MCS5202 Control Engineering III

State space description of linear systems. Concepts of controllability and observability.

Canonical realization of systems having specified transfer functions. Stability in the sense of
Lyapunov. State feedback, modal control, pole assignment and the optimal quadratic regulator.
Full-order state observers. Multi-lateral systems. Introduction to sampled-data systems. Digital
compensation. Introduction to microprocessor-based Control.

Introduction

One of the most important tasks in the analysis and design of control systems is mathematical
modeling of the system in which physical systems are represented by Dynamic equations in form
of differential equations. The differential equations describing the dynamic performance of a
physical system are obtained by utilizing the physical laws of the process (such as Newton’s
Laws, Kirchhoff’s laws, Ohm’s law etc.).

There are two approaches for analysis and design of control systems namely:

- Classical (frequency domain) approach

- Modern (time domain) approach

In classical approach the systems differential equations are converted in to transfer function.

Advantage of classical approach

- Rapidly provide stability and transient response information

Disadvantage of classical approach

- Restricted to single input single output (SISO) systems only

- Can only be applied to linear time invariant (LTI) systems

In modern approach the systems differential equations are converted in to a set of first-order
differential equations, which may be combined into a first-order vector-matrix differential
equation.

Modern approach is a unified approach for modelling, analysing and designing of a wide range
of control systems.

Advantage of modern approach

- Can be applied to nonlinear systems

- Applicable to time varying systems
- Easily tackled by the availability of advanced digital computer.
- Applicable to multi input multi output (MIMO) systems

1
State Space representation

A system is represented in state space using two state equations as follows:

The dimensions of the vectors and matrices are illustrated below:

State Space representation of nth order systems of linear differential equation

2
The following examples illustrate how to represent a system in state space from its differential
equation:

Example 1

A system is described by the differential equation

3 2
d y d y dy
3
+5 2 + 9 +10 y=6 u (t)
dt dt dt

Where y is the output and u is the input of the system. Obtain the state space representation of
the system.

Solution

Step 1 The differential equation can be rewritten as:

y⃛ =6u ( t )−5 ÿ −9 ẏ−10 y (1)

Step 2 The state variables can be chosen as:

x 1= y

x 2= ẏ (2)

x 3= ÿ

Step 3 The derivatives of the state variables can be obtained as:

ẋ 1= ẏ

ẋ 2= ÿ (3)

ẋ 3= ⃛y

Step 4 Substituting (2) and (1) in (3) and rearrange gives:

ẋ 1=x 2

ẋ 2=x 3 (4)

ẋ 3=6 u ( t )−5 ÿ−9 ẏ −10 y

¿−10 x 1−9 x 2−5 x 3+6 u ( t )

3
From (2) the output equation is:

y ¿ x1 (5)

Step 5 The state equations can be obtained by expressing equations (4) and (5) in matrix form as:

[][ ][ ] [ ]
ẋ 1 0 1 0 x1 0
ẋ 2 = 0 0 1 x2 + 0 u
ẋ3 −10 −9 −5 x 3 6

[]
x1
y= [ 1 0 0 ] x 2
x3

[ ] []
0 1 0 0
A= 0 0 1 B= 0 C=[ 1 0 0 ] D=0
−10 −9 −5 6

Example 2

A single input single output system has the transfer function

Y (s) 1
= 3
U ( s) s +10 s +29 s+18
2

Obtain the state space representation of the system.

Solution

The corresponding differential equation can be obtained as:

⃛y +10 ÿ +29 ẏ +18 y=u

Step 1 The differential equation can be rewritten as:

y⃛ =u−10 ÿ−29 ẏ−18 y (1)

Step 2 The state variables can be chosen as:

x 1= y

x 2= ẏ (2)

4
 x 3= ÿ

Step 3 The derivatives of the state variables can be obtained as:

ẋ 1= ẏ

ẋ 2= ÿ (3)

ẋ 3= ⃛y

Step 4 Substituting (2) and (1) in (3) and rearrange gives:

ẋ 1=x 2

ẋ 2=x 3 (4)

ẋ 3=u−10 ÿ−29 ẏ−18 y

¿−18 x 1−29 x 2−10 x 3+u

From (2) the output equation is:

y ¿ x1 (5)

Step 5 The state equations can be obtained by expressing equations (4) and (5) in matrix form as:

