Scribd
Upload
2 views2 pages

ME(Math)577_HW05

The document outlines a homework set for a course on Stochastic Systems in Science and Engineering, consisting of five problems focused on Monte Carlo simulations, stochastic differential equations, covariance functions, ergodicity, and linear minimum mean-square estimation. Each problem requires analytical solutions, simulations, and comparisons of results, as well as theoretical explorations of properties of Gaussian processes. The problems emphasize the application of mathematical concepts in the context of stochastic processes and their implications in engineering.

Uploaded by

Nick
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
Download as pdf or txt
2 views2 pages

ME(Math)577_HW05

The document outlines a homework set for a course on Stochastic Systems in Science and Engineering, consisting of five problems focused on Monte Carlo simulations, stochastic differential equations, covariance functions, ergodicity, and linear minimum mean-square estimation. Each problem requires analytical solutions, simulations, and comparisons of results, as well as theoretical explorations of properties of Gaussian processes. The problems emphasize the application of mathematical concepts in the context of stochastic processes and their implications in engineering.

Uploaded by

Nick
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
Download as pdf or txt
Download as pdf or txt
You are on page 1/ 2

ME/MATH 577

Stochastic Systems in Science and Engineering


Home Work Set #05

Problem 05.01
Show, by Monte Carlo simulation of the integral of white zero-mean Gaussian noise
of constant diffusion α, that the variance of the Wiener process βt is αt. Repeat
the simulation exercise if the white noise is not of constant diffusion but it is time-
varying, say, α(t) = e−t . In that case, verify whether the variance of the non-
constant diffusion Wiener process βt is (1 − e−t ) or not.

Problem 05.02
Let a dynamical system be governed by the stochastic differential equation:
[ ] [ ] [ ] [ ]
d2 yt dyt y0 1.0 y0 0.0 0.0
+3 +2yt = wt ; given E = and Cov =
dt2 dt ẏ0 0.0 ẏ0 0.0 0.0
and wt is the zero-mean stationary white Gaussian noise of intensity 0.01.
( i) Write a state space model of the system and then find an analytical solution.
( ii) Simulate the system on a computer and compare the results of Monte Carlo
simulation with the analytical results. Explain how you have chosen the step
size of integration.

Problem 05.03
Let Xt be a WSS zero-mean Gaussian random process and the ms derivative of Xt
be denoted as Ẋt ; let Yt , Xt2 and the ms derivative of Yt be denoted as Ẏt .
( i) Show that Ẏt = 2Xt Ẋt in the ms sense.
( ii) Find the covariance function of Ẏt in terms of RXX and its derivatives.
iii) What happens if Xt is not restricted to be Gaussian?
[Hint: express the fourth moment of Xt in terms of its second moment(s).]

1
Problem 05.04
Let Xt be a zero-mean WSS Gaussian process.

( i) State the general definition of ”ergodicity in the mean” for Xt and compare
this definition with that given by Birkhoff ergodicity.
( ii) Let RXX (τ ) = σ 2 exp(−α|τ |) cos(2πf τ ), where σ 2 , α, and f are positive
constants. Show that xt is ergodic in the mean. What can we say about
Birkhoff ergodicity of Xt ?

Problem 05.05
This problem makes use of the K-L expansion to obtain the linear minimum mean-
square estimate (LMMSE) of a Gaussian signal that is contaminated with additive
white Gaussian noise. Let st be the true signal that is zero-mean with covariance
KS=ss (t1 , t2 ). Let the measured signal xt over the time interval [0, T ] be contami-
nated with zero-mean constant-intensity white Gaussian noise wt as:

xt = st + wt and s⊥w

( i) Show that any set of orthonormal functions {φ(t)} satisfies the K-L integral
∫T
equation 0 dτ Kww (t, τ )φk (τ ) = λk φk (t) for all t ∈ [0, t].
( ii) Show that the same set of orthonormal functions {φ(t)} may be used for the
K-L expansion of xt and st .
( iii) Show that, for xk = sk + wk , the LMMSE estimate of sk is given as

b k |xk ] = σs2k
E[s xk
σs2k + σw2

where xk , sk , and wk are the K-L coefficients of the respective random pro-
cesses.
( iv) Using the above relation, establish that
∞
∑ σs2k
sbt = xk φk (t) ∀t ∈ [0, T ]
σs2k + σw2
k=1

∑∞
[Hint: Expand st in a K-L expansion, st = k=1 sk φk (t)].

You might also like