Controllability examples
Dr. Matthew Turner
January 24, 2018
Linear System
u(t) y(t)
G
U(s) Y(s)
Consider the following linear time invariant (LTI) system
ẋ(t) = Ax(t) + Bu(t)
x(0) = x0
y(t) = Cx(t) + Du(t)
where x(t) is the state vector, u(t) the input vector and y(t) the output vector.
A system is completely controllable, roughly speaking, if the input vector u(t) influences all of the states.
The study of a system’s controllability only involves the state equation and, therefore, the matrices A
and B. Formal definitions of controllability have been given previously. The aim of this work-sheet is to
illustrate controllability and the related concept of stabilisability through some examples.
Example 1: Consider a state-space system in which
0 1 0
A= B=
0 3 1
Is this system completely controllable?
In this case the state equation becomes:
ẋ1 0 1 x1 0
= + u
ẋ2 0 3 x2 1
This can be written long-hand as
ẋ1 = x2
ẋ2 = 3x2 + u
Because of the particularly simple form of these equations, they can be solved explicitly to obtain
Z t
x1 (t) = x1 (0) + x2 (τ )dτ
0
Z t
3t
x2 (t) = e x2 (0) + e3(t−τ ) u(τ )dτ
0
1
From the above equations we can see that
• u directly influences the evolution of x2
• u does not directly influence the evolution of x1 , but because x1 depends on x2 , x1 therefore depends
on u. Thus the system is completely controllable.
EXERCISE: Verify the above conclusion by checking the rank of the controllability matrix C = [B AB]
Example 2: Consider a state-space system in which
−1 0 0
A= B=
0 3 1
This is similar to the first example except that the first row of the A-matrix has changed a little. These
matrices lead to the state equation:
ẋ1 −1 0 x1 0
= + u
ẋ2 0 3 x2 1
This can be written long-hand as
ẋ1 = −x1
ẋ2 = 3x2 + u
Because of the particularly simple form of these equations, they can be solved explicitly to obtain
x1 (t) = e−t x1 (0)
Z t
3t
x2 (t) = e x2 (0) + e3(t−τ ) u(τ )dτ
0
From the above equations we can see that
• u directly influences the evolution of x2 (as with Example 1)
• The evolution of x1 is not affected by either the input u or the other state x2 . Whatever the
input does, it has no influence over x1 , so this state is not controllable. Hence the system is not
completely controllable
However, note that
• The state x2 is controllable
• The state x1 , though uncontrollable, is stable because, whatever the initial condition, the state
always converges to zero.
• Hence in this case, although the system is not completely controllable it is stabilisable because all
of its uncontrollable modes are stable.
EXERCISE: Verify the above conclusion by checking the rank of the controllability matrix C = [B AB].
In this case, we would expect C to lose rank because one state is not controllable.
2
Example 3: Consider a state-space system in which
1 0 0
A= B=
0 3 1
This is similar to the second example except that the sign of the first row of the A-matrix has changed.
These matrices lead to the state equation:
ẋ1 1 0 x1 0
= + u
ẋ2 0 3 x2 1
This can be written long-hand as
ẋ1 = x1
ẋ2 = 3x2 + u
Because of the particularly simple form of these equations, they can be solved explicitly to obtain
x1 (t) = et x1 (0)
Z t
3t
x2 (t) = e x2 (0) + e3(t−τ ) u(τ )dτ
0
From the above equations we can see that
• u directly influences the evolution of x2 (as with Examples 1 and 2)
• The evolution of x1 is not affected by either the input u or the other state x2 . Whatever the
input does, it has no influence over x1 , so this state is not controllable. Hence the system is not
completely controllable (as with Example 2)
However, note that in this case - different to Example 2 -
• The state x1 is both uncontrollable and not stable - it does not converge to zero and actually
diverges.
• Hence in this case, the system is not completely controllable, nor is it stabilisable because its
uncontrollable modes are not all stable.
