EE650 - Basics of Modern Control Systems

Solution for Assignment -7

1. Given system:  3
− 41
  
−4 1  
ẋ(t) = x(t) + u(t), y(t) = 4 2 x(t).
− 12 − 21 1

The controllability matrix is:  

C= B AB .
Here:  3
− 14
  
1 −4
B= , A= .
1 − 12 − 12
Compute AB:  3
− 14
   
−4 1 −1
AB = = .
− 12 − 12 1 −1
Thus:  
1 −1
C= .
1 −1
The rank of C is 1 (not full rank) =⇒ the system is not completely controllable.
The observability matrix is:  
C
O= .
CA
Here:  3
− 14

  −4
C= 4 2 , A= .
− 12 − 12
Compute CA:
 −3 − 14
 
  
CA = 4 2 4 = −4 −2 .
− 21 − 12
Thus:  
4 2
O= .
−4 −2

Since the rank of O is 1 (less than the number of states, 2) =⇒ the system is not completely
observable.
   
0 1 0
2. For the system ẋ = Ax + Bu with A = and B = , we have
0 0 1

   
At 1 t At t
e = and e B = .
0 1 1
1 3 1 2

3T 2T
RT T
The reachability Gramian is Wr (0, T ) = 0
eA(T −τ ) BB T eA (T −τ )
dτ = 1 2 .
2T T
1 4
Since det Wr (0, T ) = 12 T ̸= 0 for any T > 0, rankWr (0, T ) = n and (A, B) is reachable.
At
Notice again that R(e B) = R(Wr (0, T )) for every T > 0.

1
3. The state transition matrix is
 −(t−τ ) 1 (t+τ )
− e−(t+3τ ) ]

e 2 [e
Φ(t, τ ) = −(t−τ ) .
0 e
 −t 
e
Here Φ(t, τ )B(τ ) = , and the reachability Gramian of the system is
0
Z t1  −2t
(t1 − t0 )e−2t1 0
  
e 0
Wr (t0 , t1 ) = dτ = .
t0 0 0 0 0

It is clear that rankWr (t0 , t1 ) < 2 = n for any t0 < t1 and therefore the system is not reachable at t1 .

4. Similar to questions 2 and 3

5. If x ∈ R(C), this means that there exists a vector α such that [B, AB, ..., An−1 B]α = x. Then
Ax = [AB, A2 B, ..., An B]α. In view of the Cayley-Hamilton Theorem, An can be expressed as a linear
combination of An−1 , ..., A, I, which implies that Ax = β for some appropriate vector β. Therefore,
Ax ∈ R(C).

6. (a) This system can be written as ẋ = Ax + B(t)u, with

" #
 1
 1 , t ≤ 2,


  
−1 0
A= , B(t) = " #
0 −2  1
 0 , t > 2.




The homogeneous part of the system is actually time-invariant, and therefore the state-transition matrix
is given by  −(t−t ) 
e 0
0
Φ(t, t0 ) = eA(t−t0 ) = , ∀t, t0 ≥ 0.
0 e−2(t−t0 )

(b) On the interval [0, 1], we have

Z 1
Wc (0, 1) = Φ(0, τ )B(τ )B(τ )T Φ(0, τ )T dτ
0

Z 1  τ  
 eτ
 
e 0 1  0
= 1 1 dτ
0 0 e2τ 1 0 e2τ
" 2 #
1 e −1 e3 −1
 2τ
e3τ
Z 
e 2 3
= 4τ dτ = e3 −1 e4 −1
.
0 e3τ e
3 4

On the interval [3, 4], we have

Z 4
Wc (3, 4) = Φ(3, τ )B(τ )B(τ )T Φ(3, τ )T dτ
3

Z 4  −(3−τ )  
 e−(3−τ )
 
e 0 1  0
= 1 0 dτ
3 0 e−2(3−τ ) 0 0 e−2(3−τ )
Z 4  −2(3−τ )   −2
e −1

e 0 0
= dτ = 2 .
3 0 0 0 0

(c) The system is controllable on [0, 1] because Wc (0, 1) is a nonsingular matrix:

" 2 #
e −1 e3 −1 (e2 − 1)(e4 − 1) (e3 − 1)2
det Wc (0, 1) = det e32−1 e43−1 = − ̸= 0,
3 4
8 9
 but it is not controllable on [3, 4] because
e−2 −1
 
0
rank Wc (3, 4) = rank 2 = 1 < 2.
0 0

The controllability subspace on [3, 4] is given by

e−2 −1
   
1
Im Wc (3, 4) = Im 2 = ,
0 0
 
1
and a basis for this space is the vector .
0
7. The given system is:
ẋ = Ax + Bu, y = Cx,
with:    
1 1 1 0  
A = 0 −1 0 , B = 1 , C= 1 1 0 .
0 0 1 1

(A) Controllability and Observability

The controllability matrix is:
C = B AB A2 B .
 
 
0 2 2
C = 1 −1 1 .
1 1 1
Rank of C = 3, so the system is controllable.
The observability matrix is:  
C
O =  CA  .
CA2
=⇒  
1 1 0
O = 1 0 1 .
1 1 2
Rank of O = 3, so the system is observable.

(B) State-Feedback Controller Design

Transfer Function Matrix:
 
s−1 −1 −1
∆(s) =  0 s+1 0 
0 0 s−1

Characteristic Equation:
(s − 1)(s + 1)(s − 1) = 0
The poles are at -1, 1, 1. The system is not stable.
Desired Characteristic Equation:
(s + 2)(s + 2)(s + 2) = s3 + 6s2 + 12s + 8
Gain Matrix K:  
−1 −1 1
W = −1 1 0
1 0 0
  
0 2 0
T = Q−1
c W =  1 −2 1
−1 0 1
 
1 1 −1
1
T −1 = 1 0 0
2
1 1 1
 −1
 1 1 −1
1
8 − 1 T −1 =⇒ 7
  
K = 6 − (−1) 12 − (−1) 13 7 1 0 0
2
1 1 1
1   
K= 27 14 0 = 13.5 7 0
2

(C) Estimator Gain Matrix Design

Observer Design:
Desired Characteristic Equation: (s + 10)(s + 8)(s + 15) = s3 + 18s2 + 80s + 1200
Characteristic Polynomial:
ϕ(A) = Ā3 + 33Ā2 + 350Ā + 1200
where Ā = AT ; B̄ = C T
Calculation of phi(A):
 
1584 0 0
ϕ(Ā) =  351 882 0 
419 0 1584
 
1 1 1
C¯ = B̄ Ā2 B̄ . =⇒ 1
 
ĀB̄ 0 1
0 1 2
Inverse of the Matrix:
 
1 1 −1
1
C̄ −1 = 2 −2 0
2
−1 1 1

Calculation of K’:
  
 1 1 1 −1 1584 0 0
K̄ = 0 0 1 2 −2 0   351 882 0 
2
−1 1 1 419 0 1584

1
= [−814 882 1584] = [−407 441 792]
2
The observer gain matrix is L = K̄ T , i.e.,
 
−407
L = K ′ =  441 
792