utah (2001)
219. Jia, Z., Balasuriyaa, S., Challab, S.: Sensor fusion-based visual target tracking for
autonomous vehicles with the out-of-sequence measurements solution. Robot. Auton. Syst.,
Elsevier 56, 157–176 (2008)
220. Jiang, L., Wu, Q.H., Zhang, C., Zhou, X.X.: Observer-based nonlinear control of synchronous
generators with pertubation estimation. Electr. Power Energy Syst., Elsevier 23, 359–367
(2001)
221. Jiang, L., Wu, Q.H., Wen, J.Y.: Decentralized nonlinear adaptive control for multimachine
power systems via high-gain perturbation observer. IEEE Trans. Circ. Syst. 51(10), 2052–
2059 (2004)
222. Jiang, G., Guo, S., Zhang, J., Huang, W.: Development of electronic governor for diesel
engine based on embedded RTOS. In: IEEE ICVES 2006, IEEE International Conference
on Vehicular Electronics and Safety, Beijing, China (2006)
712 References

223. Johnson, B.B., Dhople, S.V., Hamadeh, A.O., Krein, P.T.: Synchronization of nonlinear
oscillators in an LTI electrical power network. IEEE Trans. Circuit. Syst. I 61(3), 834–844
(2014)
224. Julier, S., Uhlmann, J., Durrant-Whyte, H.F.: A new method for the nonlinear transformations
of means and covariances in filters and estimators. IEEE Trans. Autom. Control 45(3), 477–
482 (2000)
225. Julier, S.J., Uhlmann, H.K.: Unscented filtering and nonlinear estimation. Proc. IEEE 92,
401–422 (2004)
226. Kahrobaeian, A., Abdel-Rady, Y., Mohamed, I.: Analysis and mitigation of low-frequency
instabilities in autonomous medium-voltage converter-based microgrids with dynamic loads.
IEEE Trans. Industr. Electron. 61(4), 1643–1657 (2014)
227. Kahveci, N.E., Jankovic, M.J.: Adaptive controller with delay compensation for air-fuel ratio
regulation in SI engines. In: 2010 American Control Conference, Baltimore, MD, USA,
June-July 2010
228. Kamala, E., Aitouche, A.: Robust fault tolerant control of DFIG wind energy systems with
unknown inputs. Renewable Energy, Elsevier 56, 2–15 (2013)
229. Kamen, E.W., Su, J.K.: Introduction to Optimal Estimation. Springer, New York (1999)
230. Kalsi, K., Lian, J., Zak, S.H.: Decentralized control of multi-machine power systems. In:
American Control Conference, St. Louis, MO, USA (2009)
231. Kandepu, R., Foss, B., Imsland, M.: Applying the unscented Kalman filter for nonlinear state
estimation. J. Process Control, Elsevier 18, 753–768 (2008)
232. Kanoh, H., Tzafestas, S.G., Lee, H.G., Kalat, J.: Modelling and control of flexible robot
arms. In: Proceedings of 25th Conference on Decision and Control, pp. 1866–1870, Athens,
Greece (1986)
233. Karami, F., Poshtan, J., Poshtan, M.: Model-based fault detection in induction Motors. In:
2010 IEEE International Conference on Control Applications, CCA 2010, Yokohama, Japan,
Sept 2010
234. Katkhuda, H., Hadlar, A.: A novel health assessment technique with minimum information.
J. Struct. Control Health Monit., Wiley 15, 821–838 (2008)
235. Kenné, G., Gome, R., Nkwawo, H., Lamnabhi-Lagarrique, F., Arzandé, A., Vannier, J.C.:
An improved direct feedback linearization techique for transient stability enhancement and
voltage regulation of power generators. Electr. Power Energy Syst., Elsevier 32, 809–816
(2010)
236. Kennel, R.M.: Why do incremental encoders do a reasonably good job in electrical drives
with digital control? In: Industry Applications Conference, IEEE 41st IAS Annual Meeting.
Conference Record (2006)
237. Kerkeni, H., Lauber, J., Lendek, Z., Guerra, T.M.: Individual cylinder air/fuel ratio observer
on IC engine using Takagi-Sugeno’s fuzzy model. In: IEEE Vehicle Power and Propulsion
Conference (VPPC), Harbin, China (2008)
238. Khalil, H.K.: Nonlinear Systems, 2nd edn. Prentice Hall, Upper Saddle River (1996)
239. Khanesar, M.A., Teshnehlab, M., Kaynak, O.: Observer-based indirect model reference fuzzy
control system with application to control of chaotic systems. J. Franklin Inst., Elsevier 350,
419–436 (2013)
240. Khanesar, M.A., Teshnehlab, M., Kaynak, O.: Control and synchronization of chaotic sys-
tems using a novel indirect model reference fuzzy controller. Soft Comput., Springer 16(7),
1253–1265 (2012)
241. Kharel, R., Busawon, K., Ghassemlooy, Z.: Observer based secure communication using
indirect coupled synchronization. In: 8th IEEE, IET International Symposium on Commu-
nication Systems, Networks and Digital Signal Processing (CSNDSP’12). Poznan, Poland
(2012)
242. Khorrami, F., Jain, S., Tzes, A.: Experimental results on adaptive nonlinear control and input
preshaping for multi-link flexible manipulators. Automatica, Elsevier 31(1), 83–97 (1995)
243. Kiani, A., Fallahi, K., Pariz, N., Leung, H.: A chaotic secure communication scheme using
fractional chaotic systems based on an extended fractional Kalman filter. Commun. Nonlinear
Sci. Numer. Simul., Elsevier 14, 863–879 (2009)
References 713

244. Kim, Y., Park, C., Joo, J., Jeong, S.: Extended Kalman filter for wireless LAN based indoor
positioning. Decis. Support Syst., Elsevier 45, 960–971 (2008)
245. Kim, K., Jeung, Y., Lee, D., Kim, H.: LVRT scheme of PMSG wind power systems based
on feedback linearization. IEEE Trans. Power Electron. 27(5), 2376–2384 (2012)
246. Kim, D.E., Lee, D.C.: Feedback linearization control of three-phase UPS inverter systems.
IEEE Trans. Industr. Electron. 57(3), 963–968 (2010)
247. Kim, J.H., Hyun, C.H., Kim, E., Park, M.: Adaptive synchronization of chaotic systems
based on T-S fuzzy model. IEEE Trans. Fuzzy Syst. 15(3), 359–369 (2007)
248. Koch, G., Kloiber, T., Pellegrini, E., Lohmann, B.: A nonlinear estimator concept for active
vehicle suspension control. In: American Control Conference Marriott Waterfront, Balti-
more, MD, USA, (2010)
249. Koch, G., Kloiber, T., Lohmann, B.: Nonlinear and filter based estimation for vehicle sus-
pension control. In: 49th IEEE Conference on Decision and Control, Atlanta, GA, USA
(2010)
250. Koch, C., Lynch, A., Chung, S.: Flatness-based automotive solenoid valve control. In: Pro-
ceedings of the 6th IFAC Symposium Nonlinear Controll Systems (NOLCOS), pp. 1091–
1096 (2004)
251. Kotman, P., Bitzer, M., Kugi, A.: Flatness-based feedforward control of a two-stage tur-
bocharged diesel air system with EGR. In: 2010 IEEE International Conference on Control
Applications, Yokohama, Japan (2010)
252. Kröner, A.: Adaptive finite element methods for optimal control second order hyberbolic
equations. Comput. Methods Appl. Math., De Gruyter 11(2), 214–240 (2011)
253. Krstic, M.: Adaptive control of an anti-stable wave PDE. Dyn Continuous, Discrete Impulsive
Syst. Ser A: Math Anal. 17, 853–882 (2010)
254. Krstic, M., Guo, B.Z., Balogh, A., Smyshlyaev, A.: Output-feedback stabilization o an un-
stable wave equation. Automatica, Elsevier 44, 63–74 (2008)
255. Kröner, A.: Adaptive finite element methods for optimal control second order hyberbolic
equations. Comput. Methods Appl. Math. 11(2), 214–240 (2011)
256. Kumar, S., Prakash, J., Kanagasabapathy, P.: A critical evaluation and experimental veri-
fication of extended Kalman filter, unscented Kalman filter and neural state filter for state
estimation of three phase induction motor. Appl. Soft Comput. J., Elsevier 11(3), 3199–3208
(2011)
257. Kuo, H.C., Wu, L.J.: Prediction of heat-affected zone using Grey theory. J. Mater. Process.
Technol., Elsevier 180, 151–168 (2002)
258. Kurikawa, T., Kaneko, K.: Learning to memorize input-output mapping as bifurcation in
neural dynamics: relevance of multiple time-scales for synaptic changes. Neural Comput.
Appl., Springer 31, 725–734 (2012)
259. Kurylowicz, A., Jaworska, I., Tzafestas, S.G.: Robust stabilizing control: an overview. In:
Tzafestas, S.G. (ed.) Applied Control—Current Trends and Modern Methodologies, pp.
289–324. Marcel Dekker (1993)
260. Kwon, S., Chung, W.K.: Combined synthesis of state estimator and perturbation observer.
ASME J. Dyn. Syst. Measur. Control 125, 19–26 (2003)
261. Lai, X.Z., Pan, C.Z., Wu, M., Yang, S.X.: Unified control of n-link underactuated manipulator
with single passive joint: a reduced order approach. Mech. Mach. Theor., Elsevier 56, 170–
185 (2012)
262. Lai, X.Z., She, J.-H., Yang, S.X., Wu, M.: Comprehensive unified control strategy for under-
actuated two-link manipulators. IEEE Trans. Syst., Man Cybern. B Cybern. 39(2), 389–398
(2009)
263. Laroche, B., Martin, Ph, Petit, N.: Commande par platitude: Équations différentielles ordi-
naires et aux derivées partielles. École Nationale Supérieure des Techniques Avancées, Paris
(2007)
264. Lee, D.J.: Nonlinear estimation and multiple sensor fusion using unscented information
filtering. IEEE Sig. Process. Lett. 15, 861–864 (2008)
714 References

265. Lee, D.J.: Unscented information filtering for distributed estimation and multiple sensor
fusion. In: AIAA Guidance, Navigation and Control Conference and Exhibit, Hawai, Aug
2008
266. Lee, H.T., Chen, C.T., Wu, J.L.: Numerical and experimental investigation into effect of
temperature field on sensitization of alloy 690 butt welds fabricated by gas tungsten arc
welding and laser beam welding. J. Mater. Process. Technol., Elsevier 210, 1636–1644
(2010)
267. Lee, K.M., Woodruff, G.W., Wei, Z., Zhou, Z.: Modeling by numerical reduction of modes
for multivariable control of an optical-fiber draw process. IEEE Trans. Autom. Sci. Eng.
3(1), 119–130 (2006)
268. Lee, S.C., Ahn, H.S.: Sensorless torque estimation using adaptive Kalman filter and dis-
turbance estimator. In: 2010 IEEE/ASME International Conference on Mechatronic and
Embedded Systems and Applications, Qingdao, China, July 2010
269. Lee, S., Kim, W.J.: Active suspension control with direct-drive tubular linear brushless
permanent-magnet motor. IEEE Trans. Control Syst. Technol. 18(4), 859–870 (2010)
270. Lee, S., Jeon, M., Shin, V.: Distributed estimation fusion with application to a multisensory
vehicle suspension system with time delays. IEEE Trans. Industr. Electron. 59(11), 4475–
4482 (2010)
271. Lee, T.: Robust adaptive attitude tracking on SO(3) with an application to a quadrotor UAV.
IEEE Trans. Control Syst. Technol. 21(5), 1924–1930 (2013)
272. Lee, K., Jahns, T.J., Lipo, T.A., Blasko, V.: New control method including state observer of
voltage unbalance for grid voltage-source converters. IEEE Trans. Industr. Electron. 57(6),
2054–2065 (2010)
273. Lei, Y., Jiang, Y.Q.: A two-stage Kalman estimation approach for the identification of non-
linear structural parameters. Procedia Eng., Elsevier 14, 3088–3094 (2011)
274. Lei, Y., Jiang, Y., Xu, Z.: Structural damage detection with limited input and output mea-
surement signals. Mech. Syst. Sig. Process., Elsevier 28, 229–243 (2012)
275. Lei, Y., Wu, Y., Li, T.: Identification of nonlinear structural parameters under limited input
and output measurements. Int. J. Nonlinear Mech., Elsevier 47, 1141–1146 (2012)
276. Leland, R.P.: Adaptive control a MEMS gyroscope using Lyapunov methods. IEEE Trans.
Control Syst. Technol. 14(2), 278–283 (2006)
277. Leon, A.E., Solsona, J.A., Valla, M.I.: Comparison among nonlinear excitation control strate-
gies used for damping power system oscillations. Energy Convers. Manag., Elsevier 53,
55–67 (2012)
278. Leon, A.E., Mauricio, J.M., Solsona, J.A.: Multi-machine power system stability improve-
ment. Using an observer-based nonlinear controller. Electr. Power Syst. Res., Elsevier 89,
202–214 (2012)
279. Leon, A.E., Solsona, J.S., Busada, C., Chiacchiarini, H., Valla, M.I.: High-performance
control of a three-phase voltage-source converter including feedforward compensation of
the estimated load current. Energy Convers. Manag., Elsevier 50, 2000–2008 (2009)
280. Leonhard, W.: Control of Electrical Drives. Springer, London (1985)
281. Leroy, T., Chauvin, J., Petit, N.: Airpath control o a SI engine with variable valve timing
actuators. In: 2008 American Control Conference. Seattle USA, Washington, June 2008
282. Leu, Y.G., Lee, T.T., Wang, W.Y.: Observer-based adaptive fuzzy-neural control for unknown
nonlinear dynamical systems. IEEE Trans. Syst., Man Cybern. B Cybern. 29, 583–591 (1999)
283. Leung, H., Zhu, Z.: Performance evaluation of EKF-based chaotic synchronization. IEEE
Trans. Circ. Syst. I: Fundam. Theor. Appl. 49, 1118–1125 (2001)
284. Lévine, J.: Analysis and Control of Nonlinear Systems: A Flatness-Based Approach.
Springer, London (2009)
285. Lévine, J., Nguyen, D.V.: Flat output characterization for linear systems using polynomial
matrices. Syst. Control Lett., Elsevier 48, 69–75 (2003)
286. Lévine, J.: On necessary and sufficient conditions for differential flatness. Appl. Algebra
Eng. Commun. Comput., Springer 22(1), 47–90 (2011)
References 715

