EE 640 Multivariable Control Systems: Problem sheet 4

Instructor: Dwaipayan Mukherjee∗

Indian Institute of Technology Bombay, Mumbai- 400076, India

1. For what values of the parameter α, is the linear time invariant system
   
1 α 1 1
ẋ(t) = 0 1 1 x(t) + 1 u(t)
0 0 0 1
 
0 1 0
y(t) = x(t)
0 2 1

controllable? observable?
2. Consider the system given by
 
0 1 0 0
3ω02 0 0 2r0 ω0 
ẋ(t) = 
 0
 x(t),
0 0 1 
0 −2ω0 /r0 0 0
 
1 0 0 0
y(t) = x(t)
0 0 1 0

(a) Is the system observable by only y1 ?

(b) Is the system observable by only y2 ?

3. For the system given by

   
−1 1 0 0
ẋ(t) =  1 −3 1  x(t) + 1 u(t),
0 1 −1 0
 
y(t) = 0 1 0 x(t)

Obtain its minimal realization.

4. For what values of the parameter α is the following state equation minimal?

   
1 0 2 1
ẋ(t) = 0 3 0 x(t) + 1 u(t),
0 α 1 1
 
y(t) = 1 0 1 x(t).

5. Consider a linear system with dynamics given by ẋ = Ax, x(t0 ) = x0 and y = Cx. Show that the system is
observable on [t0 , tf ] if and only if the matrix
Z tf
T
M (t0 , tf ) = eA (t,t0 )
C T CeA(t,t0 ) dt
t0

is non-singular.

6. Consider the system

   
0 ω 0
ẋ(t) = x(t) + u(t)
−ω 0 1
 
y(t) = 1 1 x(t)

Evaluate its observability Gramian and comment on the observability using the notion of observability Gramian.
∗ Associate Professor, Electrical Engineering, Office: EE 214D, e-mail: dm@[Link]

1
  
A − λI
7. Show that the pair (C, A) is observable if and only if rank = n, for all λ ∈ C. (Note: The proof
C
should proceed independently, without relying on controllability or duality properties.)
8. Prove that there always exists a matrix L so that the eigenvalues of A + LC can be assigned freely if and only
if (C, A) is observable.
9. Show that there always exists a transformation T such that the pair (A, C) can be transformed to
 
−1 A11 0  
T AT = , CT = C1 0
A21 A22
where (C1 , A11 ) is observable. (Note: The proof should proceed independently, without relying on controllability
or duality properties.)
10. Show that the matrix A22 in Question. 9 is Hurwitz if and only if
 
A − λI
rank = n, ∀λ ∈ C+
C

11. Suppose a system is given by

   
−1 0 0 1 0  
1 1 0
ẋ =  0 −1 0 x + 0 1 u, y= x.
1 0 0
0 0 2 0 0
Determine the uncontrollable and the unobservable eigenvalues (if any).
12. Consider the spring mass given in Fig. 1 with the state space equations
      
ẋ1 0 1 0 0 x1 0 0  
ẋ2  −0.1910 −0.0536 0.0910 0.0036 
 x2  + 1 0 f1 ,
  
 =
ẋ3   0 0 0 1  x3  0 0  f2
ẋ4 0.0910 0.0036 −0.1910 −0.0536 x4 0 −1
 
  " # x1
y1 1 0 0 0 x2 
=  
y2 0 1 0 0 x3 
x4
Is the system observable from y? If yes, reduce (A, C) to observer form.

Figure 1: Spring Mass System

13. Consider the state-space description of the lateral motion of an aircraft: ẋ = Ax + Bu given by
      
ẋ1 −0.746 0.006 −0.999 0.03690 x1 0.0012 0.0092  
ẋ2   −12.9 −0.746 0.387 0   x2  +  6.05 0.952 
 u1 ,
  
 =
ẋ3   4.31 0.024 −0.174 0  x3  −0.416 −1.76  u2
ẋ4 0 1 0 0 x4 0 0
 T
where the control inputs u1 u2 := [δA , δR ]T denote the aileron and rudder deflections, respectively, and
the state variables in the vector x := [β, p, r, ϕ]T are the side-slip angle, roll rate, yaw rate, and roll angle,
respectively.
 
