AE-777 (Optimal Space Flight

Control)

Quiz No. 1 (Solution)

1. A system is governed by the following dif-
ferential equations:
1 3
ξ̇1 = ξ1 + ξ2 + η1
3
1
ξ̇2 = − ξ22 − η2
2
where ξ1(t) ∈ R and ξ2(t) ∈ R are the state
variables, and η1(t) ∈ R and η2(t) ∈ R are
the control inputs. The output equations
of the system are the following:
1 2
ζ1 = ξ 1 + ξ2
2
ζ2 = −ξ1 + η2
where ζ1(t) ∈ R and ζ2(t) ∈ R are the out-
put variables. Linearize the system about
the reference solution, ξ1 = 1, ξ2 = 0, and
determine the state-space coefficient ma-
trices, A, B, C, D of the linearized system.
Solution:
The state equations can be expressed in
the following vector form:

ξ̇ = f (ξ, η) (1)
where ξ = (ξ1, ξ2)T , and
 
1 3
 3 ξ 1 + ξ2 + η 1

 


f (ξ, η) = (2)
−1 2−η
 
ξ
 
2
 
2 2
To linearize the state equations about the
reference solution, ξr = (1, 0)T and the
corresponding reference input, ηr = (−1/3, 0),
we carry out their following truncated Tay-
lor series expansion about the reference so-
lution:
ẋ = Ax + Bu , (3)
where
   
 x1 
   ξ1 − 1 
 
x= = ξ − ξr = (4)

 x   
2
  ξ2 
and u = η − ηr ∈ R2, with A, B being the
following Jacobian matrices:
2
 
ξ 1
∂f  1
A = (ξr , ηr ) = 


∂ξ 0 −ξ2 ξ1 =1,ξ2 =0
 
1 1
=  (5)
 

0 0
 
1 0
∂f
B= (ξr , ηr ) =  (6)
 

∂η 0 −1
The output equation of the system can be
expressed in the following vector form:

ζ = h(ξ, η) (7)
where ζ = (ζ1, ζ2)T ∈ R2 is the output vec-
tor, and
1
 
2
 2 ξ1 + ξ 2 
 
h(ξ, η) = (8)
 

−ξ1 + η2 
The following is the truncated Taylor series
expansion of the output equation about the
reference output, ζr = (1/2, −1):

y = Cx + Du , (9)
where y = ζ − ζr ∈ R2, x = ξ − ξr and
u = η − ηr , with C, D being the following
Jacobian matrices:
 
ξ 1
∂h
C = (ξr , ηr ) = 
 

∂ξ −1 0 ξ1 =1,ξ2 =0
 
1 1
=  (10)
 

−1 0
 
0 0
∂h
D= (ξr , ηr ) =  (11)
 

∂η 0 1
2. A linear system has the following state tran-
sition matrix:
!
1−t −t
Φ(t, 0) =
t 1+t
What is the state dynamics matrix, A, of
this system?

Solution:
One of the properties of a state-transition
matrix is that it satisfies its own homoge-
neous state equation (see Eq.(26) of Lec-
ture 2):
dΦ(t, 0)
= A(t)Φ(t, 0) . (12)
dt
Another property of a state-transition ma-
trix is that its inverse can be obtained by
reversing the time (see Eq.(24) of Lecture
2):
Φ(t, 0)−1 = Φ(0, −t) (13)
Here we have
!
dΦ(t, 0) −1 −1
= (14)
dt 1 1
and
!
1+t t
Φ−1(t, 0) = Φ(0, −t) =
−t 1−t
(15)
Substituting Eqs.(14) and (15) into Eq.(12),
we get
dΦ(t, 0)
A(t) = Φ(t, 0)−1
dt ! !
−1 −1 1+t t
=
1 1 −t 1−t
!
−1 − t + t −t − 1 + t
=
1+t−t t+1−t
!
−1 −1
= (16)
1 1
Hence, the state-dynamics matrix is time-
invariant, given by
!
−1 −1
A= (17)
1 1
3. A system has the following state equations:

ẋ1 = x1 − x2
ẋ2 = u
where x1(t) ∈ R and x2(t) ∈ R are the state
variables, and u(t) ∈ R is the control input.

(a) If the initial condition at t = 0 is x1(0) =

1, x2(0) = −1 and the input is u(t) = 0
for t < 0, u(t) = 1 for t ≥ 0, solve the
state equations for t ≥ 0.

Solution:
We express the state equations in the
following vector form:

ẋ = Ax + Bu , (18)
where
   
1 −1 0
A= , B= (19)
   

0 0 1
with x(t) = [x1(t), x2(t)]T ∈ R2 being
the state vector, and u(t) ∈ R the con-
trol input.
The general solution of an LTI system
to an arbitrary, Laplace transformable
input, which begins to act at time t = 0
when the system’s state is x(0) = x0, is
given by (see Lecture 2)
Z t
x(t) = eAtx0 + eA(t−τ )Bu(τ )dτ . (20)
0
Here, it is given that the unit step input,
u(t) = 0 for t < 0, u(t) = 1 for t ≥ 0, is
applied at t = 0, with initial condition,
x0 = (1, −1)T .
We first calculate the state-transition
matrix using the partial fraction expan-
sion (Lecture 2) as follows:
 −1
s−1 1
eAt = L−1(sI − A)−1 = L−1 
 

0 s
 
s −1
1
= L−1
 
s(s − 1) 0
 
s−1
1 −1
 

−1  s−1 s(s−1) 
= L  


0 1
 s 
1 1− 1
−1  s−1 s s−1 
= L  


0 1
 s
et 1 − et


=  , (t ≥ 0) (21)
 
0 1
 Then, substituting eAt into Eq.(20), we
have the following state solution of the
linearized system:
  
et 1− et  1 
x(t) =  
0 1  −1 
  
Z t e (t−τ ) 1−e (t−τ ) 0
+    (1)dτ
0 0 1 1
   
t
 e −1+e  t (t−τ )
Z t 1−e 
= + dτ
 −1  0  1 
   t
 2et − 1  τ − et (−e−τ )
= + 
 −1  τ
  0
 t + et 
= , (t ≥ 0) (22)
 t−1 

(b) A stability analysis of the system reveals

the following characteristic equation:
 
s−1 −1
det(sI−A) = det   = s(s−1) = 0
 
0 s
(23)
 whose roots, s1,2 = 0, 1, have a positive
real part, s = 1. Hence, the system is
unstable.

(c) Controllability is analyzed using the fol-

lowing test matrix for controllability:

P = 
(B, AB) 
0 −1
=  (24)
 

1 0
whose rank equals two, the order of the
system. Hence, the system is control-
lable.

(d) The linear state-feedback control law is

given by:
u = −Kx (25)
where x = (x1, x2)T is the state vector
and K = (k1, k2) is the regulator gain
matrix. The state equations expressed
in Part (a) are the following:

ẋ = Ax + Bu , (26)
where the state coefficient matrices are
the following:
   
1 −1 0
A= , B= (27)
   

0 0 1
Substituting Eq.(25) into Eq.(26), we
have the following closed-loop state equa-
tion:
ẋ = (A − BK)x , (28)
whose characteristic equation is given
by
  
s−1 1
det(sI − A + BK) = det (29)
k1 s + k2
= s2 + (k2 − 1)s − k1 − k2 = 0

Comparing Eq.(29) with the given char-

acteristic polynomial, s2 +2s+1, we get

k1 = −4 , k2 = 3 , (30)
or K = (−4, 3).
Note: The same result as Eq.(30) is
obtained by Ackermann’s formula (see
Lecture 5):

K = (â − a)(P W )−1 (31)