287. Liao, X., Wong, K.W., Wu, Z.: Bifurcation analysis of a two-neuron system with distributed
delays. Physica D, Elsevier 149, 123–141 (2001)
288. Li, Y., Jiang, Y.: Real-time control of the robot manipulators by neural networks. Integr.
Comput. Aided Eng., IOP Press 2(3), 241–248 (1995)
289. Li, H.X., Tong, S.: A hybrid adaptive fuzzy control for a class of nonlinear MIMO systems.
IEEE Trans. Fuzzy Syst. 11(1), 24–35 (2003)
290. Li, P., Kadirkamanathan, V.: Particle filtering based likelihood ratio approach to fault di-
agnosis in nonlinear stochastic systems. IEEE Trans. Syst. Man Cybern. C 31, 337–343
(2001)
291. Li, T.S., Tong, S.C., Feng, G.: A novel robust adaptive-fuzzy-tracking control for a class of
nonlinear multi-input multi-output systems. IEEE Trans. Fuzzy Syst. 18(1), 150–160 (2010)
292. Li, W., Lin, P.X.: Robust adaptive tracking control of uncertain electrostatic microactuators
with H-infinity performance. Mechatronics, Elsevier 19, 591–597 (2009)
293. Li, S., Kolmanovsky, I.V., Ulsoy, A.G.: Direct optimal design for component swapping
modularity in control systems. IEEE/ASME Trans. Mechatron. 18(1), 297–306 (2013)
294. Lian, K.Y., Chiang, T.S., Chiu, C.S., Liu, P.: Synthesis of fuzzy model-based designs to
synchronization and secure communications for chaotic systems. IEEE Trans. Syst., Man
Cybern. Part B: Cybern. 31, 66–83 (2001)
295. Lim, J.S., Park, C., Han, J., Lee, J.L.: Robust tracking of a three-phase DC-AC inverter for
UPS applications. IEEE Trans. Industr. Electron. 61(8), 4142–4151 (2014)
296. Lin, F.J., Wai, R.J., Lin, C.H., Liu, D.C.: Decoupled stator-flux-oriented induction motor
drive with fuzzy neural network uncertainty observer. IEEE Trans. Industr. Electron. 47(2),
356–367 (2000)
297. Lin, Y.J., Wang, W.: Adaptive fuzzy control for a class of uncertain non-affine nonlinear
systems. Inf. Sci., Elsevier 177, 3901–3917 (2007)
298. Lin, J.M., Lin, M.C., Wang, H.P.: LEQG/LTR controller design with Extended Kalman
Filter for sensorless brushless DC driver. Comp. Methods Appl. Mech. Eng., Elsevier 190,
5481–5494 (2001)
299. Lin, Y.J., Wang, W.: Adaptive fuzzy control for a class of uncertain non-affine nonlinear
systems. Inf. Sci., Elsevier 177, 3901–3917 (2007)
300. Lin, W.M., Hong, C.M., Cheng, F.S.: On-line designed hybrid controller with adaptive
observer for variable-speed wind generation system. Energy, Elsevier 35, 3022–3030 (2010)
301. Lin, C.M., Ting, A.B., Hsu, C.F., Chung, C.M.: Adaptive control for MIMO uncertain non-
linear systems using recurrent wavelet neural network. Int. J. Neural Syst., World Scientific
21(6), 443–457 (2011)
302. Lin, Y.J., Wang, W.: Adaptive fuzzy control for a class of uncertain non-affine nonlinear
systems. Inf. Sci., Elsevier 177, 3901–3917 (2007)
303. Lin, J.H., Lin, J.H.: Decoding chaotic secure communication system with extended Kalman
filter based observers. In: IEEE ISIE 2013 International Symposium on Industrial Electronics,
Taipei, Taiwan (2013)
304. Lin, W.M., Hong, C.M., Cheng, F.S.: On-line designed hybrid controller with adaptive
observer for variable-speed wind generation system. Energy, Elsevier 35, 3022–3030 (2010)
305. Lin, S.S., Horng, S.C.: A more general parallel dual-type method and application to state
estimation. Electr. Power Energy Syst., Elsevier 33, 799–804 (2011)
306. Lions, J.L.: Pointwise control for distributed systems. In: Banks, H.T. (ed.) Control and
Estimation in Distributed Parameter Systems. SIAM Frontiers in Applied Mathematics.
SIAM, Philadelphia (1992)
307. Liu, Y., Zhu, Z.Q., Howe, D.: Instantaneous torque estimation in sensorless direct-torque-
controlled brushless DC motors. IEEE Trans. Ind. Appl. 42(5), 1275–1283 (2006)
308. Liu, M., Zhang, S.: An LMI approach to design H∞ controllers for discrete-time nonlinear
systems based on unified models. Int. J. Neural Syst., World Sci. 18(5), 443–452 (2008)
309. Liu, J., Laghrouche, S., Wack, M.: Differential flatness-based observer design for a PEM fuel
cell using adaptive-gain sliding mode differentiators. In: 2013 European Control Conference
(ECC), Zurich, Switzerland, July 2013
716 References

310. Liu, W., Gu, W., Sheng, W., Meng, X., Wu, Z., Chen, W.: Decentralized multi-agent system-
based cooperative frequency control for autonomous microgrids with communication con-
straints. IEEE Trans. Sustain. Energy 5(2), 446–456 (2014)
311. Liu, D., Javaherian, H., Kovalenko, O., Huang, T.: Adaptive critic learning techniques for
engine torque and airfuel ratio control. IEEE Trans. Syst., Man Cybern. Part B: Cybern.
38(4), 988–993 (2008)
312. Liu, Y., Guo, C., Zhou, R.: Robust feedback stabilization control of an underactuated sur-
face vessel. In: IEEE WRI 2009 World Congress on Computer Science and Information
Engineering, Los Angeles, California, March–April (2009)
313. Locatelli, M., Alfieri, E., Onder, C.H., Geering, H.P.: Identification of the relevant parameters
of the wall- wetting system by extended Kalman filter. Control Eng. Pract., Elsevier 14(3),
235–241 (2006)
314. Loria, A.: Control of the 4th order hyper-chaotic system with one input. Commun. Nonlinear
Sci. Numer. Simul., Elsevier 15(6), 1621–1630 (2010)
315. Low, K.S., Cao, R.: Model predictive control of parallel-connected inverters for uninterrupt-
ible power supplies. IEEE Trans. Industr. Electron. 55(8), 2884–2893 (2008)
316. Lublin, L., Athans, M.: An experimental comparison of and designs for interferometer test-
bed. In: Francis, B., Tannenbaum, A. (eds.) Lectures Notes in Control and Information
Sciences: Feedback Control, Nonlinear Systems and Complexity, pp. 150–172. Springer,
London (1995)
317. Lu, W.C., Duan, L., Hsiao, F.B., Mora-Camino, F.: Differential flatness applied to vehicle
trajectory tracking. In: Proceedings of the 27th Chinese Control Conference, 16–18 July,
Kunming, Yunnan, China (2008)
318. Lu, Q., Mei, S., Hu, W., Wu, F.F., Ni, Y., Shen, T.: Nonlinear decentralized disturbance
attenuation excitation control via new recursive design for multi-machine power systems.
IEEE Trans. Power Syst. 16(4), 729–736 (2001)
319. Machowski, J., Robak, S., Bialek, J.W., Bumby, J.R., Abi-Samra, N.: Decentralized stability-
enhancing control of synchronous generator. IEEE Trans. Power Syst. 15(4), 1336–1344
(2000)
320. Maidi, A., Corriou, J.P.: Distributed control of nonlinear diffusion systems by input-output
linearization. Int. J. Robust Nonlinear Control, Wiley 26, 389–405 (2014)
321. Malinkowski, M., Kazimierkowski, M.P., Hansen, S., Blaabjerg, F., Marques, G.D.: Virtual-
flux-based direct power control of three-phase PWM rectifiers. IEEE Trans. Ind. Appl. 37(4),
1019–1027 (2001)
322. Malinkowski, M., Jasinksi, M., Kazmeirkowski, M.P.: Simple direct power control of three-
phase PWM rectifier using space-vector modulation (DPC-SVM). IEEE Trans. Ind. Electron.
51(2), 447–454 (2004)
323. Mammar, S., Glaser, S., Netto, M.: Vehicle lateral dynamics estimation using unknown input
proportional-integral observers. In: American Control Conference (2006)
324. Mathuswamy, J., Okandan, M., Jain, T., Gilletti, A.: Electrostatic microactuators for precise
positioning of neural microelectrodes. IEEE Trans. Biomed. Eng. 52(10), 1742–1755 (2005)
325. Mahindrakar, A.D., Rao, S., Banavar, R.N.: Point-to-point control of a 2R planar horizontal
underactuated manipulator. Mech. Mach. Theor., Elsevier 41, 838–844 (2006)
326. Mahler, R.P.S.: Statistical Multisource-multitarget Information Fusion. Artech House Inc.,
Norwood (2007)
327. Mahmoud, M.A., Rota, H.K., Hussain, M.J.: Full-order nonlinear observer-based excitation
controller design for interconnected power systems via exact linearization approach. Electr.
Power Energ. Syst., Elsevier 41, 54–62 (2012)
328. Mahmud, M.A., Pota, H.R., Hossain, M.J.: Nonlinear controller design for single-phase
Grid-Connected photovoltaic systems using partial feedback linearization. In: Australian
Control Conference, Sydney, Australia, Nov 2012
329. Makarenko, A., Durrany-Whyte, H.: Decentralized Bayesian algorithms for active sensor
networks. Inf. Fusion, Elsevier 7, 418–433 (2006)
References 717

