Modern Control Systems

Matthew M. Peet
Illinois Institute of Technology

Lecture 5: Controllability and Observability

State-Space

The standard state-space form is

ẋ(t) = Ax(t) + Bu(t)

y(t) = Cx(t) + Du(t)

State-space reflects an approach based on internal dynamics as opposed to

input-output maps.
• For a given mapping, G : u 7→ y, the choice of A, B, C, D is not unique.

M. Peet Lecture 5: State-Space Theory 2 / 16

Solving the Equations
Find the output given the input

State-Space:
ẋ = Ax(t) + Bu(t) Input Output
u State-Space y
y(t) = Cx(t) + Du(t) x(0) = 0 System

Basic Question: Given an input function, u(t), what is the output?

Solution: Solve the differential Equation.
Example: The equation

ẋ(t) = ax(t), x(0) = x0

has solution
x(t) = eat x0 ,
But we are interested in Matrices. Can we define the matrix exponential?

M. Peet Lecture 5: State-Space Theory 3 / 16

The Solution to State-Space
Ignore Inputs and Outputs:
The equation

ẋ(t) = Ax(t), x(0) = x0

has solution x(t) = eAt x0

The function eAt must satisfy the following

d At
e0 = I, and e = AeAt
dt
For scalars, the matrix exponential is defined as
an
ea = 1 + a + a2 /2 + · · · + + ···
n!
We define the exponential for matrices is defined the same way as scalars
1 1 1
e A = I + A + A2 + A3 + · · · + Ak + · · ·
2 6 k!
M. Peet Lecture 5: State-Space Theory 4 / 16
The Solution to State-Space

The matrix exponential has the following properties

• e0 = I
1 1
e0 = I + 0 + 02 + 03 + · · · = I
2 6
∗
∗
• eM = eM

∗ 1 1 1
eM = I + M ∗ + (M ∗ )2 + (M ∗ )3 + · · · + (M ∗ )k + · · ·
2 6 k!
 ∗  ∗
∗ 1 2 ∗ 1 3 1 k
=I +M +( M ) + M + ··· + M + ···
2 6 k!
 ∗
1 1 1
= I + M + M2 + M3 + · · · + Mk + · · ·
2 6 k!

M. Peet Lecture 5: State-Space Theory 5 / 16

The Solution to State-Space

d At
• dt e = AeAt
 
d At d 1 2 1 k
e = I + (At) + (At) + · · · + (At) + · · ·
dt dt 2 k!
2 3 k
= 0 + A + A(At) + A(At)2 + · · · + A(At)k + · · ·
 2 6 k! 
1 2 1 k−1
= A I + (At) + (At) + · · · + (At) + ···
2 (k − 1)!

However,
eM +N 6= eM eN
Unless, M N = N M .

M. Peet Lecture 5: State-Space Theory 6 / 16

Find the output given the input

The equation ẋ(t) = Ax(t), x(0) = x0

has solution x(t) = eAt x0

Proof.
Let x(t) = eAt x0 , then
• ẋ(t) = AeAt x0 = Ax(t).
• x(t) = e0 x0 = x0

What happens when we add an input instead of an initial condition?

M. Peet Lecture 5: State-Space Theory 7 / 16

Find the output given the input
State-Space:
ẋ = Ax(t) + Bu(t)
y(t) = Cx(t) + Du(t) x(0) = 0

The equation
ẋ(t) = Ax(t) + Bu(t), x(0) = 0

has solution Z
t
x(t) = eA(t−s) Bu(s)ds
0

Proof.
Check the solution:
Z t
ẋ(t) = e0 Bu(t) + A eA(t−s) Bu(s)ds
0
= Bu(t) + Ax(t)

M. Peet Lecture 5: State-Space Theory 8 / 16

Find the output given the input
Solution for State-Space

State-Space:
ẋ = Ax(t) + Bu(t)
y(t) = Cx(t) + Du(t) x(0) = 0

Now that we have x(t), finding y(t) is

easy
y(t) = Cx(t) + Du(t)
Z t
= CeA(t−s) Bu(s)ds + Du(t)
0

Conclusion: Given u(t), only one integration is needed to find y(t)!