[][ ][ ] [ ]
ẋ 1 0 1 0 x1 0
ẋ 2 = 0 0 1 x2 + 0 u
ẋ3 −18 −29 −10 x 3 1

[]
x1
y= [ 1 0 0 ] x 2
x3

Example 3

Consider a system described by the following pair of differential equations

2
d θ dθ
J 2
+f −k m I a=0
dt dt

5
 dθ d Ia
km + R a I a + La =V (t)
dt dt

Obtain the state space representation of the system. Assume the output of the system to be θ

Solution

The differential equation can be rewritten as:

km f
θ̈= I − θ̇
J a J

1 Ra km
I˙a= V − I a − θ̇ (1)
La La La

The state variables can be chosen as:

x 1=θ

x 2=θ̇ (2)

x 3=I a

The derivatives of the state variables can be obtained as:

ẋ 1=θ̇

ẋ 2=θ̈ (3)

ẋ 3= I˙a

Substituting (2) and (1) in (3) and rearrange gives:

ẋ 1=x 2

km f
ẋ 2= I a− θ̇
J J

km f
¿ x− x
J 3 J 2

1 Ra km
ẋ 3= V − I a − θ̇ (4)
La La La

6
 1 Ra km
¿ V − x 3− x 2
La La La

From (2) the output equation is:

θ=x1 (5)

The state equations can be obtained by expressing equations (4) and (5) in matrix form as:

[ ][ ] [ ]
0 1 0

[]
ẋ 1 k m x1 0
−f
0 0
ẋ 2 = J J x2 + V
1
ẋ3 −k m −Ra x 3
0 La
La La

[]
x1
θ=[ 1 0 0 ] x 2
x3

Example 4

For a given complex system differential equation is

3 2
d x d x dx
3
+ 3 2 +5 +7 x (t)=u1 ( t ) +3 u2 ( t ) +4 u3 (t)
dt dt dt

And the outputs are

2
dx d x
y 1=2 +3 u1 ( t ) and y 2= +6 u2 ( t ) +u3 (t)
dt dt
2

Represent the system in its state space form.

Solution

The differential equation can be rewritten as:

⃛x =u1 ( t )+3 u 2 ( t )+ 4 u3 (t)−3 ẍ−5 ẋ−7 x (1)

The state variables can be chosen as:

x 1=x

x 2= ẋ (2)

7
 x 3= ẍ

The derivatives of the state variables can be obtained as:

ẋ 1= ẋ

ẋ 2= ẍ (3)

ẋ 3= ⃛x

Substituting (1) and (2) in (3) and rearrange gives:

ẋ 1=x 2

ẋ 2=x 3 (4)

ẋ 3=u1 ( t ) +3 u2 ( t )+ 4 u3 ( t )−3 ẍ−5 ẋ−7 x

¿−7 x1 −5 x 2−3 x 3+u1 ( t ) +3 u2 ( t )+ 4 u3 (t )

Substituting (2) in the output equations gives:

y 1=2 x 2+ 3u 1 ( t )

y 2=x 3 +6 u2 ( t ) +u 3 (t) (5)

The state equations can be obtained by expressing equations (4) and (5) in matrix form as:

[][ ][ ] [ ][ ]
ẋ 1 0 1 0 x 1 0 0 0 u1
ẋ 2 = 0 0 1 x 2 + 0 0 0 u2
ẋ3 −7 −5 −3 x 3 1 3 4 u3

[] []
x1 u1
[ ][
y1
y2
= 0 2 0
x +
3 0
] [
0
0 0 1 2 0 6 1 2
x3
u
u3
]
State Space representation of nth order systems of linear differential equation involving
derivatives of forcing function

Consider a system whose differential equation involves derivatives of the forcing (input)
function such as:

8
 n n−1 n n −1
d y d y d u d u du
n
+a 1 n−1 +... …+an y=b0 n +b1 n−1 + ...…+ bn−1 +b n u (1)
dt dt dt dt dt

The state variables are chosen as:

x 1= y− β0 u
x 2= ẏ −β 0 u̇−β1 u
x 3= ÿ −β 0 ü−β 1 u̇−β2 u
(2)
⋮
⋮
(n−1) (n−1) (n−2)
x n= y −β 0 u −β 1 u −...…−β n−2 u̇−β n−1 u

Where

β 0=b0
β 1=b1 −a1 β0
β 2=b2−a 1 β 1−a 2 β 0
β 3=b3 −a1 β2 −a2 β1 −a3 β 0 (3)
⋮
⋮
β n=b n−a 1 β n−1−...−an−1 β 1−an β 0

Example 5

Consider the control system whose closed loop transfer function is

Y (s) s+ 3
= 3
U ( s) s + 9 s +24 s +20
2

Obtain a state space representation of the system.