330. Malis, E., Chaumette, F., Boudet, S.: Multi-cameras visual servoing. In: IEEE International
Conference on Robotics and Automation, ICRA 2000 International Conference on Robotics
and Automation, pp. 3183–3188 (2000)
331. Mammar, S., Glaser, S., Netto, M.: Vehicle lateral dynamics estimation using unknown input
proportional-integral observers. In: American Control Conference (2006)
332. Manyika, J., Durrant-Whyte, H.: Data Fusion and Sensor Management: A Decentralized
Information Theoretic Approach. Prentice Hall, Englewood Cliffs (1994)
333. Maroteaux, F., Saad, C.: Diesel engine combustion modelling for hardware in the loop
applications: effects of ignition delay time model. Energy, Elsevier 57(1), 641–652 (2013)
334. Marino, R.: Adaptive observers for single output nonlinear systems. IEEE Trans.
Autom.Control 35(9), 1054–1058 (1990)
335. Marino, R., Tomei, P.: Global asymptotic observers for nonlinear systems via filtered trans-
formations. IEEE Trans. Autom. Control 37(8), 1239–1245 (1992)
336. Marino, R., Peresada, S., Valigi, P.: Adaptive input-output linearizing control of induction
motors. IEEE Trans. Autom. Control 38(2), 208–221 (1993)
337. Marino, R., Cinili, F.: InputOutput decoupling control by measurement feedback in four-
wheel-steering vehicles. IEEE Trans. Control Syst. Technol. 17(5), 1163–1172 (2009)
338. Martin, P., Rouchon, P.: Two remarks on induction motors. In: CESA ’96 IMACS Multicon-
ference, vol. 1, pp. 76–79. Lille, France (1996)
339. Maroteaux, F., Saad, C.: Diesel engine combustion modelling for hardware in the loop
applications: effects of ignition delay time model. Energy, Elsevier 57(1), 641–652 (2013)
340. Martin, Ph. Rouchon, P.: Systèmes plats: planification et suivi des trajectoires. Journées
X-UPS, École des Mines de Paris, Centre Automatique et Systèmes, Mai (1999)
341. Marzbanrada, J., Ahmadib, G., Zohoorc, H., Hojjatd, Y.: Stochastic optimal preview control
of a vehicle suspension. J. Sound Vib., Elsesier 275, 973–990 (2004)
342. Medeiros, H., Park, J., Kak, A.C.: Distributed object tracking using a cluster-based Kalman
filter in wireless camera networks. IEEE J. Sel. Top. Sign. Process. 2, 448–463 (2008)
343. Mendalek, N., Al-Haddad, K., Fnaiech, F., Dessaint, L.A.: Nonlinear control techique to
enhance dynamic performance of a shunt active power filter. IEE Proc. Electr. Power Appl.
150(4), 373–379 (2003)
344. Mercorelli, P., Lehmann, K., Liu, S.: Robust flatness based control of an electromagnetic
linear actuator using adaptive PID controller. In: Proceedings of IEEE Decision and Control
Conference, pp. 3790–3795 (2003)
345. Meyer, J., Yurkovich, S., Midlam-Mohler, S.: Air-to-fuel ratio switching frequency control
for gasoline engines. IEEE Trans. Control Syst. Technol. 21(3), 636–648 (2013)
346. Micheau, P., Oddo, R., Lecours, G.: Engine speed limiter for watercrafts. IEEE Trans. Control
Syst. Technol. 14, 579–585 (2006)
347. W. Mielczarski, Observing the state of a synchronous generator. Int. J. Control. Part 1:
Theory, Taylor and Francis. 45(3), 987–1000 (1987)
348. Mielczarski, W., Observing the state of a synchronous generator. Part 2, applications. Int. J.
Control, Taylor and Francis 45(3), 1001–1021 (1987)
349. Menhour, L., d’Andr’ea-Novel, B., Boussard, C., Fliess, M., Mounier, H.: Algebraic nonlin-
ear estimation and flatness-based lateral/longitudinal control for automotive vehicles. In: 14th
International IEEE Conference in Intelligent Transportiation Systems, Washington (2011)
350. Menhour, L.: Brigitte d’Andréa-Novel, Michel Fliess and Hugues Mounier, Commande
couplée longitudinale/latérale de véhicules par platitude et estimation algébrique, Manuscrit
publié dans, 7e Conférence Internationale Francophone d’Automatique, Grenoble, France
(2012)
351. Menhour, L., d’Andre’a-Novel, B., Fliess, M., Mounier, H.: Coupled nonlinear vehicle
control: Flatness-based setting with algebraic estimation techniques. Control Eng. Pract.,
Elsevier 22, 135–146 (2014)
352. Meurer, T., Zeitz, M.: A modal approach to flatness-based control of flexible structures.
PAMM—Proc. Appl. Math. Mech., Wiley 4, 133–134 (2004)
718 References

353. Mianzo, L., Peng, H.: Output feedback H∞ preview control of an electro-mechanical valve
actuator. IEEE Trans. Control Syst. Technol. 15(3), 428–437 (2007)
354. Miklosovic, R., Radke, A., Gao, Z.: Discrete implementation and generalization of the
Extended State Observer. In: Proceedings of the Americal Control Conference, Minneapolis,
Minnesota, USA (2006)
355. Mohseni, M., Islam, S.M.: Transient control of DFIG-based wind power plants in compliance
with the australian grid code. IEEE Trans. Power Electron. 27(6), 2813–2824 (2012)
356. Montanaro, U., di Gaeta, A., Giglio, V.: Robust discrete-time MRAC with minimal controller
synthesis of an electronic throttle body. IEEE/ASME Trans. Mechatron. 19(2), 524–537
(2014)
357. Moulin, P., Chauvin, J.: Modeling and control of the air system of a turbocharged gasoline
engine. Control Eng. Pract. 19, 287–297 (2011)
358. Mounier, H., Rudolph, J.: Trajectory tracking for π -flat nonlinear dealy systems with a motor
example. In: Isidori, A., Lamnabhi-Lagarrigue, F., Respondek, W. (eds.) Nonlinear Control
in the Year 2000, Lecture Notes in Control and Information Science, vol. 258, pp. 339–352.
Springer (2001)
359. Moreira, L., Fossen, T.I., Soares, G.G.: Path following control system for a tanker ship model.
Ocean Eng., Elsevier 34(14–15), 2074–2085 (2007)
360. Moriwaki, K., Tanaka, K.: Navigation control for electric vehicles using nonlinear state
feedback H-infinity control. Nonlinear Anal., Elsevier 71, 2920–2933 (2009)
361. Moulin, P.: Modélisation et commande des systèmes d’ air des moteurs suralimentés. Thèse
de doctorat, Ecole Nationale Supérieure des Mines de Paris (2010)
362. Mounier, H., Rudolph, J.: Trajectory tracking for π -flat nonlinear dealy systems with a motor
example. In: Isidori, A., Lamnabhi-Lagarrigue, F., Respondek, W. (eds.) Nonlinear Control
in the Year 2000, Lecture Notes in Control and Information Science, vol. 258, pp. 339–352.
Springer (2001)
363. Mounier, H., Rudolph, J., Woittenneck, F.: Boundary value problems and convolutional
systems over rings of ultradistributions. In: Advances in the Theory of Control, Signal and
Systems with Physical Modelling. Lecture Notes in Control an Information Sciences, pp.
179–188. Springer (2010)
364. Mounier, H.: Document de synthèse pour l’ habilitation à diriger des recherches. Université
de Paris XI, Juillet (2005)
365. Moon, H.S., Kim, Y.B., Beattie, R.J.: Multi sensor data fusion for improving performance
and reliability of fully automatic welding system. Int. J. Adv. Manuf. Technol., Springer 28,
286–293 (2006)
366. Narikiyoa, T., Sahashib, J., Misao, K.: Control of a class of underactuated mechanical sys-
tems. Nonlinear Anal. Hybrid Syst., Elsevier 2, 231–241 (2008)
367. N’Doye, L., Voos, H., Darouach, M.: Observer-based approach for fractional-order chaotic
synchronization and secure communication. IEEE J. Emerg. Sel. Top. Circ. Syst. 3, 442–450
(2013)
368. Negrea, A.C., Imecs, M., Incze, I.I., Pop, A., Szabo, C.: Error compensation methods in speed
identification using incremental encoder. In: 2012 International Conference and Exposition
on Electrical and Power Engineering (EPE), lasi, Romania (2012)
369. Nettleton, E., Durrant-Whyte, H., Sukkarieh, S.: A robust architecture for decentralized
data fusion. In: ICAR03, 11th International Conference on Advanced Robotics, Coimbra,
Portugal, 2003
370. Niknam, T., Firouzi, B.: A practical algorithm for distribution state estimation including
renewable energy sources. Renewable Energy, Elsevier 34, 2309–2316 (2009)
371. Ning, Z.H., Yan, Q.X.: Stabilization of the wave equation with variable coefficients and a
delay in dissipative boundary feedback. J. Math. Anal. Appl., Elsevier 367, 167–173 (2010)
372. Ng, G.W.: Intelligent Systems—Fusion, Tracking and Control. Research Studies Press,
England (2003)
373. Nguyen, L.H., Hong, K.S.: Synchronization of coupled chaotic FitzHughNagumo neurons
via Lyapunov functions. Math. Comput. Simul., Elsevier 82, 590–603 (2011)
References 719

374. Nguyen, A., Lauber, J., Dambrine, M.: Multi-objective control design for turbo-charged
spark ignited air system: a switching Takagi-Sugeno model approach. In: 2013 American
Control Conference, Washington DC, USA (2013)
375. Nitsche, R., Schwarzmann, D., Hanscke, J.: Nonlinear internal model control of diesel air
systems. Oil Gas Sci. Technol. Rev. IFP Energies Nouvelles 62(4), 573–586 (2011)
376. Nosrati, K., Azemi, A., Rostami, A.S., Pariz, N.: Unscented Kalman filter applied to noisy
synchronization of Rossler chaotic system. In: 3rd International Conference on Advanced
Computer Control (ICACC), Harbin, China (2011)
377. Nounou, H.N., Rehman, H.: Application of adaptive fuzzy control to AC machines. Appl.
Soft Comput., Elsevier 7(3), 899–907 (2007)
378. Okatan, A., Hajiyev, C., Hajiyeva, U.: Fault detection in sensor information fusion Kalman
Filter. Int. J. Electron. Commun., Elsevier 47(4), 1657–1665 (2001)
379. Okou, F., Dessaint, L.-A., Akhrif, O.: Power systems stability enhancement using a wide-area
signals based hierarchical controller. IEEE Trans. Power Syst. 20(3), 1465–1477 (2005)
380. Olfati-Saber, R.: Distributed Kalman filter with embedded consensus filters. In: Proceedings
of 44th IEEE Conference on Decision and Control, pp. 8179–8184, Seville, Spain (2005)
381. Olfati-Saber, R.: Distributed Kalman filtering and sensor fusion in sensor networks. Lect.
Notes Control Inf. Sci. 331, 157–167 (2006)
382. Oriolo, G., De Luca, A., Vendittelli, M.: WMR control via dynamic feedback linearization:
design, implementation and experimental validation. IEEE Trans. Control Syst. Technol.
10(6), 835–852 (2002)
383. Ovalle, A., Ramos, G., Bacha, S., Hably, A., Rameau, A.: Decentralized control of voltage
source converters in microgrids based on the application of instantaneous power theory.
IEEE Trans. Industr. Electron. (2015). doi:10.1109/TIE.2014.2336638
384. Pan, C.Z., Lai, X.Z., Yang, S.X., Wu, M.: An efficient neural network approach to track-
ing control of an autonomous surface vehicle with unknown dynamics. Expert Syst. Appl.
Elsevier, 40, 1629–1635 (2013)
385. Pena, R., Cardenas, R., Reyes, E., Clare, J., Wheeler, P.: Control of a doubly fed induction
generator via an indirect matrix converter with changing DC voltage. IEEE Trans. Industr.
Electron. 58(10), 4664–4674 (2011)
386. Peresada, S., Tilli, A., Tonielli, A.: Power control of a doubly fed induction machine via
output feedback. Control Eng. Pract., Elsevier 12, 4157 (2004)
387. Petkovic, I.: Dynamique de la phase dans des jonctions Josephson ferromagnétiques. Thèse
de doctorat, Université Paris-Sud 11, (2009)
388. Pinsky, M.: Partial Differential Equations and Boundary Value Problems. Prentice-Hall,
Englewood Cliffs (1991)
389. Piyabongkarn, D., Sun, Y., Rajamani, R., Sezen, A., Nelson, B.J.: Travel range extension
of a MEMS electrostatic microactuator. IEEE Trans. Control Syst. Technol. 13(1), 138–145
(2005)
390. Postma, M., Nagamune, R.: Air-fuel ratio control of spark ignition engines using a switching
LPV controller. IEEE Trans. Control Syst. Technol. 20(5), 1175–1187 (2012)
391. Powell, J.D., Fekete, N.P., Chang, C.-F.: Observer-based air-fuel ratio control. IEEE Control
Syst. Mag. 18(5), 72–83 (1998)
392. Posznyak, A.S., Yu, W., Sanchez, E.N.: Identification and control of unknown chaotic systems
via dynamic neural network. IEEE Trans. Circuits Syst. I 46(12), 1491–1495 (1999)
393. Purwar, S., Kar, I.N., Jha, A.N.: Adaptive control of robot manipulators using fuzzy logic
systems under actuator constraints. Fuzzy Sets Syst., Elsevier 152, 651–664 (2004)
394. Puscasu, G., Codres, B.: Nonlinear system identification and control based on modular neural
networks. Int. J. Neural Syst., World Scientific 21(4), 319–334 (2011)
395. Puthusserypady, S., Kurian, A.P.: Variants of Kalman filter for synchronization of chaotic
systems. In: Kordic, V. (ed.) Kalman Filter. InTech Croatia (2010)
396. Qi, R., Tao, G., Tan, C., Yao, X.: Adaptive control of discrete-time state-space TS fuzzy
systems with general relative degree. Fuzzy Sets Syst., Elsevier 217, 2240 (2013)
720 References