Note that the state, x, doesn’t appear!!
We have solved the problem.

M. Peet Lecture 5: State-Space Theory 9 / 16

Calculating the Output
Numerical Example, u(t) = sin(t)
State-Space:
input

ẋ = −x(t) + u(t) 1

0.8

y(t) = x(t) − .5u(t) x(0) = 0 0.6

0.4

A = −1; B = 1; C = 1; D = −.5 0.2

u(t)
0

−0.2

Solution:Z −0.4

t −0.6

CeA(t−s) Bu(s)ds + Du(t)

−0.8

y(t) = −1
0 2 4 6 8 10
0 t

Z t
−t 1
=e es sin(s)ds − sin(t)
0 2 0.6
output

1 −t s
1
e (sin s − cos s)|t0 − sin(t)
0.4
= e
2 2 0.2

1 −t t
1 0

= e e (sin t − cos t) + 1 − sin(t)

y(t)
2 2 −0.2

1 −t
−0.4

= e − cos t −0.6

2 −0.8
0 2 4 6 8 10
t

M. Peet Lecture 5: State-Space Theory 10 / 16

Stability

ẋ(t) = Ax(t)
There are several notions of stability.
• All notions are equivalent for linear systems.

Definition 1.
A differential equation is stable if any solution x(t) satisfies

lim x(t) = 0
t→∞

M. Peet Lecture 5: State-Space Theory 11 / 16

Stability

The unique solution has the form x(t) = eAt x0 .

ẋ(t) = Ax(t)

Question: Is it stable?
Suppose A is diagonalizable, so A = T ΛT −1 , so that

Ak = T ΛT −1 T ΛT −1 · · · T ΛT −1 = T Λk T −1

We conclude that
 
1 1
eAt = T T −1 + (T ΛT −1 )t + (T Λ2 T −1 )t2 + · · · + (T Λk T −1 )tk + · · ·
2 k!
 
1 1
= T I + (Λt) + (Λt)2 + · · · + (Λt)k + · · · T −1
2 k!
= T eΛt T −1

M. Peet Lecture 5: State-Space Theory 12 / 16

Stability

But we can see that eΛt converges

1
eΛt = I + Λt + · · · + Λk tk + · · ·
 k!   λt 
1 + λ1 t + · · · e 1
=
 . .. =
  .. 
. 
1 + λn t + · · · eλn t

The solution is a linear combination of the functions of the form eλi t .

• limt→∞ eλi t → 0 if and only if Re λi (A) < 0 for all i.

M. Peet Lecture 5: State-Space Theory 13 / 16

Stability
Inconveniently, not all matrices are diagonalizable.
• However, all matrices are Jordan diagonalizable.
I Ak = T J k T −1 , where J
• Hence eAt = T eJt T −1
Consider a single Jordan block Ji = λi I + N .
• Convenient because λi I and N commute.
I Hence eλi I+N = eλi I eN .
eJi t = eλi t+N t = eλi t eN t
Conveniently N d = 0, so the series expansion terminates
1 1
eN t = 1 + N + N 2 + · · · + N k−1 tk−1
 λt  2 (k − 1)!
e i  
. 1 2 1
eJ i t = ..  1 + N + N + ··· + N k−1
t k−1
 
λi t
2 (k − 1)!
e
This time all terms have the form ti eλt for i ≤ k
• limt→∞ ti eλt = 0 if and only if Re λ < 0
M. Peet Lecture 5: State-Space Theory 14 / 16
Stability

Now consider the general case A = T JT −1 where

 
J1
J =
 .. 
. 
Jn

Then  Jt 
e 1
eAt = T eJt T −1 =T
 ..  −1
T
.
eJ n t
eAt is entirely composed of terms of the form

eλi t tk
k!
We conclude that ẋ(t) = Ax(t) is stable if and only if Re λ < 0.

M. Peet Lecture 5: State-Space Theory 15 / 16

Stability

Definition 2.
A is Hurwitz if Re λi (A) < 0 for all i.

Theorem 3.
ẋ(t) = Ax(t) is stable if and only if A is Hurwitz.

M. Peet Lecture 5: State-Space Theory 16 / 16