Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

2005-01-0417

SAE TECHNICAL
PAPER SERIES

Implementing the Milliken Moment Method

using Controlled Dynamic Simulation
Dongsoo Kang and Jeffrey L. Stein
University of Michigan

Robert C. Hoffman
ArvinMeritor, Inc., USA

Loucas S. Louca
University of Cyprus, Cyprus

Kunsoo Huh
Hanyang University, Korea

Reprinted From: Vehicle Dynamics and Simulation 2005

(SP-1916)

2005 SAE World Congress

Detroit, Michigan
April 11-14, 2005

400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel: (724) 776-4841 Fax: (724) 776-5760 Web: www.sae.org
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

The Engineering Meetings Board has approved this paper for publication. It has successfully completed
SAE’s peer review process under the supervision of the session organizer. This process requires a
minimum of three (3) reviews by industry experts.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of SAE.

For permission and licensing requests contact:

SAE Permissions
400 Commonwealth Drive
Warrendale, PA 15096-0001-USA
Email: permissions@sae.org
Tel: 724-772-4028
Fax: 724-772-4891

For multiple print copies contact:

SAE Customer Service

Tel: 877-606-7323 (inside USA and Canada)
Tel: 724-776-4970 (outside USA)
Fax: 724-776-1615
Email: CustomerService@sae.org

ISBN 0-7680-1561-8
Copyright © 2005 SAE International

Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE.
The author is solely responsible for the content of the paper. A process is available by which discussions
will be printed with the paper if it is published in SAE Transactions.

Persons wishing to submit papers to be considered for presentation or publication by SAE should send the
manuscript or a 300 word abstract to Secretary, Engineering Meetings Board, SAE.

Printed in USA
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

2005-01-0417

Implementing the Milliken Moment Method using Controlled

Dynamic Simulation
Dongsoo Kang and Jeffrey L. Stein
University of Michigan

Robert C. Hoffman
ArvinMeritor, Inc., USA

Loucas S. Louca
University of Cyprus, Cyprus

Kunsoo Huh
Hanyang University, Korea

Copyright © 2005 SAE International

ABSTRACT IZ – Yaw moment of inertia (kg-m^2)

W – Vehicle weight (N)
The Milliken Moment Method (MMM) can be VX – Vehicle longitudinal velocity (m/sec)
used to quantify the constraints imposed on vehicle VY – Vehicle lateral velocity (m/sec)
stability and controllability by front and rear tire traction r – Vehicle yaw velocity (rad/sec)
limitations. The main aspect of the Milliken Moment l – Wheel base (m)
Method is the plot of vehicle’s yaw moment versus a – Distance from front axle to center of gravity (m)
lateral acceleration for given vehicle sideslip and b – Distance from rear axle to center of gravity (m)
steering angle ranges. This plot is typically called the β – Vehicle sideslip angle (degree)
Milliken Moment Diagram (MMD). This paper proposes a α – Tire slip angle (degree)
dynamic simulation approach to implementing the MMM δ – Front wheel steering angle (degree)
that emulates the traditional experimental one. The FX – longitudinal resultant forces (N)
approach embeds a vehicle dynamics model in a control FY – Lateral resultant forces (N)
loop that maintains a constant desired sideslip angle, N – Resultant yaw moment (N-m)
and integrates the resulting controlled vehicle system Maero – Externally applied aero yaw moment (N-m)
model in time to generate the MMD. A numerical AY – Vehicle lateral acceleration (g’s)
example demonstrates the approach and shows that the N
while the proposed approach may not be as efficient as CN – Non-dimensional yaw moment coefficient ( ≡ )
the traditional computer based approach, it has the Wl
advantage of being able to used a variety of existing
vehicle dynamics models with commonly available INTRODUCTION
numerical simulation packages in order to generate the
MMD’s. Stability and controllability are among a vehicle’s
most important handling characteristics. Whether a
NOMENCLATURE vehicle properly follows severe control inputs and rejects
disturbances in extreme maneuvers depends in part on
MMM – The Milliken Moment Method its stability and controllability margins. Therefore, one
MMD – The Milliken Moment Diagram (CN-AY diagram should quantify these two properties both when
created by Milliken Moment Method) assessing vehicle handling performance and when
SSAB – Steering and Sideslip Angle to Boundary metric generating vehicle designs.
in [1] The Milliken Moment Method (MMM) is an
DOF – Degree of freedom efficient rendering of vehicle stability and controllability
m – Vehicle mass (kg) limitations. It generates the Milliken Moment Diagram
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