Solution

The corresponding differential equation can be obtained as:

⃛y +9 ÿ +24 ẏ+ 20 y =u̇+3 u

From the differential equation:

a 1=9, a 2=24 , a 3=20, b 0=0,b 1=0, b 2=1, b 3=3

From which:

9
 β 0=b0=0
β 1=b 1−a1 β 0=0−0=0
β 2=b 2−a1 β 1−a2 β 0=1−0−0=1

The differential equation can be rewritten as:

⃛y =u̇+ 3u−9 ÿ−24 ẏ−20 y (1)

The state variables can be chosen as:

x1 = y−β 0 u= y
x 2= ẏ−β 0 u̇−β 1 u= ẏ (2)
x 3= ÿ− β0 ü− β1 u̇−β 2 u= ÿ −u

The derivatives of the state variables can be obtained as:

ẋ 1= ẏ

ẋ 2= ÿ (3)

ẋ 3= ⃛y −u̇

From (2)

y=x 1
ẏ=x 2 (4)
ÿ=x 3 +u

Substituting (4) and (1) in (3) and rearrange gives:

ẋ 1=x 2

ẋ 2=x 3+ u (5)

ẋ 3=u̇+3 u−9 ÿ−24 ẏ−20 y−u̇

¿−20 x1−24 x 2−9 ( x 3+ u ) +3 u

¿−20 x1−24 x 2−9 x 3−6 u

From (4) the output equation is:

y=x 1 (6)

The state equations can be obtained by expressing equations (5) and (6) in matrix form as:

10
 [][ ][ ] [ ]
ẋ 1 0 1 0 x1 0
ẋ 2 = 0 0 1 x2 + 1 u
ẋ3 −20 −24 −9 x3 −6

[]
x1
y= [ 1 0 0 ] x 2
x3

Example 6

A system is represented by the following differential equation:

⃛y +5 ÿ +4 ẏ+ y=u⃛ +3 ü+ u̇+ 2u

Obtain a state space representation of the system.

Solution

From the differential equation:

a 1=5, a 2=4 , a 3=1, b 0=1,b 1=3, b 2=1, b 3=2

From which:

β 0=b 0=1
β1=b1−a1 β 0=3−5 ( 1 )=3−5=−2
β 2=b 2−a1 β 1−a2 β 0=1−5 (−2 ) −4 ( 1 ) =1+ 10−4=7

The differential equation can be rewritten as:

⃛y =u⃛ + 3 ü+ u̇+2 u−5 ÿ−4 ẏ− y (1)

The state variables can be chosen as:

x 1= y− β0 u= y−u
x 2= ẏ−β 0 u̇−β 1 u= ẏ−u̇+ 2u (2)
x 3= ÿ− β0 ü−β1 u̇−β 2 u= ÿ −ü+2 u̇−7 u

The derivatives of the state variables can be obtained as:

ẋ 1= ẏ−u̇

ẋ 2= ÿ−ü+ 2 u̇ (3)

ẋ 3= ⃛y −u⃛ +2 ü−7 u̇

11
From (2)

y=x 1 +u

ẏ=x 2 + u̇−2u (4)

ÿ=x 3 + ü−2 u̇+ 7 u

Substituting (4) and (1) in (3) and rearrange gives:

ẋ 1=x 2+ u̇−2 u−u̇

¿ x 2−2 u

ẋ 2=x 3+ ü−2 u̇+7 u−ü+2 u̇

¿ x 3+ 7 u (5)

ẋ 3=u⃛ +3 ü+ u̇+ 2u−5 ÿ−4 ẏ− y−u⃛ +2 ü−7 u̇

¿ 5 ü−6 u̇+ 2u−5 ÿ−4 ẏ− y

¿ 5 ü−6 u̇+ 2u−5( x 3 + ü−2 u̇+7 u)−4 (x2 + u̇−2 u)−(x 1 +u)

¿ 5 ü−6 u̇+ 2u−5 x3 −5 ü +10 u̇−35 u−4 x 2−4 u̇+8 u−x1−u

¿−x 1−4 x 2−5 x 3−26 u

From (4) the output equation is:

y=x 1 +u (6)