397. Raffo, G.V., Ortega, M.G., Rubio, F.R.: An integral predictive/nonlinear H∞ control structure
for a quadrotor helicopter. Automatica, Elsevier 46, 29–39 (2010)
398. Raffo, G.V., Ortega, M.G., Rubio, F.R.: MPC with nonlinear H∞ control for path tracking
of a quad-rotor helicopter. In: Proceedings of the 17th World Congress The International
Federation of Automatic Control Seoul, Korea, 6–11 July 2008
399. Raffo, G.V., Ortega, M.G., Rubio, F.R.: Path tracking of a UAV via an underactuated
H-infinity control strategy. Eur. J. Control, Elsevier 17, 194–213 (2011)
400. Raffo, G.V., Ortega, M.G., Rubio, F.R.: Nonlinear H-infinity controller for the quad-rotor
helicopter with input coupling. In: 18th World Congress of the IFAC, 2011, Milan, Italy,
Aug 2011
401. Sira-Ramirez, H., Aguilar-Ibanez, C.: On the control of the hovercraft system. Dyn. Control,
Springer 10, 151–163 (2000)
402. Rao, B.S., Durrant-Whyte, H.F.: Fully decentralized algorithm for multisensor Kalman fil-
tering. IEE Proc. D 138, 413–451 (1991)
403. Raoufi, R., Zinober, A.: Smooth adaptive sliding mode observers in uncertain chaotic com-
munication. Int. J. Syst. Sci., Taylor and Francis 38, 931–942 (2007)
404. Rigatos, G.G., Tzafestas, C.S., Tzafestas, S.G.: Mobile robot motion control in partially
unknown environments using a sliding-mode fuzzy-logic controller. Robot. Auton. Syst.,
Elsevier 33, 1–11 (2000)
405. Rigatos, G., Zhang, Q.: Fuzzy Model Validation using the Local Statistical Approach, Pub-
lication Interne IRISA No 1417. Rennes, France (2001)
406. Rigatos, G.G.: Fuzzy stochastic automata for intelligent vehicle control. IEEE Trans. Industr.
Electron. 50, 76–79 (2003)
407. Rigatos, G.G., Tzafestas, S.G.: Adaptive fuzzy control for the ship steering problem. J.
Mechatron., Elsevier 16(6), 479–489 (2006)
408. Rigatos, G.G., Tzafestas, S.G.: Extended Kalman filtering for fuzzy modelling and multi-
sensor fusion. Math. Comput. Model. Dyn. Syst., Taylor & Francis 13, 251–266 (2007)
409. Rigatos, G.G., Tzafestas, S.G.: H∞ tracking of uncertain SISO nonlinear systems: an
observer-based adaptive fuzzy approach. Int. J. Syst. Sci., Taylor & Francis 38, 459–472
(2007)
410. Rigatos, G.G.: Adaptive fuzzy control with output feedback for H∞ tracking of SISO non-
linear systems. Int. J. Neural Syst., World Scientific 18(4), 1–16 (2008)
411. Rigatos, G.G.: Particle Filtering for state estimation in industrial robotic systems. IMeche J.
Syst. Control Eng., Sage Publications 222(6), 437–455 (2008)
412. Rigatos, G.G.: Particle filtering for state estimation in nonlinear industrial systems. IEEE
Trans. Instrum. Meas. 58(11), 3885–3901 (2009)
413. Rigatos, G.G.: Adaptive fuzzy control of DC motors using state and output feedback. Electr.
Power Syst. Res., Elsevier 79(11), 1579–1592 (2009)
414. Rigatos, G., Zhang, Q.: Fuzzy model validation using the local statistical approach. Fuzzy
Sets Syst., Elsevier 60(7), 882–904 (2009)
415. Rigatos, G.G., Siano, P., Merola, E.: Sensorless control of DC and induction motors using
Kalman Filtering, MASCOT 2009. In: IMACS Workshop on Scientific Computation, Italian
Institute for Calculus Applications, Roma Italy (2009)
416. Rigatos, G.G., Siano, P., Piccolo, A.: A neural network-based approach for early detection
of cascading events in electric power systems. IET Gener. Transm. Distrib. 3(7), 650–665
(2009)
417. Rigatos, G.G.: Particle and Kalman filtering for state estimation and control of DC motors.
ISA Trans., Elsevier 48(1), 62–72 (2009)
418. Rigatos, G.G., Particle and Kalman filtering for fault diagnosis in DC motors. In: IEEE VPPC
2009, IEEE 5th Vehicle Power Propulsion Conference, Michigan, USA, Sept 2009
419. Rigatos, G.G.: Sigma-point Kalman Filters and particle filters for integrated navigation of
unmanned aerial vehicles. In: International Workshop on Robotics for Risky Interventions
and Environmental Surveillance, RISE 2009, Brussels, Belgium, Jan 2009
References 721

420. Rigatos, G.G.: Model-based and model-free control of flexible-link robots: a comparison
between representative methods. Appl. Math. Model., Elsevier 33(10), 3906–3925 (2009)
421. Rigatos, G.G.: Extended Kalman and particle Filtering for sensor fusion in motion control
of mobile robots. Math. Comput. Simul., Elsevier 81(3), 590–607 (2010)
422. Rigatos, G.G.: Extended Kalman and particle Filtering for sensor fusion in motion control
of mobile robots. Math. Comput. Simul., Elsevier 81(3), 590–607 (2010)
423. Rigatos, G., Siano, P.: Design of robust electric power system stabilizer using Kharitonov’s
theorem. Math. Comput. Simul., Elsevier 82(1), 181–191 (2011)
424. Rigatos, G.G.: A derivative-free Kalman Filtering approach for sensorless control of nonlin-
ear systems. In: IEEE ISIE 2010, IEEE International Symposium on Industrial Electronics.
Bari, Italy, July, 2010
425. Rigatos, G., Siano, P.: A derivative-free extended information filtering approach for sen-
sorless control of nonlinear systems. In: MASCOT 2010. IMACS Workshop on Scientific
Computation, Italian Institute for Calculus Applications, Gran Canaria, Spain, Oct 2010
426. Rigatos, G., Al-Khazraji, A.: Flatness-based adaptive fuzzy control for MIMO nonlinear
dynamical systems. In: Nonlinear Estimation and Applications to Industrial Systems Control,
Nova Publications (2011)
427. Rigatos, G.G.: Modelling and Control for Intelligent Industrial Systems: Adaptive Algo-
rithms in Robotcs and Industrial Engineering. Springer, New York (2011)
428. Rigatos, G.G.: Adaptive fuzzy control for field-oriented induction motor drives. Neural
Comput. Appl., Springer 21(1), 9–23 (2011)
429. Rigatos, G.G.: Flatness-based adaptive fuzzy control for nonlinear dynamical systems. In:
AIM 2011, IEEE/ASME, International Conference on Advanced Intelligent Mechatronics,
Budapest, Hungary, July 2011
430. Rigatos, G.G.: Distributed nonlinear filtering under packet drops and variable delays for
robotic visual servoing. In: Robot Arms, In-Tech Publications, Vienna, Austria (2011)
431. Rigatos, G.G., Siano, P.: Design of robust electric power system stabilizers using
Kharitonov’s theorem. Math. Comput. Simul., Elsevier 81(1), 181–191 (2011)
432. Rigatos, G.G.: Differential flatness theory and Extended Kalman Filtering for sensorless
control of induction motors. In: 43rd ISCIE International Symposium on Stochastic Systems
Theory and Its Applications, Japan (2011)
433. Rigatos, G.G.: Adaptive fuzzy control of MIMO dynamical systems using differential flat-
ness theory. In: IEEE ISIE 2012, 21st International Symposium on Industrial Electronics.
Hangzhou, China, May 2012
434. Rigatos, G.G.: A derivative-free Kalman Filtering approach to state estimation-based control
of nonlinear dynamical systems. IEEE Trans. Industr. Electron. 59(10), 3987–3997 (2012)
435. Rigatos, G.G.: Distributed filtering over sensor networks for autonomous navigation of
UAVs. Intel. Serv. Robot., Springer 5(3), 179–198 (2012)
436. Rigatos, G.G.: Derivative-free nonlinear Kalman Filtering for MIMO dynamical systems:
Application to multi-DOF robotic manipulators. J. Adv. Rob. Syst. (special issue on Robot
Manipulators) (2012). doi:10.5772/10679
437. Rigatos, G.G.: Nonlinear Kalman Filters and particle filters for integrated navigation of
unmanned aerial vehicles. Robot. Auton. Syst., Elsevier 60(7), 978–995 (2012)
438. Rigatos, G., Siano, P., Zervos, N.: A Distributed state estimation approach to condition
monitoring of nonlinear electric power systems. Asian J. Control., Wiley 15(3), 849–860
(2013)
439. Rigatos, G. Siano, P.: DFIG control using differential flatness theory and extended kalman
filtering. In: IFAC INCOM 2012, 14th IFAC International Conference on Information Control
Problems in Manufacturing, Bucharest, Romania, May 2012
440. Rigatos, G., Siano, P.: Sensorless nonlinear control of induction motors using unscented
Kalman Filtering. In: IEEE IECON 2012, 38th international Conference of the Industrial
Electronics Society, Montreal, Canada, Oct 2012
441. Rigatos, G., Siano, P., Zervos, N.: Derivative-free nonlinear Kalman filtering for PMSG
sensorless control. In: Habib, M. (ed.) Mechatronics Engineering: Research Development
and Education, Wiley (2012)
722 References

442. Rigatos, G. Siano, P.: Validation of fuzzy Kalman Filters using the local statistical approach
to fault diagnosis. In: IMACS Mascot 2012, Annual Conference of the Italian Institute for
Calculus Applications, Gran Canaria, Spain, Oct 2012
443. Rigatos, G.G.: Adaptive fuzzy control for non-linear dynamical systems based on differential
flatness theory. IET Control Theor. Appl. 6(17), 2644–2656 (2012)
444. Rigatos, G., Siano P., Zervos, N.: PMSG sensorless control with the use of the derivative-free
nonlinear Kalman Filter. In: IEEE ICCEP 2013, IEEE International Conference on Clean
Electrical Power, Alghero, Sardinia, Italy, June 2013
445. Rigatos, G, Siano, P.: Control of underactuated robotic systems with the use of the derivative-
free nonlinear Kalman filter. In: ICNAAM 2013, 11th International Conference on Numerical
Analysis and Applied Mathematics, AIP Conference Proceedings, vol. 1558, pp. 2551–2556
(2013)
446. Rigatos, G., Zervos, N.: Channel equalization and synchronization in chaotic communica-
tions using a dual Kalman filtering scheme. In: ICNAAM 2013, 14th International Confer-
ence on Numerical Analysis and Applied Mathematics, AIP Conference Proceedings, vol.
1558, pp. 2539–2544 (2013)
447. Rigatos, G., Siano, P., Zervos, N., Cecati, C.: Derivative-free nonlinear Kalman Filtering
for control of three-phase voltage source converters. In: IEEE IECON 2013, 39th IEEE
Conference on Industrial Electronics, Vienna, Austria, Nov 2013
448. Rigatos, G., Siano P., Zervos N.: Control and disturbances compensation for doubly-fed
induction generators using derivative-free nonlinear Kalman Filter. In: IEEE IECON 2013,
39th IEEE Conference on Industrial Electronics, Vienna, Austria, Nov 2013
449. Rigatos, G., Siano, P., Zervos, N.: PMSG sensorless control with the derivative-free nonlinear
Kalman Filter for distributed generation units. In: 11th IFAC International Workshop on
Adaptation and Learning in Control and Signal Processing (ALCOSP’2013), Caen, France,
July 2013
450. Rigatos, G.: Derivative-free distributed filtering for MIMO robotic systems under delays and
packet drops. In: Yin-Tien Wang (ed.) International Journal of Advanced Robotic Systems,
Intech Publications (2013). doi:10.5772/54186
451. Rigatos, G.G.: Sensor fusion-based dynamic positioning of ships using Extended Kalman
and Particle Filtering. Robotica, Cambridge University Press 31(3), 389–403 (2013)
452. Rigatos, G.G.: Robust control of valves in ship diesel engines with the use of the derivative-
free nonlinear Kalman Filter. IMeche J. Syst. Control Eng., Sage Publications 228(9), 631–
644 (2014)
453. Rigatos, G.G.: Advanced Models of Neural Networks: Nonlinear Dynamics and Stochasticity
in Biological Neurons. Springer, Berlin (2014)
454. Rigatos, G.G.: A differential flatness theory approach to observer-based adaptive fuzzy con-
trol of MIMO nonlinear dynamical systems. Nonlinear Dyn., Springer 76(2), 1335–1354
(2014)
455. Rigatos, G.: A differential flatness theory approach to adaptive fuzzy control of chaotic
dynamical systems. In: IEEE SSCI 2014, Orlando, Florida, USA (2014)
456. Rigatos, G., Siano, P., Zervos, N.: Sensorless control of distributed power generators with the
derivative-free nonlinear Kalman Filter. IEEE Trans. Industr. Electron. 61(11), 6369–6382
(2014)
457. Rigatos, G., Zhu, G., Youssef, H., Boulkroune, A.: Flatness-based adaptive fuzzy control
of electrostatically actuated MEMS using output feedback. In: ISI Internal Research Report
(2014)
458. Rigatos, G., Siano, P.: A nonlinear H-infinity feedback control approach for asynchronous
generators. In: IEEE ICCEP 2015, 5th International Conference on Clean Electrical Power,
Taormina, Sicily, Italy, June 2015
459. Rigatos, G.: A new concept on flatness-based control of nonlinear dynamical systems, ISI
Internal Report (2015)
460. Rigatos, G., Siano, P., Zervos, N., Cecati, C.: Nonlinear control of the three-phase inverter
using the derivative-free nonlinear Kalman Filter. In: IET ACDC 2015, The 11th IET Inter-
national Conference on AC and DC Power Transmission, Birmingham UK, Feb 2015
References 723