(MMD) also known as the CN-AY diagram, where a CONSTRAINED TESTING

vehicle’s yaw moment coefficient (CN) versus its lateral
acceleration (AY) is mapped for different vehicle sideslip
(β) and steering wheel angles (δ). Tire saturation limits
bound that portion of the MMD where the vehicle is
stable and controllable. Milliken et al. [2-4] originally
developed the MMM and the MMD as a compact
portrayal of the results of experimental vehicle handling
performance tests. They subsequently developed
computer software that plots MMD’s and uses them to
analyze directional stability by quantifying steady-state
vehicle yawing moments [8]. Shibahata et al. [5] used
similar reasoning to create the β-method, where the
stabilizing yaw moment and the resultant side force at
the center of gravity are plotted individually versus the
vehicle sideslip angle β. Hoffman et al. [1] quantified a
vehicle’s stability and controllability over the course of a
maneuver in terms of the closest distances between the
maneuver’s closed-loop dynamic simulation data and
the MMD tire saturation boundaries. This approach
predicted vehicle instability during extreme maneuvers
more accurately than the phase plane method [6], but
two different vehicle models were used to generate the Figure 1: Illustration of constrained testing
MMD and closed-loop simulation data. Discrepancies
between models used to simulate vehicle maneuvers The MMM is based on the assumption that most
and models used to create MMD’s can degrade the vehicle handling properties can be obtained from the
accuracy of stability and controllability margins obtained steady-state forces and moments on the vehicle. These
by comparing the maneuvers and MMD’s. forces and moments are not only those associated with
In this paper, a new method for implementing equilibrium conditions (e.g., steady-state cornering) but
the MMM is presented. This method uses the same also those unbalanced conditions, where forces and
vehicle model to generate both the MMD and vehicle moments are accelerating the vehicle linearly and
maneuver response, thereby overcoming the model angularly. The unbalanced forces and moments
consistency problem. Furthermore, the proposed conditions are available in constrained testing shown in
method enables users to study the influence of nonlinear Figure 1. Analogous to wind tunnel testing of airplanes,
and linear features of existing vehicle models on stability the vehicle forces and moments can be measured by
and controllability of a vehicle. The proposed method constraints (tethers), which balance the tire forces and
embeds nonlinear vehicle dynamics models within a moments that are not in equilibrium. The vehicle is
control loop that modulates the vehicle yaw moment to attached to the platform with three tethers and it is held
maintain a desired sideslip angle. This method
in a configuration with sideslip angle (β) with steering
simulates the constrained experimental testing, where
angle (δ) at a constant forward speed (V). The tethers
tethers apply an external yaw moment to maintain a
hold the vehicle from moving forward and enable the
desired sideslip angle (see Background section).
user to measure unbalanced forces. This test is
In the remainder of this paper, we describe the
repeated for a range of conditions in order to generate
constrained experimental tests and the diagram’s
the data needed for the MMD.
general features. We then proceed to its proposed use
by Hoffman et al. [1] to quantify vehicle stability and
GENERAL FEATURES OF MMD’S
controllability over the course of a maneuver. Finally, we
describe the proposed MMD generation algorithm in
detail and demonstrate it using a numerical example. In The MMD is a way of presenting the results from
conclusion, we show that the proposed method a constrained test on a 2-D plane relating both the yaw
eliminates the aforementioned model consistency moment coefficient and lateral acceleration as shown in
problem by allowing the user to select from the vehicle Figure 2. Given a constant forward velocity, the
dynamics model of their choice and can be used for unbalanced lateral force and yaw moment coefficient are
vehicle parameter design using commonly available mapped according to both steering angle and vehicle
numerical simulation software. sideslip angle. This diagram is composed of two sets of
constant value lines: β-line and δ-line. The limits of this
diagram are generated by front and rear tire saturation,
BACKGROUND where each tire will not produce additional lateral forces
for additional tire slip angle. The limits seen in quadrant
This section presents relevant background of 1 and 3 / quadrant 2 and 4 show the front and rear tire
the original MMM and the recently developed “Steering saturation limit, respectively. These properties led to a
Sideslip Angle to Boundary” (SSAB) metric [1]. vision of comparing the MMD with closed-loop vehicle
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