The state equations can be obtained by expressing equations (5) and (6) in matrix form as:

[][ ][ ] [ ]
ẋ 1 0 1 0 x1 −2
ẋ 2 = 0 0 1 x2 + 7 u
ẋ3 −1 −4 −5 x 3 −26

[]
x1
y= [ 1 0 0 ] x 2 +u
x3

Transfer function from state space equations

Consider the state space model

12
 ẋ= Ax+ Bu (1)

y=Cx + Du (2)

The transfer function can be obtained as follows:

Taking the Laplace transform of (1) assuming zero initial condition gives:

sX ( s )= AX ( s )+ BU ( s) (3)

Y (s)=CX (s)+ DU (s) (4)

From (3)

sX ( s )− AX ( s )=BU ( s)

( sI− A ) X ( s )=BU ( s )
−1
X ( s ) =( sI− A) BU ( s) (5)

Where is I identity matrix

Substituting (5) in (4) gives:

−1
Y ( s)=C ( sI− A) BU ( s)+ DU ( s)
−1
Y ( s )=[C ( sI −A ) B+ D ]U (s ) (6)

The transfer function is obtained as:

Y ( s)
G ( s )= =C ( sI− A )−1 B+ D
U (s)

C[adj ( sI −A ) ]B
¿ +D
|sI− A|

C [ adj ( sI− A ) ] B+ D|sI− A|

¿
|sI− A|

Example 1

Determine the transfer function relating the output y (t ) to the input u(t ) for the system described
by the state space equation below.

13
 [][
ẋ1
ẋ2
=
3 −4 x2][ ] [ ]
2 −1 x 1
+
1
−1
u

y= [ 1 0 ]
[]
x1
x2

Solution

From the state space equations:

A=
[ 23 −1
−4 ]
, B=[ ], C=[ 1
1
−1
0 ], D=0, I = [10 01]
The transfer function can be obtained using the relation:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

(sI− A)=s [ 10 01]−[23 −1

−4 ] =[ ] −[
s 0 2 −1
0 s 3 −4 ]=
[ ( s−2 )
−3
1
s+ 4 ]
adj ( sI −A )=
[ s +43 s−2
−1
]
|sI− A|=|( s−2 ) 1 |=( s−2 )( s+ 4 )−3 (−1 )
−3 s+ 4
2 2
¿ s +4 s−2 s−8+3=s +2 s−5

Therefore the transfer function can be obtained as:

G ( s )=
Y ( s)
=
[3
[ 1 0 ] s +4 −1 1
s−2 −1][ ]
U (s) 2
s +2 s−5

¿
[ s+ 4 −1 ] 1
2
[−1]
s + 2 s−5

( s +4 ) (1 )−1(−1)
¿ 2
s +2 s−5

s+5
¿ 2
s + 2 s−5

14
Example 1 can be solved using the following Matlab code:

A=[2 -1;3 -4]

B=[1;-1]
C=[1 0]
D=0
[num,den]=ss2tf(A,B,C,D)
G=tf(num,den)

Example 2

Determine the transfer function for the system described by the following state space equation:

[][
ẋ1 −3 1 x 1 0
= + u
ẋ2 −2 0 x 2 1 ][ ] [ ]
y= [ 1 0 ]
[]
x1
x2

Solution

From the state space equations:

A=
[−3 −1
−2 0 ], B=[ ], C=[ 1
0
1
0 ], D=0, I = [10 01]
The transfer function can be obtained using the relation:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

(sI− A)=s [ 10 01]−[−3 −1

−2 0 ] =[ ] −[
s 0 −3 −1
0 s −2 0 ]=
[ s ]
( s +3 ) −1
2

adj ( sI −A )=
[−2s s+31 ]
|sI− A|=|( s+3 ) −1|=s ( s+ 3 )−2 (−1 )=s +3 s+2
2
2 s

Therefore the transfer function can be obtained as:

15
 G ( s )=
Y ( s)
=
[1 0 ] s [−2 ][ ]
1 0
s+3 1
U (s) 2
s +3 s +2

¿
[ s 1] 0
2
[1 ]
s + 3 s+ 2

1
¿ 2
s + 3 s+ 2

Example 2 can be solved using the following Matlab code:

A=[-3 1;-2 0]
B=[0;1]
C=[1 0]
D=0
[num,den]=ss2tf(A,B,C,D)
G=tf(num,den)