461. Rigatos, G., Raffo, F.: Input-output linearizing control of the underactuated hovercraft using
the Derivative-free nonlinear Kalman Filter. Unmanned Systems, World Scientific (2015)
462. Rosencrantz, M., Gordon, G., Thrun, S.: Decentralized data fusion with distributed particle
filtering. In: Proceedings of the Conference of Uncertainty in AI (UAI), Acapulco, Mexico
(2003)
463. Rouchon, P., Laroche, B., Martin, P.: Motion planning of the heat equation. Int. J. Robust
Nonlinear Control, Wiey 10:629–643 (2000)
464. Rouchon, P.: Flatness-based control of oscillators. ZAMM J. Appl. Math. Mech. 85(6),
411–421 (2005)
465. Rudolph, J.: Flatness based control of distributed parameter systems. In: Examples and
Computer Exercises from Various Technological Domains, Shaker Verlag, Aachen (2003)
466. Saadatpour, A., Levi, M.: Travelling waves in chains of pendula. Physica D 244, 68–73
(2013)
467. Salberg, S.A., Maybeck, P.S., Oxley, M.E.: Infinite-dimensional sampled-data Kalman Fil-
tering and stochastic heat equation. In: 49th IEEE Conference on Decision and Control,
Atlanta, Georgia, USA, Dec 2010
468. Saleh, M.H., Aldvidyan, K.M., Tatlicioglu, E., Dawson, D.M.: Robust backstepping nonlin-
ear control for parallel-plate micro electrostatic actuators. In: Proceedings of the 49th IEEE
Conference on Decision and Control, Atlanta, Georgia, USA, Dec 2010
469. Särrkä, S.: On unscented Kalman Filtering for state estimation of continuous-time nonlinear
systems. IEEE Trans. Autom. Control 52(9), 1631–1641 (2007)
470. Sepulveda, R., Montiel, O., Castello, O., Melin, P.: Embedding a high-speed interval type-2
fuzzy controller for a real plant into a FPGA. Appl. Soft Comput., Elsevier 12(3), 988–998
(2012)
471. Shafiee, Q., Vasquez, J., Guerrero, J.M.: Distributed secondary control for islanded micro-
grids: a networked control systems approach. In: IEEE IECON 2012, 38th Annual Confer-
ence of the IEEE Industrial Electronics Society, Montr’eal, Canada (2012)
472. Shcheglov, K., Jiang, X., Toch, R., Chang, Z., Yang, E.H.: Hybrid linear microactuators and
their control models for mirror shape correction. J. Micro-Nano Mechatron., Springer 4,
159–167 (2008)
473. Schiffer, J., Goldin, D., Raisch, J., Sezi, T.: Synchronization of droop-controlled microgrids
with distributed rotational and electronic generation. In: IEEE Conference on Decision and
Control CDC 2013, pp. 2334–2339, Firenze, Italy, Dec 2013
474. Schiffer, J., Ortega, R., Astolfi, A., Raisch, J., Sezi, T.: Conditions for stability of droop-
controlled inverter-based microgrids. Automatica, Elsevier 50(10), 2457–2469 (2014)
475. Schiffer, J., Ortega, R., Astolfi, A., Raisch, J., Sezi, T.: Stability of synchronized motions of
inverter-based microgrids under droop control. In: 19th IFAC World Congress, Cape Town,
South Africa (2014)
476. Schenato, L.: Optimal sensor fusion for distributed sensors subject to random delay and
packet loss. In: Proceedings of the 46th IEEE Conference on Decision and Control, New
Orleans, Luisiana, USA, Dec 2007
477. Schenato, L.: Optimal estimation in networked control systems subject to random delay and
pachet drop. IEEE Trans. Autom. Control 53, 1311–1317 (2008)
478. Schuurman, D.C., Capson, D.W.: Robust direct visual servo using network-synchronized
cameras. IEEE Trans. Rob. Autom. 20, 319–334 (2004)
479. Sekhavat, S., Rouchon, P., Hermosillo, P.J.: Computing the flat outputs of engel differential
systems: the case study of the bi-steerable car. In: American Control Conference (2001)
480. Seguritan, A., Rotuno, M.: Torque pulsation compensation for a DC motor using an extended
Kalman Filter approach. In: Proceedings of the 41st IEEE Conference on Decision and
Control, Las Vegas, Nevada, USA, Dec 2002
481. Sepulveda, R., Montiel, O., Castello, O., Melin, P.: Embedding a high-speed interval type-2
fuzzy controller for a real plant into a FPGA. Appl. Soft Comput., Elsevier 12(3), 988–998
(2012)
724 References

482. Serrano, M.E., Scaglia, G.J.E., Godoy, S.A., Mut, V., Ortiz, O.A.: Trajectory tracking of
underactuated surface vessels: a linear algebra approach. IEEE Trans. Control Syst. Tech.
22(3), 1103–1111 (2014)
483. Shafiee, Q., Stefanovic, C., Draginevic, T., Popovskii, P., Vasquez, J., Guerrero, J.M.: Robust
networked control scheme for distributed secondary control of islanded microgrids. IEEE
Trans. Industr. Electron. 61(10), 5363–5374 (2014)
484. Shih, M.C., Wang, T.Y.: Active control of electro-rheological fluid embedded pneumatic
vibration isolator. Integr. Comput. Aided Eng. 15(3), 267–276 (2008)
485. Shima, T., Rasmussen, S.J., Chandler, P.: UAV team decision and control using efficient
collaborative estimation. ASME J. Dyn. Syst. Measur. Control 129(5), 609–619 (2007)
486. Shimizu, Y., Ohtsuka, T., Diehl, M.: Nonlinear receding horizon control of an underactuated
hovercraft with a multiple-shooting-based algorithm. In: Proceedings of the 2006 IEEE
International Conference on Control Applications, Munich, Germany, Oct 2006
487. Shrinivashan, P., Gallash, C.O., Kraft, M.: Three-dimensional electrostatic actuators for
tunable optical micro-cavities. Sens. Actuators A, Elsevier 161, 191–198 (2010)
488. Sienel, W.: Estimation of the tire cornering stiffness and its application to active car steering.
In: Proceedings of the 36th Conference on Decision & Control, San Diego, California, USA,
Dec 1997
489. Silver, D., Salmon, R., Barbieri. E., Drakunov, S.: Towards an integrated welding testbed:
temperature field control. In: Proceedings of the IEEE ACC 98, American Control Confer-
ence, Philadelphia, Pennsylvania, June 1998
490. Simon, D.: A game theory approach to constrained minimax state estimation. IEEE Trans.
Sig. Process. 54(2), 405–412 (2006)
491. Sira-Ramirez, H., Matamoros-Sanchez, A., Goodall, R.M.: Flatness based control of a sus-
pension system: a GPI observer approach. In: 18th IFAC World Congress, Milano, Italy
(2011)
492. Sira-Ramirez, H.: Dynamic second-order sliding-mode control of the hovercraft vessel. IEEE
Trans. Control Syst. Tech. 10(6), 860–865 (2002)
493. Sira-Ramirez, H., Luviano-Juarez, A., Cortes-Romero, J.: A disturbance rejection flatness-
based linear output feedback control approach for tracking tasks of Chua’s circuit. In: 50th
IEEE Conference on Decision and Control and European Control Conference (CDC-ECC),
Orlando, Florida (2011)
494. Sira-Ramirez, H., Fliess, M.: On the output feedback control of a synchronous generator. In:
43rd IEEE Conference on Decision and Control, Bahamas, Dec 2004
495. Sira-Ramirez, H., Agrawal, S.: Differentially Flat Systems. Marcel Dekker, New York (2004)
496. M’Sirdi, N.K., Jaballah, B., Rabhi, A., Messaoud, H., Naamane, A.: Differential estimators
for state observers in vehicle dynamics: HOSM and ALIEN. In: 18th IFAC World Congress,
Milano, Italy, Aug–Sep (2011)
497. Sivaramakrishnan, S.: Simultaneous identification of tire cornering stiffnesses and vehicle
center of gravity. In: 2008 American Control Conference, Seattle, Washington, USA, June
2008
498. Slotine, J.J.: Applied Nonlinear Control, Prentice-Hall, Englewood Cliffs (1991)
499. Soltun, R.A., Ashrafiuon, H., Muske, K.R.: ODE-based obstacle avoidance and trajectory
planning for unmanned surface vessels. Robotica, Cambridge University Press 29, 691–703
(2010)
500. Song, E., Lynch, A.F., Dinavahi, V.: Experimental validation of a flatness-based control for
a voltage source converter. In: Proceedings of the 2007 American Control Conference, New
York, USA (2007)
501. Song, T., Bin, C., Wang, Y.: Fuzzy adaptive output feedback control for MIMO nonlinear
systems. Fuzzy Sets Syst., Elsevier 156, 285–299 (2005)
502. Sorensen, A.J., Egeland, O.: Design of ride control system for surface effect ships using
dissipative control. Automatica, Elsevier 31, 183–199 (1995)
503. Spooner, J.T., Passino, K.M.: Stable adaptive control using fuzzy systems and neural net-
works. IEEE Trans. Fuzzy Syst. 4, 339–359 (1996)
References 725

504. Su, C.Y., Stephanenko, Y.: Adaptive fuzzy control of a class of nonlinear systems. IEEE
Trans. Fuzzy Syst. 2, 285–294 (1994)
505. Sugihira, S., Ohmori, H.: Model-based starting control of SI engines via adaptive
feedback linearization. In: SICE Annual Conference 2008, The University of Electro-
Communications, Japan, Aug 2008
506. Sun, S.L., Deng, Z.L.: Distributed optimal fusion steady-state Kalman filter for systems with
coloured measurement noises. Int. J. Syst. Sci., Taylor & Francis 36, 113–118 (2005)
507. Sun, J., Zhang, C., Guo, B.: Distributed full-order optimal fusion filters and smoothers for
discrete-time stochastic singular systems. Int. J. Syst. Sci., Taylor & Francis 42, 507–516
(2011)
508. Suresh, S., Kannan, N., Sundararajan, N., Saratchandran, P.: Neural adaptive control for
vibration suppression in composite fin-tip of aircraft. Int. J. Neural Syst., World Scientific
18(3), 219–231 (2008)
509. Tanaka, K., Iwasaki, M., Wang, H.O.: Switching control of an R/C hovercraft: stabilization
and smooth switching. IEEE Trans. Syst. Man Cybern. B Cybern. 31(6), 853–863 (2001)
510. Tang, C.P., Miller, P.T., Krovi, V.N., Ryu, J.C., Agrawal, S.K.: Differential flatness-based
planning and control of a wheeled mobile manipulator theory and experiment. IEEE/ASME
Trans. Mechatron. 16(4), 768–773 (2011)
511. Tee, K.P., Ge, S.S., Tay, F.E.H.: Adaptive control of electrostatic microactuators with bidi-
rectional drive. IEEE Trans. Control Syst. Technol. 17(2), 340–352 (2009)
512. Tee, K.F., Koh, K.F., Quek, S.T.: Numerical and experimental studies of a substructural
identification strategy. Struct. Health Monit., Sage Publications 8, 397–410 (2009)
513. Tee, K.P., Ge, S.S., Tay, E.H.: Output feedback adaptive control of electrostatic microactu-
ators. In: 2009 American Control Conference, St. Louis, MO, USA, June 2009
514. Terzic, B., Jadric, M.: Design and implementation of the extended Kalman Filter for the
speed and rotor position estimation of brushless DC motor. IEEE Trans. Industr. Electron.
48(6) (2001)
515. Tham, Y.K., Wang, H., Teoh, E.K.: Multi-sensor fusion for steerable four-wheeled industrial
vehicles. Control Eng. Pract., Elsevier 7, 1233–1248 (1999)
516. Tang, C.P., Miller, P.T., Krovi, V.N., Ryu, J.C., Agrawal, S.K.: Differential flatness-based
planning and control of a wheeled mobile manipulator theory and experiment. IEEE/ASME
Trans. Mechatron. 16(4), 768–773 (2011)
517. Tang, H., Weng, L., Dong, Z.Y., Yan, R.: Adaptive and learning control for SI engine model
with uncertainties. IEEE/ASME Trans. Mechatron. 14(1), 93–104 (2009)
518. Tang, H., Zhang, Q.: An efficient numerical inverse scattering algorithm for generalized
Zakharov-Shabat equations with two potential functions, INRIA Research Report, inria-
00447358 (2010)
519. Tee, K.F., Koh, C.C., Quek, S.T.: Numerical and experimental studies of a substructural
identification strategy. Struct. Health Monit., Sage Publications 8, 397–410 (2009)
520. Tharmarasa, R., Kirubarajan, T., Peng, J., Lang, T.: Optimization-based dynamic sensor
management for distributed multitarget tracking. IEEE Trans. Syst. Man Cybern. Part C
Appl. Rev. 39(5), 534–546 (2009)
521. Thao, N.G.M., Uchida, K.: Control the active and reactive powers of three-phase grid-
connected photovoltaic inverters using feedback linearization and fuzzy logic. In: 2013 Aus-
tralian Control Conference, Perth, Australia, Nov 2013
522. Tian, G., Gao, Z.: Frequency response analysis of active disturbance rejection based control
system. In: 16th IEEE International Conference on Control Applications, Singapore, Oct
2007
523. Tonati, Y., Djouani, K., Amirat, Y.: Neuro-fuzzy based approach for hybrid force/position
robot control. Integr. Comput. Aided Eng., IOS Press 11(1), 85–98 (2004)
524. Tong, S., Li, H.-X., Chen, G.: Adaptive fuzzy decentralized control for a class of large-scale
nonlinear systems. IEEE Trans. Syst. Man Cybern. Part B Cybern. 34(1), 770–775 (2004)
525. Tong, S., Chen, B., Wang, Y.: Fuzzy adaptive output feedback control for MIMO nonlinear
systems. Fuzzy Sets Syst., Elsevier 156, 285–299 (2005)
726 References