simulation data in assessing stability and controllability as a distance between the current vehicle state and
of a vehicle. boundaries on the MMD.
If a vehicle is assumed to be driving smoothly
without rapid steering changes, state values on the
MMD can be compared with a driving trajectory obtained
from real vehicle motion. This leads to the mapping of a
dynamic trajectory from closed-loop simulation onto the
MMD.
As the vehicle is assumed to move along the
trajectory on the MMD (see Figure 3.), it uses up more
or less available traction, as indicated by how close it is
to each tire saturation boundary. Hoffman et al. [1]
proposed that this distance be in degrees of steering to
the front tire boundary, and degrees of vehicle sideslip to
the rear tire boundary for measuring stability and
controllability safety margins. At any given point of the
trajectory, the stability safety margin can be represented
by the distance to the rear tire boundary along a
Figure 2: Milliken Moment Diagram constant steering angle, which in the MMD is measured
in degrees of vehicle sideslip. Similarly, the controllability
In addition, steady-state properties are shown safety margin can be represented by the distance to the
on the dotted line in Figure 2 – where the yaw moment front tire boundary along a constant sideslip line,
coefficient is zero. Understeer coefficient is represented measured in degrees of steering (see Figure 4.).
by the variation of steering angle to lateral acceleration
along this line. The maximum steady-state lateral
acceleration that a vehicle can produce at this forward
velocity is easily shown by the maximum x-value on the
steady-state line. The stability index which represents
directional stability is shown as the change in yaw
moment coefficient with respect to lateral acceleration at
each steady-state point. The extreme limits in the x-
direction (not necessarily on the y-axis) encountered by
front and rear saturation limit can enable an engineer to
discern whether vehicle has a limit oversteer or
understeer behavior. If the CN has a negative sign at the
maximum lateral acceleration, then the vehicle exhibits
limit understeer.

Table 1: General MMD properties Figure 3: The MMD and dynamic trajectory
Property Expression Remarks/Units
Understeer [∂δ ] Ax =const degree/g
coefficient ∂AY
Stability ∂C N
[ ]δ =const -/g
Index ∂β
Front tire Quadrant 1 and 3
Controllability
Saturation limit boundary
Rear tire Quadrant 2 and 4
Stability
Saturation limit boundary
Limit Sign of CN at the
Negative
Understeer maximum lateral
(Positive)
(Oversteer) acceleration

SSAB METRIC Figure 4: Boundary description

This section presents an extension of the MMD As for the SSAB metric, stability and
to evaluate handling performance. Hoffman et al. [1] controllability safety margins are expressed by how
combined the MMD with closed-loop dynamic simulation close the closed-loop trajectories are to the MMD
and measured both stability and controllability properties boundaries. This metric was shown to be a very effective
and intuitive measurement of vehicle performance; much
more so than phase-plane stability analysis. In a severe
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

but stable maneuvering at a 100km/h speed, the lane and Simulink are used for constructing the nonlinear
change trajectory goes outside the stable boundary of dynamic model and control structure.
the 3-D phase plane as shown in the upper one of figure
5, whereas it stays inside the stable boundary of the
MMD. It is believed that the phase plane stability
boundary represents the open-loop tendency of the
vehicle but it is not guaranteed when the driver closes
the loop.