Example 3

Determine the transfer function of a linear system described by the following state equations:

[][ ][ ] [ ]
ẋ 1 −2 0 1 x1 1
ẋ 2 = 1 −2 0 x 2 + 0 u
ẋ3 1 1 −1 x 3 1

[]
x1
y= [ 2 1 −1 ] x 2 +u
x3

Solution

From the state equations given

[ ] []
−2 0 1 1
A= 1 −2 0 , B= 0 ,C=[ 2 1 −1 ], D=1
1 1 −1 1

[ ][ ][ ]
s 0 0 −2 0 1 s +2 0 −1
( sI− A ) = 0 s 0 − 1 −2 0 = −1 s+ 2 0
0 0 s 1 1 −1 −1 −1 s +1

16
 [ ]
| | | | | |
T
s+2 0 −1 0 −1 s +2
−
−1 s +1 −1 s+1 −1 −1
adj ( sI −A )= −
0
−1| −1
s +1 | | s +2
−1
−1
s+1 | | −
s+ 2 0
−1 −1 |
0
|
s+2
−1
0 | | − s +2
−1
−1
0 || s+2 0
−1 s +2 |

[ ]
2 T
s + 3 s+ 2 s+ 1 s +3
2
= 1 s +3 s +1 s +2
2
s+2 1 s + 4 s+ 4

[ ]
s2 +3 s+2 1 s+2
2
¿ s +1 s + 3 s+ 1 1
2
s +3 s+2 s +4 s +4

| |
s+ 2 0 −1
|sI− A|= −1 s +2 0
−1 −1 s+ 1

¿ ( s+2 ) |s+2
−1
0
s +1| |
−( 0 )
−1 0
−1 s+1 | |
+ (−1 )
−1 s +2
−1 −1 |
¿ ( s+2 ) ( s2 +3 s +2 )−0−( s+3 )
3 2 2
¿ s +3 s +2 s+ 2 s +6 s+ 4−s−3
3 2
¿ s +5 s +7 s +1

The transfer function can be obtained using the relation:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

[ ][ ]
s2 +3 s +2 1 s+2 1
[ 2 1 −1 ] s+ 1 2
s +3 s+1 1
3 2
0 + s +5 s +7 s+1
s+3 s +2 s 2 +4 s+ 4 1
¿
s 3+5 s2 +7 s+1

[]
1
[(2 s¿¿ 2+6 s+ 4)+ ( s+1 )−(s+3) ¿ 2+ ( s 2+ 3 s+ 1 )−(s +2) ( 2 s +4 ) +1−( s2 + 4 s +4 ) ] 0 +s 3 +5 s 2 +7 s +1
1
¿ 3 2
s +5 s + 7 s+ 1

17
 []
1
[ 2 s2 +6 s+ 2 s +2 s +1 −s −2 s+1 ] 0 +s +5 s + 7 s+ 1
2 2 3 2

1
¿ 3 2
s +5 s +7 s+1
2 3 2
(2 s ¿¿ 2+6 s+2)+0+(−s −2 s +1)+ s +5 s +7 s+1
¿ ¿
s3 +5 s 2 +7 s +1
2 3 2
s + 4 s+3+ s + 5 s +7 s+1
¿ 3 2
s +5 s +7 s +1
3 2
s + 6 s +11s +4
3 2
s +5 s + 7 s +1

Example 3 can be solved using the following Matlab code:

A=[-2 0 1;1 -2 0;1 1 -1]
B=[1;0;1]
C=[2 1 -1]
D=1
[num,den]=ss2tf(A,B,C,D)
G=tf(num,den)

Example 4

Find the transfer function of the system represented by the following state and output equations:

[][ ][ ] [ ]
ẋ 1 0 1 0 x1 1
=
ẋ 2 0 0 +
1 x 2 −1 u
ẋ3 0 −4 −6 x 3 5

[]
x1
y= [ 1 0 0 ] x 2
x3

Solution

From the state equations given

[ ] []
0 1 0 1
A= 0 0 1 , B= −1 , C=[ 1 0 0 ], D=0
0 −4 −6 5

18
 [ ][ ][ ]
s 0 0 0 1 0 s −1 0
( sI− A ) = 0 s 0 − 0 0 1 = 0 s −1
0 0 s 0 −4 −6 0 4 s +6