526. Tong, S., Li, Y.: Observer-based adaptive control for strict-feedback nonlinear systems. Fuzzy
Sets Syst., Elsevier 160, 1749–1764 (2009)
527. Song, T., Bin, C., Wang, Y.: Fuzzy adaptive output feedback control for MIMO nonlinear
systems. Fuzzy Sets Syst., Elsevier 156, 285–299 (2005)
528. Tsai, M.T., Tsai, W.I.: Analysis and design of three-phase AC-to-DC converters with high
power factor and near-optimum feedforward control. IEEE Trans. Industr. Electron. 46(3),
535–543 (1999)
529. Tsai, J.S.H., Lu, F.C., Provence, R.S., Shieh, L.S., Han, Z.: A new approach for adaptive
blind equalization of chaotic communication: the optimal linearization technique. Comput.
Math. Appl., Elsevier 58(9), 1687–1698 (2009)
530. Tsygankov, D.: Spontaneous synchronization of Josephson junctions and fiber lasers. Ph.D.
Thesis, Georgia Institute of Technology (2005)
531. Tyner, D.R., Lewis, A.D.: Controllability of a hovercraft model (and two general results). In:
43rd IEEE Conference on Decision and Control, Atlantis, Paradise Island, Bahamas, Dec
2004
532. Tzafestas, S.G., Rigatos, G.G., Kyriannakis, E.J.: Geometry and thermal regulation of GMA
welding via conventional and neural adaptive control. J. Intell. Rob. Syst., Springer 19(2),
153–186 (1997)
533. van der Merwe, R., Wan, E.A., Julier, S.I.: Sigma-point Kalman Filters for nonlinear estima-
tion and sensor-fusion applications to intergrated navigation. In: Proceedings of the AIAA
Guidance, Navigation and Control Conference, Providence, RI, USA, Aug 2004
534. Vercauteren, T., Wang, X.: Decentralized sigma-point information filters for target tracking
in collaborative sensor networks. IEEE Trans. Signal Process. 53(8), 2997–3009 (2005)
535. Villagra, J., d’Andrea-Novel, B., Mounier, H., Pengov, M.: Flatness-based vehicle steering
control strategy with SDRE feedback gains tuned via a sensitivity approach. IEEE Trans.
Control Syst. Technol. 15, 554–565 (2007)
536. Vural, C., Cetinel, C.: Blind equalization of single-input single-output FIR channels for
chaotic communication systems. Digital Signal Process., Elsevier 20, 201–211 (2010)
537. Wai, R.J., Chang, J.M.: Implementation of robust wavelet-neural-network sliding-mode con-
trol for induction servo motor drive. IEEE Trans. Ind. Electron. 50(6), 1317–1334 (2003)
538. Wai, R.J., Chang, H.H.: Backstepping wavelet neural network control for indirect field-
oriented induction motor drive. IEEE Trans. Neural Networks 15(2), 367–382 (2004)
539. Wai, R.J., Chang, J.M.: Intelligent control of induction servo motor drive via wavelet neural
network. Electr. Power Syst. Res., Elsevier 61(1), 67–76 (2001)
540. Wai, R.J., Wang, W.H., Lin, C.Y.: High-performance stand-alone photovoltaic generation
system. IEEE Trans. Ind. Electron. 55(1), 240–250 (2008)
541. Wang, J., Boussaada, I., Cela, A., Mounier, H., Niculescu, S.I.: Analysis and control of
quadrotor via a normal form approach. In: IEEE International Symposium on Mathematical
Theory of Networks and Systems, Melbourne, Australia (2012)
542. Wan, F., Wang, L-X.: On the persistent excitation conditions of adaptive fuzzy system in
nonlinear identifications. In: Proceedings of the 39th IEEE Conference on Decision and
Control, Sydney, Australia, Dec 2000
543. Wang, L.X.: Adaptive Fuzzy Systems and Control: Design and Stability Analysis. Prentice
Hall, Englewood Cliffs (1994)
544. Wang, L.X.: A Course in Fuzzy Systems and Control. Prentice-Hall, Englewood Cliffs (1998)
545. Wang, Y., Nam, K., Fujimoto, H., Hori, Y.: Robust roll and yaw integrated control using
4-wheel steering based on yaw moment and lateral force observers. In: Proceedings of the
IEEJ Technical Meeting Record, IIC-11-138 (2011)
546. Wang, H.P., Bosche, J., Tian, Y., El Hajjaji, A.: Two-loop based dynamical feedback sta-
bilization of a diesel engine with EGR & VGT. In: 50th IEEE conference of Decision and
Control and European Control Conference (CDC-ECC), Orlando, Florida (2011)
547. Wang, F.Y., Gao, Y.: Advanced Studies of Flexible Robotic Manipulators. World Scientific,
New Jersey (2004)
References 727

548. Wang, X., Yaz, E.E.: A new nonlinear-filter-based modulation/demodulation technique for
chaotic communication. In: 2009 American Control Conference, St. Louis, Missouri, USA
(2009)
549. Wang, X., Yaz, E.: Improved chaotic communications using nonlinear filtering. Int. J. Innov.
Comput. Inf. Control 6(5), 2127–2136 (2010)
550. Wang, Q., Jing, Y., Wang, L., Kong, Z.: Backstepping-based direct adaptive fuzzy control
for SISO nonlinear systems. In: Proceedings of the 46th IEEE Conference on Decision and
Control, New Orleans, LA, USA, Dec 2007
551. Wahlström, J., Erikson, L.: Nonlinear EGR and VGT control with integral action for diesel
engines. OGST - Rev. IFP Energies Nouvelles 66(4), 573–586 (2011)
552. Watanabe, K., Tzafestas, S.G.: Filtering, smoothing and control in discrete-time stochastic
distributed-sensor networks, In: Tzafestas, S.G., Watanabe, K. (eds.) Stochastic Large-Scale
Engineering Systems, pp. 229–252, Marcel Dekker (1992)
553. Wei, D.Q., Luo, X.S., Zhang, B., Qin, Y.H.: Controlling chaos in spaceclamped FitzHugh-
Nagumo neuron by adaptive passive method. Nonlinear Anal. Real World Appl., Elsevier
11(3), 1752–1759 (2010)
554. Wesemeier, D., Isermann, R.: Identification of vehicle parameters using stationary driving
maneuvers. Control Eng. Pract., Elsevier 17(12), 1426–1431 (2009)
555. Wiggins, S.: Introduction to Applied Nonlinear Dynamical Systems and Chaos. Series: Texts
in Applied Mathematics, vol. 2. Springer, New York (2003)
556. White, A., Choi, J., Zhu, G.: Dynamic output feedback gain scheduling of an electric variable
valve timing system. In: 2013 American Control Conference, Washington DC, USA, June
2013
557. Winkler, F.J., Krause, I., Lohmann, B.: Flatness-based control of a continuous furnace. In:
18th International Conference on Control Applications, Part of 2009 IEEE Multi-Conference
on Systems and Control, Saint Petersburg, Russia, July 2009
558. Winkler, F.J., Lohmann, B.: Design of a decoupling controller structure for first order hy-
perbolic PDEs with distributed control action. In: 2010 American Control Conference, Bal-
timore, MD, USA, July 2010
559. Woittennek, F., Rudolph, J.: Controller canonical forms and flatness-based state feedback for
1D hyperbolic systems. In: 7th Vienna International Conference on Mathematical Modelling,
MATHMOD (2012)
560. Woitteneck, F., Mounier, H.: Controllability of networks of spatially one-dimensional second
order pdes—an algebraic approach. SIAM J. Control Optim. 48(6), 3882–3902 (2010)
561. Wu, S.L., Chen. P.C., Hsu, C.H., Chang, K.Y.: Gain-scheduled control of PVTOL aircraft
dynamics with parameter-dependent disturbance. J. Franklin Inst., Elsevier 345(8), 906–925
(2008)
562. Wu, F., Zhang, X.P., Ju, P., Sterling, M.J.H.: Decentralized nonlinear control of wind turbine
with doubly-fed induction generator. IEEE Trans. Power Syst. 23(2), 613–621 (2008)
563. Wu, S., Zhou, L., Yang, J.: Experimental study of an adaptive extended Kalman Filter for
structural damage identification. In: 4th International Conference on Earthquake Engineer-
ing, Taipei, Taiwan, Oct 2006
564. Wu, H.N., Wang, J.W., Li, H.K.: Design of distributed H∞ fuzzy controllers with constraint
for nonlinear hyperbolic PDE systems. Automatica, Elsevier 48, 2535–2543 (2012)
565. Wu, D., Chen, K.: Design and analysis of precision active disturbance rejection control for
nonsingular turning process. IEEE Trans. Industr. Electron. 56(7), 2746–2753 (2009)
566. Wu, B., Lang, Y., Zargari, N., Kouro, S.: Power Conversion and Control of Wind Energy
Systems. Wiley, NJ (2011)
567. Wu, C.S., Wang, H.L., Zhang, Y.M.: Numerical analysis of the temperature profiles and
weld dimension in high power direct-diode laser welding. Comput. Mater. Sci., Elsevier 46,
49–56 (2009)
568. Wu, Z.G., Shi, P., Su, H., Chu, J.: Sampled-data fuzzy control of chaotic systems based on
T-S Fuzzy model. IEEE Trans. Fuzzy Syst. 22(1), 153–163 (2014)
728 References