Figure 6: Schematic block diagram

Generally, this procedure is composed of three

control/dynamic blocks and two post-processing blocks
as shown in Figure 6. “Delta/Beta Generator” generates
desired steering and sideslip angles and increments
them upon convergence. “Controller” regulates a sideslip
angle to a desired value (Beta_ref) using an externally
applied yaw moment (Mz_aero). “Nonlinear Vehicle
0.5
Big Car - Bicycle at 100 km/hr Dynamics” adopts a vehicle model from any vehicle
0.4
simulation program as a dynamic link library (DLL) file.
0.3
The converged CN-AY data at every step is stored and
0.2 4o
5 o6
7o89ooo mapped with corresponding steering and sideslip angle
2o
3o
in two post-processing blocks.
Yaw Moment Coeff.

0.1
o o o
o o 1o
2.5
-9.5 o

oo
4.54 3 .5 3 2 1.5 o o
oo o 1 o ooooo
9. 5 o

oo 5
.56.55.5
ooo6 δ0.=5 0o 0 oo-o.5 .5
8977.5 β = o oo 6 6-798.5
--7
0 -0.5 o oo - 5-5- .5
-4.5

Table 2: Function of control/dynamic blocks

o o
-1 o -1
-1.5 o o .5-4
-2 -2.5 -3 -3
- 2o
-3 o
-0.1 -4 o
-6oo-5o o

Control/dynamic
-7
-9
-8

-0.2
Function
-0.3
blocks
-0.4 Delta/Beta Setup and increment steering and
-0.5
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Generator sideslip angle
Lateral Accel, g
PID controller: focusing on settling
Controller
Figure 5: Severe lane change trajectory comparison time and steady-state error
between 3-D phase-plane and MMD Vehicle CarSim dynamic model
Dynamics (any dynamic model)
However, Hoffman et al. [1] used two different Post-processing Store and mapping CN-AY data
vehicle models in generating closed-loop trajectories
and the MMD. Therefore, the model consistency Specifically, the role of tethers in the general
problem still remained. Nonetheless, this metric was a constrained testing model is replaced by an externally
valuable first step in developing a way to combine the applied yaw moment (See Figure 7). Since we do not
MMD and closed-loop dynamic simulation. focus on transient but steady-state values, the controller
To ensure a valid and useful implementation of can be designed as a simple PID controller with the
the SSAB metric, the same vehicle model should be gains chosen to minimize the steady-state error and the
used for generating closed-loop trajectories and the time to reach this minimum error. Through this feedback
MMD. The next section will describe a new method for control loop, a vehicle’s unbalanced force/moment
implementing the MMM using existing dynamic vehicle condition can be regarded as a steady-state equilibrium
models such as CarSim. condition, mimicking a real constrained test. This
constrained moment has a reverse direction but the
SIMULATING CONSTRAINED TESTING same magnitude as the generated vehicle yaw moment.

This section describes a new procedure

developed to generate the MMD by using commercial
vehicle simulation and control software. Nonlinear
dynamics models are embedded in a structure
implemented in a control software (see Figure 6). For a
given set of steering and sideslip angle, the system is
numerically integrated until it has settled to a steady-
state, mimicking a constrained test. The stored steady-
state CN-AY data at every steering and sideslip angle is Figure 7: Externally applied yaw moment (Aero moment,
then used to compose the MMD. In this paper, CarSim Mz_aero)
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

In Figure 8 an MMD is shown for a passenger vehicle. the proposed procedures is a way to ensure accurate
The steering wheel angle was varied from -320 to 320 and intuitive results that best represent reality.
degree by 16 degree step and sideslip angle was from -
9 to 9 degree by 0.1 degree step. The forward velocity of Table 3: Example vehicle parameter
the vehicle is constant at 70 km/h.
Property Value Units
Sprung Mass 1527 kg
Front Unsprung
100 kg
Mass
Rear Unsprung
80 kg
Mass
CG Height 0.542 m
Wheelbase 2.69 m
a 1.014 m
Front Roll Center
0.07 m
Height
Front Tire Stiffness 220 N/mm

Rear Tire Stiffness 220 N/mm

Track Width 1.54 m

Front Stabar
Figure 8: The Milliken Moment Diagram 384 N-m/deg
Stiffness
Rear Stabar
This procedure simulates the effect of the 344 N-m/deg
Stiffness
constrained physical experiment. As mentioned before, Steering Ratio 16 deg/deg
using a nonlinear dynamic model to create the MMD
makes it possible to predict a vehicle’s handling
performance a straight forward, convenient manner. In
addition, it can be a flexible way to analyze stability and
controllability of a vehicle by mapping dynamic
trajectories on the MMD as discussed previously. In the
next section, this benefit will be shown with examples.