[ ]
| | | | | |
T
s −1 0 −1 0 s
−
4 s+6 0 s +6 0 4
adj ( sI −A )= −
−1
4 | 0
s+ 6 | | | | | s
0
0
s +6
−
s −1
0 4
−1
s| 0
−1 | | | | | −s
0
0
−1
s −1
0 s

[ ]
2 T
s + 6 s +4 0 0
= s+ 6 2
s +6 s −4 s
1 s s

[ ]
s2 +6 s+ 4 s +6 1
¿ 0
2
s +6 s s
0 −4 s s

| |
s −1 0
|sI− A|= 0 s −1
0 4 s+ 6

¿ ( s) |4s −1
s+ 6 | |
−(−1 )
0 −1
0 s+ 6
+( 0 )| | |
0 s
0 4

¿ s ( s 2 +6 s+ 4 ) −0−0
3 2
¿ s +6 s + 4 s

The transfer function can be obtained using the relation:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

[ ][ ]
s 2+ 6 s +4 s+6 1 1
[1 0 0 ] 0
2
s +6 s s −1
0 −4 s s 5
¿ 3 2
s +6 s +4 s

19
 []
1
[ s 2+ 6 s +4 s+6 1 ] −1
5
¿ 3 2
s +6s +4 s

( s 2 +6 s +4 )−(s+6)+5
¿ 3 2
s +6 s + 4 s
2
s +5 s +3
¿ 3 2
s +6s +4 s

Example 4 can be solved using the following Matlab code:

A=[ 0 1 0;0 0 1;0 -4 -6]
B=[1;-1;5]
C=[1 0 0]
D=0
[num,den]=ss2tf(A,B,C,D)
G=tf(num,den)

Characteristic equation

Consider the transfer function:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

The characteristic equation is given by:

|sI− A|=0

Eigenvalues and Eigenvectors

Eigenvalues

The eigenvalues of an nxn system matrix A are the roots of the characteristic equation:

|λI − A|=0
Eigenvalues of system matrix (A) are actually poles of the system’s transfer function. For a
stable system all the eigenvalues must have negative real part.

Example 1

20
A system has it’s a matrix given by

[ ]
4 −1 −1
A= −1 2 1
−1 1 2

Determine:

i. The characteristics equation

ii. Eigenvalues of A
iii. Stability of the system

Solution

i. |λI − A|=¿

|[ ] [ ]| |[ ]|
λ 0 0 4 −1 −1 λ−4 1 1
0 λ 0 − −1 2 1 = 1 λ−2 −1 =0
0 0 λ −1 1 2 1 −1 λ−2

| | | | |
( λ−4 ) λ−2 −1 −( 1 ) 1 −1 + ( 1 ) 1 λ−2 =0
−1 λ−2 1 λ−2 1 −1 |
( ( λ−4 ) ( ( λ−2 ) ( λ−2 )−1) )−( ( λ−2 )+1 )+(−1−( λ−2 ))=0

( ( λ−4 )( ( λ2−4 λ +4 )−1)) −( λ−1 )−( λ−1 )=0

( ( λ 3−4 λ2 +3 λ )−4 λ2 +16 λ−12)−( 2 λ−2 )=0

3 2
λ −8 λ +17 λ−10=0

ii. Factorizing the characteristics equation gives:

( λ−1 ) ( λ−2 )( λ−5 ) =0

Therefore the eigenvalues are: λ=1, λ=2 and λ=5

iii. Since the real parts of the eigenvalues are positive the system is unstable.

The eigenvalues in example 1 can be solved using the following Matlab code:
A=[ 4 -1 -1;-1 2 1;-1 1 2]
E=eig(A)

21
Eigenvectors

For each eigenvalue λ i, there is a corresponding eigenvector x i associated with it. The
eigenvector is a set of solutions in the form of a non-zero vector. Eigenvalues are unique but
eigenvectors are non-unique.