569. Xia, Y., Shang, J., Chen, J., Liu, G.P.: Networked Data Fusion with packet losses and variable
delays. IEEE Trans. Syst. Man Cybern. B Cybern. 39, 1107–1119 (2009)
570. Xia, Y., Zhu, Z., Fu, M., Wang, S.: Attitude tracking of rigid spacecraft with bounded
disturbances. IEEE Trans. Industr. Electron. 58(2), 647–659 (2011)
571. Xie, Z., Feng, J., Li, Z.: A multi-user chaotic communication scheme based on feedback
square root Unscented Kalman Filter. Int. J. Nonlinear Sci. Numer. Simul. 11, 1059–1068
(2010)
572. Xin, X., Tanaka, S., She, J., Yamasaki, T.: New analytical results of energy-based swing-up
control for the Pendubot. Int. J. Non-Linear Mech., Elsevier 52, 110–118 (2013)
573. Xiong, J.: An Introduction to Stochastic Filtering Theory. Oxford University Press, Oxford
(2008)
574. Xiong, K., Zhang, H.Y., Chan, C.W.: Performance evaluation of UKF-based nonlinear fil-
tering. Automatica, Elsevier 42(2), 261–270 (2006)
575. Xu, H., Hu, J., He, Y.: Operation of wind-turbine-driven DFIG systems under distorted
grid voltage conditions: analysis and experimental validations. IEEE Trans. Power Electron.
27(5), 2354–2366 (2012)
576. Xu, X., Hu, H.Y., Wang, H.L.: Stability switches, Hopf bifurcations and chaos of a neuron
model with delay dependent parameters. Phys. Lett. A, Elsevier 354(1–2), 126–136 (2006)
577. Xue, Y., Tai, N.: System frequency regulation in doubly fed induction generators. Electr.
Power Energy Syst., Elsevier 43(1), 977–983 (2012)
578. Yadaiah, N., Venkata Ramara, N.: Linearization of multi-machine power system: modelling
and control—a survey. Electr. Power Energy Syst., Elsevier 29(4), 297–311 (2007)
579. Yamashita, M., Fujimori, K., Hayakaw, K., Kimura, H.: Application of H∞ control to active
suspension systems. Automatica, Elsevier 30(11), 1717–1729 (1994)
580. Yang, S.K.: An experiment of state estimation for predictive maintenance using Kalman
Filter on a DC motor. Reliab. Eng. Syst. Saf., Elsevier 75(1), 103–111 (2002)
581. Yang, Y., Zhou, C., Jia, X.: Robust adaptive fuzzy control and its application to ship roll
stabilization. Inf. Sci., Elsevier 142(1–4), 177–194 (2002)
582. Yang, S., Ajjarapu, V.: A speed-adaptive reduced-order observer for sensorless vector control
of doubly-fed induction generator-based variable-speed wind turbines. IEEE Trans. Energy
Convers. 25(3), 891–900 (2010)
583. Yang, Q., Jagannathan, S.: A suite of robust controllers for the manipulation of microscale
objects. IEEE Trans. Syst. Man Cybern. B Cybern. 38(1), 113–125 (2008)
584. Yang, X.S., Huang, Y.: Complex Dynamics in Simple Hopfield Networks. AIP Chaos 16
(2006)
585. Yang, S., Lei, Q., Peng, F.Z., Qian, Z.: A robust control scheme for grid-connected voltage-
source inverters. IEEE Trans. Industr. Electron. 58(1), 202–212 (2011)
586. Yao, J., Li, H., Chen, Z., Xia, X., Li, X.Q., Liao, Y.: Enhanced control of a DFIG-based
wind-power generation system with series grid-side converter under unbalanced grid voltage
conditions. IEEE Trans. Power Electron. 28(7), 3167–3181 (2013)
587. Yao, J., Li, H., Chen, Z., Xia, X.: Enhanced control of a DFIG-based wind-power generation
system with series grid-side converter under unbalanced grid voltage conditions. IEEE Trans.
Power Electron. 28(7), 3167–3181 (2013)
588. Yildiz, Y., Annaswamy, A., Yanakiev, D., Kolmanovsky, I.: Adaptive air fuel ratio control for
internal combustion engines. In: Proceedings of American Control Conference, pp. 2058–
2063 (2008)
589. Yoon, Y., Shin, J., Kim, H.J., Park, Y., Sastry, S.: Model-predictive active steering and obstacle
avoidance for autonomous ground vehicles. Control Eng. Pract., Elsevier 17(7), 741–750
(2009)
590. Yoshimoto, Y., Watanabe, K., Iwatani, Y., Hashimoto, K.: Multi-camera visual servoing of a
micro helicopter under occlusions. In: Fung, R.-F. (ed.) Visual Servoing. InTech Publications
(2010)
591. Yousef, H.A., Hamdy, M., Shafiq, M.: Flatness-based adaptive fuzzy output tracking exci-
tation control for power system generators. J. Franklin Inst., Elsevier 350(8), 2334–2353
(2013)
References 729

592. Yu, D., Chakravotry, S.: A randomly perturbed iterative proper orthogonal decomposition
technique for filtering distributed parameter systems. In: American Control Conference,
Montreal, Canada, June 2012
593. Yue, H., Li, J.: Output-feedback adaptive fuzzy control for a class of nonlinear time-varying
delay systems with unknown control directions. IET Control Theory Appl. 6, 1266–1280
(2012)
594. Yoshimoto, Y., Watanabe, K., Iwatani, Y., Hashimoto, K.: Multi-camera visual servoing of a
micro helicopter under occlusions. In: Fung, R.-F. (ed.) Visual Servoing. InTech Publications
(2010)
595. Zamani, M.A., Sidha, T.S., Yazdani, A.: Investigations into the control and protection of
existing distribution network to operate as a microgrid: a case study. IEEE Trans. Industr.
Electron. 61(4), 1904–1915 (2014)
596. Zarei, J., Poshtan, J., Poshtan, M.: Robust fault detection of nonlinear systems with unknown
disturbances. In: 2010 IEEE International Conference on Control Applications (Part of 2010
IEEE Multi-Conference on Systems and Control), Yokohama, Japan, Sept 2010
597. Zhang, C., Franch, J., Agrawal, S.K.: Differentially flat design of a closed-chain planar
underactuated 2-DOF system. IEEE Trans. Robot. 29(1), 277–282 (2013)
598. Zhang, J., Shen, T., Marino, R.: Model-based cold-start speed control scheme for spark
ignition engines. Control Eng. Pract., Elsevier 18(11), 1285–1294 (2010)
599. Zhang, Z., Sun, Z.: Rotational angle based pressure control of a common rail fuel injec-
tion system for internal combustion engines. In: 2009 American Control Conference Hyatt
Regency Riverfront. St. Louis, MO, USA, June 2009
600. Zhang, Q., Basseville, M., Benveniste, A.: Fault detection and isolation in nonlinear dynamic
systems: a combined input-output and local approach. Automatica, Elsevier 34(11), 1359–
1373 (1998)
601. Zhang, X., Khadra, A., Li, D., Yang, D.: Impulsive stability of chaotic systems represented
by Takagi-Sugeno model. Chaos, Solitons Fractals, Elsevier 41(4), 1863–1869 (2009)
602. Zhang, B., Ma, B.: Robust stabilization of underactuated surface vessels with parameter
uncertainties. In: 29th Chinese Control Conference, Beijing, China, July 2010
603. Zhang, Y., Ma, H.: Theoretical and experimental investigation of networked control for
parallel operation of inverters. IEEE Trans. Industr. Electron. 59(4), 1961–1970 (2012)
604. Zheng, G., Boutat, D., Floquet, T., Barbot, J.P.: Secure communication based on multi-input
multi-output chaotic system with large message amplitude. Chaos, Solitons Fractals, Elsevier
41(3), 1510–1517 (2009)
605. Zheng, P., Tang, W., Zhang, J.: Some novel double-scroll chaotic attractors in Hopfield
networks. Neurocomputing, Elsevier 73(10–12), 2280–2285 (2010)
606. Zhong, Q.C., Nguyen, P.L., Ma, Z., Sheng, W.: Self-synchronized synchronverters: invert-
ers without a dedicated synchronization unit. IEEE Trans. Power Electron. 29(2), 617–629
(2014)
607. Zhong, Q.C.: Robust droop controller for accurate proportional load sharing among inverters
operated in parallel. IEEE Trans. Industr. Electron. 60(4), 1281–1290 (2013)
608. Zhong, Q.C., Hornik, T.: Control of Power Inverters in Renewable Energy and Smart Grid
Integration. Wiley, NY (2013)
609. Zhong, Q.C., Weiss, G.: Static synchronous generators for distributed generation and renew-
able energy. In: Proceedings of IEEE PES Systems Control Conference Exhibition, pp. 1–6
(2009)
610. Zhong, Q.C., Weiss, G.: Synchronverters: inverters that mimic synchronous generators. IEEE
Trans. Industr. Electron. 58(4), 1259–1267 (2011)
611. Zhou, L., Wu, S., Yang, J.N.: Experimental study of an adaptive extended Kalman filter for
structural damage identification. ASCE J. Infrastruct. Syst. 14(Special Issue: New Sensors,
Instrumentation, and Signal Interpretation), 4251 (2008)
612. Zhou, Z., Wang, C., Liu, Y., Holland, P.M., Igic, P.: Load current observer based feed-forward
DC bus voltage control for active rectifiers. Electr. Power Syst. Res., Elsevier 84(1), 165–173
(2012)
730 References

613. Zhu, G., Lévine, J., Praly, L., Peter, Y.A.: Flatness-based control of electrostatically actuated
MEMS with application to adaptive optics: a simulation study. IEEE J. Microelectromech.
Syst. 15(5), 1165–1174 (2006)
614. Zhu, G., Packirisamy, M., Hosseini, M., Peter, Y.A.: Modelling and control of an elec-
trostatically actuated torsional micromirror. J. Micromech. Microeng., Institute of Physics
Publishing 16(10), 2044–2052 (2006)
615. Zhu, G., Saydy, L., Hosseini, M., Chiannetta, J.F., Peter, Y.A.: A robustness apporach for
handling modelling errors in parallel-plate electrostatic MEMS control. J. Microelectromech.
Syst. 17(6), 1902 (2007)
616. Zhu, G., Agudelo, G.G., Saydy, L., Packirisamy, M.: Torque multiplication and singularity
avoidance in the control of electrostatic torsional micro-mirrors. In: Proceedings of 17th
IFAC World Congress, Seoul, Korea, July 2008
617. Zhu, Z., Leung, H.: Combined demodulation with adaptive blind-channel equalization for
chaotic-modulation communication systems. IEEE Trans. Circ. Syst. I 49(12), 1811–1820
(2002)
618. Zuazua, E.: Propagation, observation and control of waves approximated by finite difference
methods. SIAM Rev. 47(2), 197–243 (2005)
619. Zwart, H., Le Gorrec, Y., Moschke, B.: Linking hyperbolic and parabolic PDEs. 2011 50th
IEEE Conference on Decision and Control and European Control Conference, CDC-ECC,
Orlando, Florida, USA, Dec 2011
Index

Symbols B
H∞ Kalman Filtering, 299 Backstepping control, 672, 679
H∞ criterion, 106 Barbalat’s Lemma, 131, 212, 591
H∞ tracking performance, 104, 438, 542 Bendixson’s theorem, 21
χ 2 distribution, 662 Bifurcations of fixed points, 21
χ 2 test, 648 Bit Error Rate, 599
0-flat system, 83 Boundary control of nonlinear PDEs, 672,
687, 693
Brunovsky canonical form, 80, 118, 154,
A 165, 403, 406
Active control of suspensions, 240 Brunovsky-form, 57
Adaptation scheme in MIMO control, 541
Adaptive fuzzy control of chaotic systems,
579 C
Adaptive fuzzy control of diesel engines, Canonical form of robotic unicycles, 247
492, 528, 529 Canonical (normal) form, 32
Adaptive fuzzy control of MEMS, 429 Cart-pole balancing, 111
Adaptive fuzzy control of SI engines, 492, Central Limit Theorem, 648
560 Channel distortion, 602
Air–fuel ratio measurement delays, 574 Channel equalization, 603
ARMAX description of the Kalman Filter, Chaos, 2
646 Chaotic carrier signal, 599
Asymptotic stability, 10 Chaotic communication system, 579, 598
Asynchronous motor dynamics, 411 Chaotic modulation, 579
Attenuation coefficient, 111, 131 Chaotic modulator, 598
Autonomous 4-wheeled ground vehicles, Chen chaotic system, 585
240 Closed-chain underactuated robots, 218
Autonomous dynamical system, 58 Codistribution, 29
Autonomous ground vehicle, 242 Communication delays, 177
Autonomous hovercraft, 62 Compensation of delays and packet losses,
Autonomous Underwater Vessels, 239 230
Autonomous vehicle localization, 242 Compensation of measurement delays, 493

© Springer International Publishing Switzerland 2015 731

G.G. Rigatos, Nonlinear Control and Filtering Using Differential
Flatness Approaches, Studies in Systems, Decision and Control 25,
DOI 10.1007/978-3-319-16420-5
732 Index

