ME 433 – STATE SPACE CONTROL
Lecture 4
ME 433 - State Space Control 54
State Transformation
We consider the linear, time-invariant system
We define the state transformation
Then we can write
to obtain
The state-space representation is NOT unique!
ME 433 - State Space Control 55
1
Solution of State Equation
Time-invariant Dynamics:
We consider the linear, time-invariant, homogeneous system
where A is a constant n×n matrix. The solution can be written as
Time-variant Dynamics:
We consider the linear, time-variant, homogeneous system
where A is a time-variant n×n matrix. The solution can be written as
ME 433 - State Space Control 56
Transfer Function
We consider the linear, time-invariant system
We Laplace transform the state equation to obtain
And we solve to obtain
We Laplace transform the output equation to obtain
ME 433 - State Space Control 57
2
Solution of State Equation
We inverse Laplace transform to obtain
−1
⎡ t ⎤
[ ]
where L e At = ( sI − A) ; L ⎢
⎣
∫ f (t − τ )g(τ )dτ ⎥⎦ = f (s)g(s)
0
For LTI systems, the exponential matrix eAt is the transition matrix of the
system. Its Laplace transform Φ(s)=(sI-A)-1 is called the resolvent of the
system. Faddeeva’s algorithm allows for a fast computation of Φ(s).
€
How can we compute the transition matrix eAt ?
1. Compute sI-A.
2. Obtain the resolvent Φ(s) by inverting sI-A.
3. Obtain the transition matrix by computing the inverse Laplace
transform of the resolvent, element by element.
ME 433 - State Space Control 58
Transfer Function
Assuming x(0)=0 and combining state and output equations we obtain
the transfer function
[ −1
]
Y (s) = C ( sI − A) B + D U(s)
The transfer function does NOT depend on the state choice because it
represents the input-output relationship.
€
Given two state space representations
[ −1
] [ −1
Y (s) = C˜ ( sI − A˜ ) B˜ + D˜ U(s) = C ( sI − A) B + D U(s) ]
ME 433 - State Space Control 59
3
Controllability and Observability
Let us consider the system
with
and transfer function
ME 433 - State Space Control 60
Controllability and Observability
The transfer function can be written as
What’s going on? Let’s see the system from a different angle. Consider:
with
to obtain
ME 433 - State Space Control 61
4
Controllability and Observability
with
ME 433 - State Space Control 62
Controllability and Observability
Let us assume the following modal form:
Only the controllable and observable mode appears in the transfer function
ME 433 - State Space Control 63
5
Controllability and Observability
ME 433 - State Space Control 64
Controllability
Problem Definition: “A system is said to be controllable if and only if it
is possible, by means of the input, to transfer the system from any initial
state x(0) to any other state x(t) in a finite time t≥ 0”
t
x ( t ) = e At x(0) + ∫e A ( t −τ )
Bu(τ ) dτ
0
Let us take
T
€ u(τ ) = BT e A ( t −τ )
P −1 ( t )[ x ( t ) − e At x (0)]
where
t
T
€
P(t) = ∫e A ( t −τ )
BBT e A ( t −τ )
dτ
0
is invertible.
ME 433 - State Space Control
€ 65
6
Controllability
Replacing u in the expression for x we will demonstrate that this is an
input that drives the state from x(0) to x(t).
t
x ( t ) = e At x(0) + ∫e
0
A ( t −τ )
[
B BT e A
T
( t −τ )
P −1 ( t )[ x ( t ) − e At x (0)] dτ ]
t
T
= e At x(0) + ∫e A ( t −τ )
BBT e A ( t −τ )
dτP −1 ( t )[ x ( t ) − e At x (0)]
0
= e x(0) + P ( t ) P −1 ( t )[ x ( t ) − e At x (0)]
At
= x (t )
It is indeed possible to show that this the only input that drives the state
from x(0) to x(t).
Theorem: A system is controllable if and only if the matrix
t
€ T
P(t) = ∫e A ( t −τ )
BBT e A ( t −τ )
dτ Controllability Gramian
0
is nonsingular for some t≥ 0.
Proof: In class
ME 433 - State Space Control 66
€
Controllability
The time derivative of the Controllability Gramian
t
T
P(t) = ∫e A ( t −τ )
BBT e A ( t −τ )
dτ
0
is given by (Leibniz’s rule)
d T
P ( t ) = e A ( t −τ ) BBT e A ( t −τ ) +
dt € τ =t
t t
d A ( t −τ ) d A T ( t −τ )
+∫
0 dt
[ e ]
T
BBT e A ( t −τ ) dτ + ∫e
0
A ( t −τ )
BBT
dt
e [ dτ ]
t t
d A ( t −τ ) d A T ( t −τ )
= BBT + ∫ [
0 dt
e
T
]
BBT e A ( t −τ ) dτ + ∫e
0
A ( t −τ )
BBT
dt
[
e ] dτ
Then, the Controllability Gramian satisfies
€ P˙ ( t ) = BBT + AP ( t ) + P ( t ) AT , P (0) = 0 Lyapunov Equation
ME 433 - State Space Control 67
7
Controllability
Theorem: “A system is controllable if and only if the matrix
C = [ B AB A 2 B A n −1B] Controllability Matrix
is full-rank.”
Proof: In class
€
ME 433 - State Space Control 68
Observability
Problem Definition: “An unforced system is said to be observable if
and only if it is possible to determine any (arbitrary initial) state x(0) by
using only a finite record, y(τ) for 0≤τ ≤T, of the output”
Theorem: “A system is observable if and only if the matrix
t
A T ( t −τ )
Q( t ) = ∫e C T Ce A ( t −τ) dτ Observability Gramian
0
is nonsingular for some t≥ 0.”
The Observability Gramian satisfies
Q˙€
( t ) = C T C + AT Q(t ) + Q( t ) A, Q(0) = 0 Lyapunov Equation
Theorem: “A system is observable if and only if the matrix
⎡ C ⎤
€ ⎢ ⎥
⎢ CA ⎥
O= Observability Matrix
⎢ ⎥
⎢ n −1 ⎥
⎣CA ⎦
is full-rank.”
ME 433 - State Space Control 69
8
Controllability & Observability
Theorem (Popov-Belevitch-Hautus): Eigenvector Tests
A pair {A,B} will be noncontrollable if and only if there exists a row
vector q≠0 such that
qA = λq, qB = 0
In other words, {A,B} will be controllable if and only if there is no row (or
left) eigenvector of A that is orthogonal to B.
€
A pair {C,A} will be nonobservable if and only if there exists a column
vector p≠0 such that
Ap = λp, Cp = 0
In other words, {C,A} will be observable if and only if there is no column
(or right) eigenvector of A that is orthogonal to C.
€
Proof: In class
ME 433 - State Space Control 70
Controllability & Observability
Theorem (Popov-Belevitch-Hautus): Rank Tests
A pair {A,B} will be controllable if and only if
rank [ sI − A B] = n for all s
A pair {C,A} will be observable if and only if
€ ⎡ C ⎤
rank ⎢ ⎥ = n for all s
⎣ sI − A ⎦
Proof: In class
ME 433 - State Space Control 71
9
Controllability & Observability
Examples :
ME 433 - State Space Control 72
10