0 1 0 0
Let y = Cx with C = . Is the system observable from y? If yes, reduce (A, C) to the observer
0 0 1 0
canonical form.
14. Consider the pair (A, C) for the system given by
 
−1 1 0 0 0
0 −1 1 0 0  
  0 0 0 0 1
A= 0 0 −1 0 0 C =

0 0 1 1 1 0
0 0 0 1
0 0 0 0 0
Show that the pair (A, C) is detectable. Further, design a matrix L ∈ R5×2 such that the A + LC has all the
eigenvalues in the left-half plane.

2
 dx
15. Consider a system dt = Ax, y = Cx where
 
8 17 31 10 31
 16 18 44 −2 33   
  0 1 1 2 2
A=
−15 −21 −46 −4 −39 and C = .
 −6 −5
 3 5 9 2 7
−14 4 −7 
6 8 18 1 15

Design a matrix L such that the A + LC has all the eigenvalues in the left-half plane.
 
0 1 0  
1 1 1
16. Given A = −1 −2 1  and C = , Reduce the system to the observer form.
0 1 2
1 1 −1
17. Consider the system given in Fig. 2.
 
y2
(a) Prove that the system is observable from y1 , but not from y2 or .
y3
 
u
(b) Prove that the system is controllable from u3 , but not from u1 or 1 .
u2

Figure 2: System for problem 14

18. Consider the system given in Fig. 3. Prove that the system is always controllable, but is observable if and only
if
c20 − a1 c1 c0 + a0 c21 ̸= 0.

Figure 3: System for problem 15

19. Consider the second-order system

   
0 1 0 0
ẋ(t) =  0 0 1  x(t) + 0 u(t)
−1 −2 −6 7
 
y(t) = 5 −4 0 x(t) + [0]u(t).
Verify that the system is observable. If so, determine the observer gain matrix required to place the observer
poles at s1,2 = −2 ± j3 and s3 = −6.
20. Consider the third-order system
   
0 1 0 0
ẋ(t) =  0 0 1  x(t) +  0  u(t)
−4.3 −1.7 −6.7 0.35
 
y(t) = 0 1 0 x(t) + [0]u(t).
Determine a full-state feedback gain matrix and an observer gain matrix to place the closed-loop system poles
at s1,2 = −1.4 ± j1.4, s3 = −2, and the observer poles at s1,2 = −18 ± j5, s3 = −20.

3
21. Consider the system in state variable form

ẋ(t) = Ax(t) + Bu(t)

where    
0 1 0 0 0
0 0 1 0  0
A=
0
, B= 
0 0 1  0
−4 −7 −8 −11 1
and  
y(t) = 1 0 0 0 x(t) + [0]u(t).
Design a full-state feedback gain matrix and an observer gain matrix to place the closed-loop system poles at
s1,2 = −2 ± j2, s3,4 = −5 ± j, and the observer poles s1,2 = −9 ± j2, s3,4 = −15. Draw a complete schematic
diagram for the interconnection between your designed observer-based controller and the given system.
   
0 −1 1 1 0  
22. For the matrices A = 1 −2 1 , B = 1 1 and C = 0 1 0 , perform Kalman decomposition and
0 1 −1 1 2
comment on observable/unobservable, controllable/uncontrollable subsystems.
23. Comment on observable/unobservable, controllable/uncontrollable subsystems for the system given by
   
0 0 −1 1
ẋ = 1 0 −3 x + 1 u
0 1 −3 0
 
y = 0 1 −2 x

24. A system

ẋ(t) = Ax(t) + Bu(t), x(0) = x0

y(t) = Cx(t) + Du(t)

is said to be detectable if it does not have any state which is both unobservable and unstable. State and prove
a PBH like test for detectability.
25. Consider the state space system given by

ẋ(t) = Ax(t) + Bu(t), x(0) = x0

y(t) = Cx(t) + Du(t)

Let an observer for the above system be given by

ẇ(t) = M w(t) + N y(t) + P u(t), w(0) = w0

x̂(t) = Qw(t) + Ry(t) + Su(t)

Show that an observer exists if and only if (C, A) is detectable.