Completely integrable, 30 Differential flatness of nonlinear wave PDE,

Controllability condition, 40 621, 690, 691
Control of wave-type PDEs, 613 Differential flatness of PMSG, 337, 347
Coriolis and centrifugal forces matrix, 123 Differential flatness of quadropters, 240
Coupled nonlinear oscillators, 616, 657 Differential flatness of quadrotor, 313
Covector field, 26 Differential flatness of robotic unicycle, 245
Cramer–Rao Lower Bound, 158 Differential flatness of the air–fuel ratio sys-
tem, 570
Differential flatness of the DC motor, 154
D Differential flatness of the nonlinear heat
Decoupling, 44 PDE, 100, 634
Defect, 87 Differential flatness of the n-storey building,
Derivative-free distributed nonlinear 657, 659
Kalman Filter, 173, 185, 227, 239, Differential flatness of the spark-ignited en-
262, 263, 271 gine, 492, 551
Derivative-free nonlinear Kalman Filter, Differential flatness of vehicle’s suspension,
141, 158, 242, 329, 350, 370, 443, 240, 296
491, 579, 600, 613, 624, 637, 661
Differential flatness of voltage source con-
Derivative-free nonlinear Kalman Filter for
verters, 449, 451
DFIG, 337
Differentially flat system, 48, 49
Derivative-free nonlinear Kalman Filter for
Differentially flat underactuated robot, 196
distributed PMSGs, 338
Diffiety, 71, 88
Derivative-free nonlinear Kalman Filter for
Discrete-time Kalman filter, 144
PMSG, 337
Distributed filtering for AGVs, 266
Derivative-free nonlinear Kalman Filter for
Distributed filtering for cooperating vehi-
UGVs, 247
cles, 262, 263
Derivative-free nonlinear Kalman Filter for
unicycle, 253 Distributed filtering over communication
Derivative-free nonlinear Kalman Filtering, network, 230, 238
219, 284, 300, 316 Distributed nonlinear filtering, 141
Detection of parametric changes in PDE, 653 Distributed nonlinear Kalman Filtering, 173
DFIG control in cascading loops, 381 Distributed parameter systems, 47, 96, 613
Diffeomorphism, 28 Distributed Particle Filter, 142
Differential flatness, 671 Distributed state estimation-based control,
Differential flatness for the heat PDE, 101 184
Differential flatness of 4-wheel vehicle, 240, Distribution, 28
276, 280 Disturbance estimator, 221
Differential flatness of air–fuel ratio system, Disturbance observer, 184, 219
493, 567 Dual Kalman Filtering, 603
Differential flatness of DFIG, 337, 367, 378 Duffing chaotic system, 579, 584
Differential flatness of Diesel engines, 491, Duffing oscillator, 5
521 Dynamic extension, 45
Differential flatness of distributed inverters, Dynamic extension for air–fuel ratio system,
444, 480 571
Differential flatness of distributed PMSGs, Dynamic extension for diesel engines, 517
338, 393 Dynamic extension for underactuated vessel,
Differential flatness of gas exchange valve, 324
491, 501 Dynamic feedback linearization, 45, 47, 61,
Differential flatness of hovercraft, 241 573
Differential flatness of implicit systems, 90 Dynamic feedback linearization for diesel
Differential flatness of inverter, 466 engines, 518
Differential flatness of MEMS, 404, 427 Dynamics of 4-wheel vehicle, 277
Differential flatness of nonlinear DC motor, Dynamics of DFIG, 360
406 Dynamics of diesel engine, 512
Index 733

Dynamics of distributed inverters, 479 Flatness-based control in cascading loops,

Dynamics of electrostatic microactuator, 403, 414, 672, 694
424 Flatness-based control of 1-DOF robot, 680
Dynamics of gas exchange valves, 497 Flatness-based control of 4-wheel vehicle,
Dynamics of interconnected PMSGs, 390 283
Dynamics of inverter, 459 Flatness-based control of air–fuel ratio, 493,
Dynamics of n-storey building, 657 572
Dynamics of PMSG, 341 Flatness-based control of chaotic dynamics,
Dynamics of quadrotor, 313 579, 591
Dynamics of SI engine, 548 Flatness-based control of converter, 452
Dynamics of the air–fuel ratio system, 569 Flatness-based control of cooperating vehi-
Dynamics of voltage source converters, 446 cles, 261, 265
Flatness-based control of DFIG, 369
Flatness-based control of diesel engines,
E 491, 524
Endogenous feedback, 78 Flatness-based control of distributed invert-
Equivalence between inverters and synchro- ers, 481, 483
nous generators, 477 Flatness-based control of distributed
Equivalent model, 51 PMSGs, 396, 398
Equivalent systems, 59 Flatness-based control of gas exchange
Estimation of communication channel coef- valve, 491
ficients, 609 Flatness-based control of gas exchange
Euler-Lagrange equation, 111, 311 valves, 504
Euler-Lagrange method, 187, 190 Flatness-based control of inverter, 443, 470
Extended Information Filter, 141, 173, 185, Flatness-based control of PMSG, 682
262, 271 Flatness-based control of SI engines, 492
Extended Kalman Filter, 142, 145, 601 Flatness-based control of spark-ignited en-
Extended Kalman Filter for DFIG, 383 gines, 552
Extended Kalman Filtering for induction Flatness-based control of underactuated ves-
motors, 416 sel, 329, 331
Extended State Observer, 224 Flat output, 50, 671, 692
Flat subsystem, 86
Fréchet topology, 58
F Frobenius theorem, 29
Fault diagnosis in distributed parameter sys- Fusion of distributed estimates, 235
tems, 614, 640
Fault diagnosis in sensors network, 652
Fault isolation test, 646 G
Fault threshold, 614 Gas metal arc welding process, 613
Feedback control of nonlinear heat PDE, 636 Generalized Likelihood Ratio, 649
Feedback control of the nonlinear wave Genesio -Tesi chaotic system, 584
PDE, 615, 623 Global asymptotic stability, 10
Field-oriented induction motor, 411 Global change detection test, 646
Filtering for nonlinear wave PDE, 623 Gradient algorithm, 110, 210
Filtering for vehicle localization, 248
Filtering in chaotic communications, 601
Filtering of wave-type PDEs, 613 H
Finite escape time, 1 Hamiltonian matrix, 132, 213
Fisher Information matrix, 172, 173 Heat diffusion equation, 96
Flatness-based adaptive control of MEMS, H-infinity performance index, 111, 131, 564,
403 590
Flatness-based adaptive fuzzy control, 103, Hopf bifurcations, 21, 24, 26
403, 407 Hurwitz stable system, 44
Flatness-based control, 47, 244 Hyperbolic equilibrium point, 12, 24
734 Index

Hyper-regular matrices, 90 Kernel functions, 126, 129, 209, 536

Hypothesis testing, 648

L
I Lagrangian, 189, 311
Implicit control system, 88 Lie algebra-based observer for nonlinear
Industrial robotic manipulators, 183 systems, 343
Inertia matrix, 123 Lie-Backlünd condition, 672
Information Filter, 141 Lie-Backlünd equivalence, 89
Information Matrix, 174 Lie-Backlünd isomorphism, 57, 71
Information state vector, 174 Lie-Bakclünd transformation, 71
Input–output linearization, 30, 34, 245 Lie Bracket, 27, 118
Input–output linearization for underactuated Lie derivative, 26, 117, 342
vessel, 324, 328 Limit cycles, 2, 8, 19, 21
Input–output linearization of converters, 447 Linear canonical (Brunovsky) form, 195,
Input–output linearization of DFIG, 363 428, 451, 471, 480, 503, 520, 552,
Input–output linearization of distributed 585, 614, 621, 672, 674
PMSGs, 391, 393 Linear DC motor, 152
Input–output linearization of gas exchange Linearizing transformation for nonlinear
valves, 498, 500 PDEs, 643
Input–output linearization of inverter, 463 Linear Matrix Inequalities, 591
Input–output linearization of MEMS, 425 Linear regressor models, 645
Input–output linearization of PMSG, 346
Linearizing transformation for nonlinear
Input–output linearization of the SI engine,
PDEs, 643
549, 550
Liouvillian systems, 86
Input-state linearizable, 40
Local linearization, 9, 591
Input-to-state linearization, 38, 39
Local linearization of Lorenz oscillator, 593
Integrity monitoring of sensors, 240
Local statistical approach to fault diagnosis,
Inverters, 443
614, 643
Involutive distribution, 28
Lorenz chaotic system, 579, 582
Involutivity condition, 30, 40
Lyapunov function, 127, 207, 678, 696
Isoclines, 6
Lyapunov stability analysis, 104, 109, 183,
403, 404, 434, 493, 538, 562, 579,
J 588
Jacobian, 146 Lyapunov stability criterion, 10
Jacobian matrix, 10, 12, 18, 19, 141
Jacobi’s identity, 41
Josephson junction, 617 M
Josephson transmission line, 618 Manifold of jets, 71, 88
Mean Square Error (MSE), 158, 599
MIMO canonical form, 119
K MIMO differentially flat systems, 47
Kalman Filter-based control, 184 MIMO nonlinear dynamics, 116
Kalman Filter-based disturbance estimator, Minimization problem in optimal control,
219, 222, 525 687
Kalman Filter-based disturbance observer, Minimum mean squares estimation problem,
240, 241, 285, 301, 317, 331, 338, 144
351, 372, 397, 443, 453, 472, 483, Min–max test for fault isolation, 650
574, 665 Multi-DOF building, 614
Kalman Filter-based disturbances estimator, Multi-DOF mechanical structure, 614
492, 505 Multi-DOF robotic manipulators, 183
Kalman Filter for delayed measurements, Multipath propagation, 602
232 Multiple isolated equilibria, 1
Index 735

N S
Neuro-fuzzy approximators, 107, 108, 127, Saddle-node bifurcation, 22
198, 432, 536, 561, 587 Saddle point, 13
Nonautonomous dynamical system, 58 Sensitivity matrix, 648
Nonflat systems, 86 Sensitivity test for fault isolation, 650
Nonlinear DC motor, 152, 225, 403, 405, 673 Sensor fusion for AGVs, 266
Nonlinear electric power generation sys- Sensorless control of DFIG, 338, 383
tems, 672 Sensorless control of DFIG in cascading
Nonlinear heat diffusion PDE, 613, 631 loops, 338, 380
Nonlinear partial differential equations, 96, Sensorless control of induction motor, 403
613 Sensorless control of PMSG, 337
Nonlinear PDEs control in cascading loops, Sensors fusion for vehicle localization, 265
688, 694 Sensors integrity monitoring for AGVs, 267
Nonlinear wave PDE, 616, 688 Sigma-Point Kalman Filters, 147
Normal coordinates, 35 Signal to Noise Ratio, 606
Normal form, 34 Sine-Gordon equation, 616
Normalized error square, 662 Sine-Gordon nonlinear wave PDE, 625
Normal states, 35 Single-input differentially flat system, 105
Smoothing Kalman Filter, 232
Spring–damper–mass system, 52
O Stabilizing control input, 34
Observation gain matrix, 202 Stabilizing feedback control, 43
Observer-based adaptive control, 404 State estimation-based control, 158, 168
Observer-based adaptive fuzzy control, 183, State estimation-based feedback control,
199, 431, 492 672
Observer canonical form, 227 State observer, 202
Optimal control, 686 State-space description of the nonlinear heat
Optimal fault threshold selection, 643 PDE, 97, 633
Optimal trajectories, 687 State-space model of Josephson junction
PDE, 619, 621
State-space model of nonlinear wave PDE,
622, 644, 689
P
State transition matrix, 231
Packet drops, 177
Static feedback linearization, 47, 192
PDE of Josephson junction dynamics, 618
Static state feedback, 83
Perturbation observer, 224
Statistical processing of residuals, 614, 643
Phase diagram, 8, 13
Strong closedness, 88
Pitchfork bifurcations, 22
Structural condition monitoring, 614
Poincaré-Bendixson’s theorem, 21
Sub-harmonic and harmonic oscillation, 2
Point of singularity, 74
Synchronization of distributed inverters,
Pointwise control of PDEs, 613
444, 480, 483
Synchronization of distributed PMSGs, 396
System in the explicit form, 88
R Systems in triangular form, 84, 671, 674
Recursive computation of control inputs, 677 Systems without drift term, 84
Regulation problem, 106, 560, 586
Relative degree, 27, 343
Residuals, 614, 646 T
Riccati equations, 104, 109, 143, 207, 213, Takagi-Sugeno fuzzy modelling, 591
435, 538, 563, 589 Takagi-Sugeno model stability condition,
Rigid-link robotic manipulator, 123, 133, 592
163 Taylor series expansion, 145
Robotic visual servoing, 184 Trajectory planning, 72
Robustness of a flatness-based controller, 82 Trajectory tracking, 78
736 Index

Transformation into the canonical form, 673 Unscented Kalman Filter, 147, 403, 602
Transmitter-receiver synchronization, 580, Unscented Transform, 148
603

V
U
Van der Pol oscillator, 19
Uncented Kalman Filtering for induction
Vector field, 26
motors, 416
Virtual synchronous generators, 477
Underactuated nonlinear robotic manipula-
Visual servoing, 178, 225, 236
tors, 183, 191
Voltage source converters, 443
Underactuated robot control, 198
VTOL aircraft, 55, 60
Underactuated robotic mechanism, 185
Underactuated vessel dynamics, 62, 241,
323
Unicycle robots, 239 W
Unknown Input Observer, 224 Weights adaptation law, 437
Unmanned Aerial Vehicles, 239
Unmanned Ground Vehicles, 239
Unmanned Surface Vessels, 239 Z
Unscented Information Filter, 142 Zero dynamics, 77