THE EFFECTS OF MODEL COMPLEXITY

This section shows one of the benefits of this

procedure by comparing two MMD’s - one was
constructed with a 3-DOF dynamic model described in
the Appendix, and the other was based on the nonlinear
dynamics model in CarSim, with 14 multi-body degrees
of freedom and nonlinear kinematics. To represent more
nonlinear effects, the 3-DOF model is extended to a four Figure 9. Comparison between 3-DOF and CarSim
wheel bicycle model, which includes lateral load transfer model
and a nonlinear Pacejka [7] tire model. The same tire
model was also used in CarSim for comparing the two
diagrams. In Figure 8, the blue MMD represents the 3-
DOF model as implemented in Matlab and the red MMD
is based on a CarSim model generated by the proposed
procedure. The parameters for this comparison
represent the “Big Car: Baseline” vehicle with limit-
understeer characteristics and the values are given in
Table 3.
As seen in Figure 9, the two MMD’s are similar
in shape but have different boundaries representing front
and rear tire saturation. Figure 10 shows the
enlargement of the circled area shown in Figure 9. The
two models predict very different limit behaviors with the
3-DOF predicting oversteer and CarSim understeer.
The simple 3-DOF provides potentially misleading or
non-intuitive results at the boundary. Therefore, using a Figure 10. Enlargement at maximum lateral acceleration
more detailed model for generating the MMD through
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

COMBINATION WITH DYNAMIC TRAJECTORY dynamic simulation, the stability and controllability
values determined based on the CarSim MMD is
In this section, a dynamic trajectory from CarSim assumed to be more accurate.
is combined with two MMD’s generated in the previous
section. The maneuver is a single lane change at 70 DISCUSSION
km/h based on the 1975 ISO double lane change. The
dynamic simulations for this maneuver provide the time This procedure enables an engineer to use a
histories of CN and AY. This trajectory of the maneuver existing (and potentially commercial simulation software)
can be easily obtained from CarSim by setting up the vehicle dynamic model to generate the MMD’s using
proper controller and trajectory variables. Although the closed-loop dynamic simulation. And while the authors’
forward velocity of the vehicle is not precisely constant, it acknowledge that the proposed procedure adds the
is assumed to be a constant (the speed at the end of the computational burden associated with evaluating the
maneuver is 0.3% lower). The reason is that this dynamic models as opposed to steady-state models for
trajectory was simulated with a closed-loop forward generating MMD’s, they believe the flexibility and
speed controller. The dynamic trajectory is overlaid on convenience afforded by their method is attractive.
the MMD as shown in Figure 11. The benefits inherent in this approach to
generate MMD’s could be used for many purposes,
including parameter sensitivity studies, optimal design,
and active chassis controls. Many options for carrying
out a realistic, practical application to vehicle dynamics
and design are feasible because the designer can
conveniently include as many nonlinear parameters and
degrees of freedom as deemed necessary. Model
consistency between a vehicle’s MMD and dynamic
simulation ensures that results are consistent and
lessens the possibility for misinterpretation.
The role of externally applied yaw moment might
be extended to constrain roll moment as well. If a
constrained roll moment coefficient were used with this
procedure, rollover stability problems would be
considered together with a varying yaw moment. In
active rollover controller design, the effect of additional
Figure 11. Mapping dynamic trajectory onto the MMD roll moment would be explained with varying yaw and
lateral acceleration.
There are also several benefits for using this
procedure to evaluate stability and controllability related
with closed-loop trajectories. It is possible to not only
make use of the power of a given commercial dynamic
software but also to guarantee model consistency
between closed-loop simulation and the MMD,
combining the best of both worlds.