26. It is obvious (why?) that any scalar rational biproper/strictly proper transfer function, gi (s), has some state
space realization given by (Ai , B i , Ci , Di ). Consider a multi-output single-input system having a transfer
g1 (s)
g2 (s)
function given by GO (s) :=  . . Similarly, also consider a single-output multi-input system having transfer
 
 .. 
gp (s)
 
function GI (s) := g1 (s) g2 (s) . . . gm (s) . Show that the state space realizations of GO (s) and GI (s) may
    
B1 D1
  B2  D2 
be given by diag(A1 , A2 , . . . , Ap ),  .  , diag(C1 , C2 , . . . , Cp ),  .  and
    
  ..   .. 
Bp D
  p 

diag(A1 , A2 , . . . , Am ), diag(B1 , B2 , . . . , Bm ), C1 C2 ... Cm , D1 D2 ...
Dm , respectively.
 
1 2
27. In accordance with Question. 26 realize the transfer function matrix given by G1 (s) = in the
s + 1 s + 1 
1
s 0
state space form. Further, extend the idea in Question 26 to realize the transfer functions G2 (s) =  1 1
s2 s

4
  
1
0
and G3 (s) =  1 1s  in the state space form. Determine which of the transfer functions G1 (s), G2 (s), and
 

s2 s
G3 (s) is/are have a minimal realization, following the method outlined in Question 26. For those that have not
been realized minimally, suggest a possible minimal realization.
 
−2 1.9
28. Consider a system given by ẋ = Ax; x(0− ) = x0 , where A = . Is the matrix A Hurwitz? Consider
0.9 −1
the Lyapunov candidate given by V (x) = x21 + 5x22 . Evaluate V̇ (x) along the trajectories of the above system.
What can you say about the stability of the above system in the sense of Lyapunov?
 
− 0 −1
29. Consider a system given by ẋ = Ax; x(0 ) = x0 , where A = . Consider the Lyapunov candidate
1 −1
 √ 
5+ 5 −2√
given by V (x) = x⊤ P x, where P = . Is V (x) positive definite? What about V̇ (x) along the
−2 5− 5
trajectories of the system? Can you infer anything about the stability of the system with the chosen V (x)?
30. (MATLAB based exercise) Write a code in MATLAB to achieve the goals stated below, when your code asks
the user to enter the following information: (1) number of states, (2) number of outputs, (3) the matrices A
and C, and (4) the desired eigenvalues of the observer.

(a) Check whether the (A, C) entered by the user is observable or not. Display this for the user. If not, ask
the user to re-enter a different pair.
(b) Display the observability indices for the entered data.
(c) Using the ‘search by columns’ (dual to the row-search in case of controllability) on the observability crate,
obtain a suitable similarity transformation that converts the entered (A, C) pair to their observer canonical
form, (Ao , Bo ). Display the (Ao , Bo ) pair.
(d) Obtain a suitable R, such that RCo is in row reduced echelon form. Display the matrix R.
(e) Choose a suitable L̂ such that Ao + L̂RCo has eigenvalues at the desired locations entered by the user.
Display L̂.
(f) Display the matrix L, corresponding to your choices of R and L̂ such that A + LC has its eigenvalues at
the desired locations.
[Note that you need to demonstrate each functionality of your code by choosing a random (A, C) pair. Present
the result of your code on the pair as your responses to this question.]