1 Output Controllability

LTI system in state space

ẋ(t) = Ax(t) + Bu(t),
y(t) = Cx(t)
Problem: Given x(0) = 0 and any x̄, compute u(t) such that y(t̄) = ȳ for
some t̄ > 0.

Solution: We know that

Z t̄
ȳ = y(t̄) = CeA(t̄−τ ) Bu(τ )dτ.
0

If we limit our search for solutions u in the form

T
(t̄−t)
u(t) = B T eA C T z̄
we have
Z t̄
T
ȳ = CeA(t̄−τ ) BB T eA (t̄−τ )
C T z̄dτ,
0
Z t
T
= CX(t)C T z̄, X(t) := eAξ BB T eA ξ dξ
0

and
¢−1 T
(t̄−t)
¢−1
z̄ = CX(t)C T u(t) = B T eA C T CX(t)C T
¡ ¡
x̄, ⇒ ȳ
As before, singularity of CX(t)C T is equivalent to the existance of
z 6= 0, z ∗ CeAτ B ≡ 0, ∀ 0 ≤ τ ≤ t.
which implies
di
¯
∗ Aτ ∗ i Aτ ¯
= z ∗ CAi B = 0,
¯ ¯
i
(i! z Ce B) ¯ = z CA e B τ =0
i = 0, . . . , n − 1.
dτ ¯
τ =0

This is equivalent to the matrix

CB CAB CA2 B . . . CAn−1 B = CC(A, B)
£ ¤

having full-row rank.

MAE 280A 1 Maurı́cio de Oliveira

1.1 Summary on Output Controllability
Theorem: The following are equivalent
1) The triplet (A, B, C) is output controllable;
2) The matrix CC(A, B) has full-row rank;
3) The matrix CX(t)C T is positive definite for some t ≥ 0.

WARNING: If C has full-row rank X(t) > 0 ⇒ CX(t)C T > 0. But there
might be cases when CX(t)C T > 0 and X(t) ≥ 0!

MAE 280A 2 Maurı́cio de Oliveira

2 A complete example: satelite in circular orbit

u2
u1
m
r










 θ


















PSfrag replacements

Earth
Satelite of mass m with thrust in the radial direction u1 and in the tangential
direction u2 . From Skelton, DSC, p. 101.

Newton’s law
m~r¨ = ~u1 + ~u2 + f~g ,
where f~g is the gravitational force
km ~r
f~g = − 2 .
r r

Using cylindrical coordinates

µ ¶ µ ¶
cos θ − sin θ
~e1 = , ~e2 = ,
sin θ cos θ
we have
km
~r = r~e1 , ~u1 = u1~e1 , ~u2 = u2~e2 , f~g = − 2 ~e1 .
r

MAE 280A 3 Maurı́cio de Oliveira

We need to compute
d2 d
~r = 2 (r~e1 ) = (ṙ~e1 + r~e˙ 1 ) = r̈~e1 + 2ṙ~e˙ 1 + r~e¨1 ,
¨
dt dt
where
µ ¶
− sin θ
~e˙ 1 = θ̇ = θ̇~e2 ,
cos θ
µ ¶ µ ¶
− sin θ − cos θ
~e¨1 = θ̈ + θ̇2 = θ̈~e2 − θ̇2~e1 .
cos θ − sin θ
That is
~r¨ = (r̈ − r θ̇2 )~e1 + (2ṙθ̇ + rθ̈)~e2 ,
so that Newton’s law can be rewritten as
km
m(r̈ − rθ̇2 )~e1 + m(2ṙθ̇ + rθ̈)~e2 = u1~e1 + u2~u2 − e~1 ,
r2
or, equivalently, as the two scalar differential equations
km
m(r̈ − rθ̇2 ) = u1 − ,
r2
m(2ṙθ̇ + rθ̈) = u2 .
In state space
ṙ
       
r x3 0 0 µ ¶
θ θ̇  x 4
  0 0  u1
x= ṙ  ,

r̈  = x1 x24 − k/x21  + 1/m
ẋ =     
 u2 .

0
θ̇ θ̈ −2x3 x4 /x1 0 1/(mx1 )

This is a nonlinear system and we look for equilibrium (r̈ = θ̈ = 0) when

u1 = u2 = 0. This can be stated as
x1 x24 − k/x21 = 0, −2x3 x4 /x1 = 0.
The second condition implies x3 = ṙ and/or x4 = θ̇ must be zero. We choose
x3 = ṙ = 0 which implies x1 = r = r̄ constant and
s r
k k
x4 = θ̇ = 3 = 3
= ω̄ ⇒ k = r̄ 3 ω̄ 2 .
x1 r̄
Note also that x2 = θ = ω̄t.