SUMMARY AND CONCLUSION

This paper presents a new procedure for

implementing the MMM using commercially available
dynamic simulation and control software programs.
Existing dynamic vehicle models can be used to
Figure 12. Different stability and controllability values generate the steady-state values required to draw an
MMD. A feedback controller through an applied yaw
As seen in Figure 12, two different MMD’s can moment is used to maintain the vehicle at a given
cause a significant difference predicted stability and sideslip angle. The new approach also eliminates any
controllability values. Looking at constant sideslip inconsistencies for calculating the stability and
(dashed line β=0.3), the controllability value based on the controllability metric since the same model is used to
3-DOF MMD (distance to the front tire saturation along calculate the MMD and the dynamic simulation.
β=0.3 degree line) is larger than the distance from the In conclusion, this procedure makes the
boundary of the CarSim model. Similarly the stability generation of the MMD accurate and easy and provides
measurement (distance to the rear tire saturation a flexible and convenient way to combine the MMD with
boundary along δ=48 degree line) is also different for the dynamic simulation.
two models. Because the MMD produced with the
CarSim model is consistent with the model used for
 Downloaded from SAE International by Zhejiang University, Tuesday, November 12, 2024

ACKNOWLEDGMENTS 7. E Bakker, H.B. Pacejka and L. Lidner, 1989, “A new

tire model with an application in vehicle dynamics
The authors would like to acknowledge the studies,” in Proc. Int. Congress and Exposition,
generous technical and financial support of the Detroit, MI, SAE paper 890087.
DaimlerChrysler Challenge Fund, as well as the 8. Milliken Moment Method Software, Milliken
technical support of the Automotive Research Center Research Associates, Inc., Buffalo, NY, 2003.
(ARC) which is a U.S. Army Center of Excellence for 9. CarSim, Version 5.16b, Mechanical Simulation
Automotive Research at the University of Michigan. Corporation, Ann Arbor, MI, 2004.

REFERENCES CONTACT

1. Robert C. Hoffman, Jeffrey L. Stein, Loucas S. Dongsoo Kang

Louca and Kunsoo Huh, 2004, “Using the Milliken AML W. E. Lay Automotive Lab Univ. of Michigan
Moment Method and Dynamic Simulation to 1231 Beal Ave Ann Arbor MI 48109-2133
Evaluate Vehicle Stability and Controllability,” Tel: 734-763-7388, Email: dskang@umich.edu
Proceedings of ASME IMECE 2004.
2. Milliken, W.F., Wright, P.G. and Milliken, D.L., 1994, APPENDIX
“Moment Method – A Comprehensive Tool for Race
Car Development,” SAE Technical Paper 942538, 3-DOF bicycle model dynamic equation
Society of Automotive Engineers, Inc., Warrendale,
PA. •
3. Milliken, W.F. and Milliken, D.L., 1995, Race Car m(V x − V y r ) = Fxf cos δ f + Fxr − Fyf sin δ f
Vehicle Dynamics, Society of Automotive Engineers,
Inc. Warrendale, PA. •
4. Milliken, W.F., Dell’Amico, F. and Rice, R.S., 1976, m(V y + V x r ) = Fyr + Fyf cos δ f + Fxf sin δ f
“The Static Directional Stability and Control of the
Automobile,” SAE Technical Paper 760712, •
Automobile Engineering Meeting, Society of I z r = aFyf cos δ f − bFyr + aFxf sin δ f
Automotive Engineers, Inc. Warrendale, PA.
5. Y. Shibahata, K. Shimada and T. Tomari, 1993,
Tire slip angle and motion variables
“Improvement of Vehicle Maneuverability by Direct
Yaw Moment Control,” Society of Automative
Engineers of Japan, Vehicle System Dynamics, 22, ar + V y br − V y
αf =δf − , αr =
pp. 465-481. Vx Vx
6. Shoji Inagaki, Ikuo Kshiro and Masaki Yamamoto,
1994, “Analysis on Vehicle Stability in Critical
Pacejka Magic tire formula
Cornering Using Phase-Plane Method,” Proceedings
of AVEC’94, pp. 412-420. −1 −1
F (α ) = D ⋅ sin{C ⋅ tan ( B ⋅ α − E ⋅ ( B ⋅ α − tan ( B ⋅ α )))} + S v