Linear System Theory, Fall 2023
ME564, EE560, AERO550, CEE571
Professors Gillespie & Li, University of Michigan
Homework #9 (Includes concepts up to and including pre-recorded lecture video 37)
Total points: 100
Assigned: Monday November 20th, 2023
Due: Monday December 4th, 2023 on Canvas at 11:59pm.
There is no HW9 Quiz. Submit a file on Canvas containing your typed/scanned/code answers.
This will be worth 100 points.
Students may work together on homework assignments, but individually prepared written solutions
(no copying!) are required for submission. Consulting any textbook solutions, or old homework
solutions (either from this course or a similar course at another university) is not allowed.
You may use as much MATLAB as you want.
1. Consider the system given by
2 1 1 1
ẋ = 5 3 6 x + 0 u
−5 −1 −4 0
h i
y= 1 1 2 x
(a) [5 pts] Identify the three modes of the system. Each mode can look like v1 eλt , or
v2 eλt + v3 teλt , if there is a defective eigenvalue. It is up to you to find values vectors vi .
(b) [2 pts] For each mode, identify whether it is controllable or uncontrollable.
(c) [2 pts] For each mode, identify whether it is observable or unobservable.
(d) [2 pts] Which eigenvalue(s) will appear as poles in the transfer function YU (s)
(s)
of the
system? You should be able to do this without any calculation. Confirm your answer
by computing the transfer function of the system.
(e) [4 pts] Transform the system into modal form. That is, find a change of basis matrix V
and define a new variable z = V −1 x that allows you to rewrite the system dynamics as
ż = Jz + B̄u, y = C̄z, where J is in Jordan normal form.
2. Consider the transfer function
s+3
G(s) =
(s + 2)(s + 1)
We want to find some state-space realizations of this transfer function, i.e.
ẋ = Ax + Bu, y = Cx
such that C(sI − A)−1 B = G(s)
(a) [4 pts] Find matrices A, B, C corresponding to the controllable canonical realization.
(b) [4 pts] Find matrices A, B, C corresponding to the observable canonical realization.
(c) [4 pts] Find matrices A, B, C corresponding to a modal realization, i.e. let A be diagonal.
(d) [4 pts] Find matrices A, B, C corresponding to a non-minimal realization with A ∈ R4×4 .
3. Consider the system given by
" # " #
−2 −1 2
ẋ = x+ u
0 −3 1
h i
y = 1 −1 x
(a) [4 pts] Transform the system into modal form (see instructions for 1h). Let the trans-
formed state variable be z = V −1 x. For the rest of this problem, we will work in z
coordinates unless otherwise stated.
(b) [5 pts] Compute the zero-state and zero-input responses Lu and Lo .
(c) [5 pts] Find the adjoint system, and its zero-state and zero-input responses Pu and Po .
(d) [4 pts] Show that L∗o = Pu .
(e) [4 pts] Show that L∗u = Po .
" # " #
0 4
(f) [5 pts] Given x(0) = and x(1) = find an input u(t) that will steer the system
4 0
from x(0) to x(1). You may equivalently think of steering from z(0) to z(1)
(g) [2 pts] Find a basis in z coordinates for the unobservable subspace.
(h) [2 pts] Find a basis in x coordinates for the unobservable subspace
4. [10 pts] Find the Kalman decomposition of the following system
3 3 0 2 0
0 87 0 60 3
ẋ = x + u
6 3 −3 2 −1
0 −126 0 −87 4
5. A discrete-time system is given by
" # " #
0 1 1 0
x(k + 1) = x+ u
−1 −2 0 1
(a) [4 pts] We will use a state-feedback law u(k) = Kx(k). Find K such that the closed-loop
eigenvalues (i.e. the eigenvalues of A + BK) are both at 0.
(b) [3 pts] Repeat part a) using only the first input u1 (k) = Kx(k) if possible, or explain
why it cannot be done. Notice that K is now a row vector.
2
(c) [3 pts] Repeat part a) using only the second input u2 (k) = Kx(k) if possible, or explain
why it cannot be done. Notice that K is now a row vector.
6. Consider the linear system described by the equations
ẋ = Ax + Bu (1)
y = Cx (2)
with the linear state feedback
u = −Kx + v (3)
(a) [2 pts] Find an expression for the closed loop system, i.e. find matrices Ā and B̄ such
that ẋ = Āx + B̄v by plugging (3) into (1). Your closed-loop matrices Ā and B̄ may
depend on K.
(b) [3 pts] Show that if the original system (A, B) is controllable, then the closed-loop system
(Ā, B̄) is also controllable for any K.
(c) [3 pts] Let A ∈ R2×2 . Find an example for which the original system (A, C) is observable
but the closed-loop system Ā, C is not.
7. [10 pts] Fill out the online course evaluation. Attach a copy of the acknowledgment for full
credit on this problem.