MAE 280A 4 Maurı́cio de Oliveira

This nonlinear system is in the form
ẋ = f (x, t) + g(x)u.
We will linearize f (x, t) and g(x)u around the equilibrium point (x̄, ū) to obtain
the linearized system
ẋ = (∇fx )T [x(t) − x̄(t)] + (∇gx )T [x(t) − x̄(t)]ū + g(x̄)u.
For this problem
     
r̄ x3 0 0
ω̄t x4  , g(x) =  0 0
   
x̄(t) = 
0
 , ū = 0, f (x, t) = 
x1 x24 − k/x21  1/m
,
0 
ω̄ −2x3 x4 /x1 0 1/(mx1 )
and
   
0 0 1 0 0 0 1 0
0 0 0 1   0 0 0 1 
(∇fx )T = 
  
2k/r̄3 + ω̄ 2 = .
0 0 2r̄ω̄  3ω̄ 2 0 0 2r̄ω̄ 
0 0 −2ω̄/r̄ 0 0 0 −2ω̄/r̄ 0
This produces the linearized system
   
0 0 1 0 0 0 µ ¶
 0 0 0 1   0 0  u1
ẋ = 
3ω̄ 2 0
x +   .
0 2r̄ω̄  1/m 0  u2
0 0 −2ω̄/r̄ 0 0 1/(mr̄)
If we looking at the satelite (from the earth) we can say that we can observe r
and θ̇ (how?), that is · ¸
1 0 0 0
y= x.
0 0 0 1
Questions:
1) Can we know the ‘state’ of the satelite is by measuring only r?
2) Can we know the ‘state’ of the satelite is by measuring only θ̇?
3) Can we know the ‘state’ of the satelite by measuring r and θ̇?
4) Can the system be controlled to remain in circular orbit using radial
thrusting (u1 ) alone?
5) Can the system be controlled using tangential thrusting (u2 ) alone?

MAE 280A 5 Maurı́cio de Oliveira

Question: Can we know the ‘state’ of the satelite by measuring only r?
£ ¤
Answer: Is the system observable when C = 1 0 0 0 ? Compute the
observability matrix
 
C
 CA 
O(A, C) =  CA2  ,


CA3
1 0 0 0
 
 0 0 1 0 
=  3ω̄ 2 0 0 2r̄ω̄ 


0 0 −ω̄ 2 0

Phsysical interpretation: measuring r does not give any information on θ or θ̄!

Note that if we that know the satelite is in equilibrium and “measure” k then
r
k
θ̇ = ω̄ = .
r̄3
But we still do not know θ since

θ(t) = θ(0) + ωt,

and we do not know θ(0)!

MAE 280A 6 Maurı́cio de Oliveira

Question: Can we know the ‘state’ of the satelite by observing θ̇ only?
£ ¤
Answer: Is the system observable when C = 0 0 0 1 ? Compute the
observability matrix
 
C
 CA 
O(A, C) = CA2  ,


CA3
0 0 0 1
 
 0 0 −2ω̄/r̄ 0 
=  −6ω̄ 3 /r̄ 0

0 −4ω̄ 2 
0 0 2ω̄ 3 /r̄ 0

Phsysical interpretation: again, if we try to reconstruct θ from θ̇ we still need to

know θ at some t̄! From that point on
Z t
θ = θ(t̄) + θ̇(τ )dτ.
t̄

MAE 280A 7 Maurı́cio de Oliveira

Question: Can we know the ‘state’ of the satelite by measuring r and θ̇?
· ¸
1 0 0 0
Answer: Is the system observable when C = ? Compute the
0 0 0 1
observability matrix
 
C
 CA 
O(A, C) = CA2  ,


CA3
 
1 0 0 0

 0 0 0 1  

 0 0 1 0  
0 0 −2ω̄/r̄ 0 
 
=

 3ω̄ 2 0 0 2r̄ω̄ 

 −6ω̄ 3 /r̄ 0 0 −4ω̄ 2 
 
0 −ω̄ 2
 
 0 0 
0 0 2ω̄ 3 /r̄ 0

Phsysical interpretation: can we know θ at all?

MAE 280A 8 Maurı́cio de Oliveira

Question: Can the system be controlled to remain in circular orbit using radial
thrusting (u1 ) alone?
 
0
 0 
Answer: Is the system controllable when B = 1/m? Compute the


0
controllability matrix

O(A, C) = B AB A2 B A3 B ,
£ ¤

−ω̄ 2
 
0 1 0
1 0 0 −2ω̄/r̄ 0 
=  
m 1 0 −ω̄ 2 0 
0 −2ω̄/r̄ 0 2ω̄ 3 /r̄
Note that
−ω̄ 2
   
1
 0   0 
−ω̄ 2 
 0 = 0 
  

−2ω̄/r̄ 2ω̄ 3 /r̄

which implies that the system is not controllable from u1 !
Phsysical interpretation: there must be a change in the angular velocity θ̇ if one
changes the radius!

MAE 280A 9 Maurı́cio de Oliveira

Question: Can the system be controlled using tangential thrusting (u2 ) alone?
 
0
 0 
Answer: Is the system controllable when B =  0 ? Compute the


1/(mr̄)
controllability matrix

O(A, C) = B AB A2 B A3 B ,
£ ¤
 
0 0 2r̄ω̄ 0
1 
0 1 0 −4ω̄ 2 
= 
3 
mr̄  0 2r̄ ω̄ 0 −2r̄ ω̄
1 0 −4ω̄ 2 0
The above matrix has full rank, so the system is controllable from u2 !
Phsysical interpretation: we can change the radius by changing the angular
velocity!

MAE 280A 10 Maurı́cio de Oliveira