To find eigenvector the following relation is used:

Ax=λx

( λI − A ) x=0

Example 2

A system matrix A given by

[ ]
2 1 1
A= 2 3 4
−1 −1 −2

Determine:

i. Characteristics equation
ii. Eigenvalues
iii. Eigenvectors

Solution

i. |λI − A|=¿

|[ ] [ ]| |[ ]|
λ 0 0 2 1 1 λ−2 −1 −1
0 λ 0− 2 3 4 = −2 λ−3 −4 =0
0 0 λ −1 −1 −2 1 1 λ+ 2

| | | | |
( λ−2 ) λ−3 −4 − (−1 ) −2 −4 + (−1 ) −2 λ−3 =0
1 λ+2 1 λ+ 2 1 1 |
( ( λ−2 ) (( λ−3 )( λ+ 2 ) +4) ) + (−( 2 λ+4 ) +4 )−(−2−( λ−3 ))=0

22
 ( ( λ3− λ2−2 λ ) −2 λ2 +2 λ+ 4 ) −λ−1=0
3 2
λ −3 λ − λ+3=0

ii. Factorizing the characteristics equation gives:

( λ−1 ) ( λ+1 ) ( λ+3 )=0

Therefore the eigenvalues are: λ=1, λ=−1 and λ=3

iii. Eigenvectors: Using ( λI − A ) x=0

For λ=1
( I − A ) x=0

[ ][ ][ ] [ ]
1 0 0 2 1 1 xa 0
0 1 0− 2 3 4 xb = 0
0 0 1 −1 −1 −2 xc 0

[ ][ ] [ ]
−1 −1 −1 x a 0
−2 −2 −4 x b = 0
1 1 3 xc 0

−x a−x b −xc =0 (1)

−2 x a−2 x b−4 x c=0 (2)
x a + x b +3 x c=0 (3)

Adding equation (1) and (3) gives: 2 x c =0

Therefore x c =0

Substituting x c =0 in (2) gives:

x b=−x a
Let
x a=1
Then
x b=−1

Therefore the first eigenvector is:

[]
1
v 1= −1
0

23
For λ=−1
(−I − A ) x=0

[ ][ ][ ] [ ]
−1 0 0 2 1 1 xa 0
0 −1 0 − 2 3 4 xb = 0
0 0 −1 −1 −1 −2 xc 0

[ ][ ] [ ]
−3 −1 −1 x a 0
−2 −4 −4 x b = 0
1 1 1 xc 0

−3 x a−x b −x c =0 (4)
−2 x a−4 x b−4 x c =0 (5)
x a + xb + x c =0 (6)

Adding equation (4) and (6) gives: −2 x a=0

Therefore x a=0

Substituting x a=0 in (2) gives:

x c =−x b

Let
x b=1
Then
x c =−1
Therefore the second eigenvector is:

[]
0
v 2= 1
−1
For λ=3
( 3 I − A ) x=0

[ ][ ][ ] [ ]
3 0 0 2 1 1 xa 0
0 3 0− 2 3 4 xb = 0
0 0 3 −1 −1 −2 xc 0

[ ][ ] [ ]
1 −1 −1 x a 0
−2 0 −4 x b = 0
1 1 5 xc 0

x a −x b−x c =0 (7)

24
 −2 x a−4 x c =0 (8)
x a + xb +5 x c =0 (9)

Adding equation (7) and (9) gives:

2 x a+ 4 x c =0
Then
x a=−2 x c
Let
x c =−1
Then
x a=2
From (7)
x b=x a−x c (10)
Substituting x c =−1 and x a=−2 in (10) gives
x b=3

Therefore the third eigenvector is:

[]
2
v 3= 3
−1
The eigenvector of A is

[ ]
1 0 2
V = [ v 1 v 2 v 3 ]= −1 1 3
0 −1 −1

The eigenvalues of example 2 can be obtained using the following Matlab code:
A=[ 2 1 1;2 3 4;-1 -1 -2]
[V,D]=eig(A)

The above code produces a diagonal matrix D of eigenvalues and a full matrix V whose columns
are the corresponding eigenvectors so that X*V = V*D.

State Equations Transformation

Consider the state space model

ẋ= Ax+ Bu (1)

y=Cx + Du (2)

It is possible to transform the state-space representation with state vector x into a state-space
representation with a state vector z .
25
Suppose the relation between the state vector is defined by the following transformation:

x=Pz (3)

Where P is the transformation matrix.

Substituting (3) into equations (1) and (2) gives:

P ż=APz +Bu
−1 −1
ż=P APz+ P Bu (4)

y=CPz + Du (5)

Hence, the transformation yields a new space representation. If the newly formed A matrix is
denoted by A z=P−1 AP , then A and A z have the following properties:

1. The same eigenvalues

2. The same determinant
3. The same trace (i.e. sum of elements on the principal diagonal).

Example 1

Given the system represented in state space as:

[][ ][ ] [ ]
ẋ 1 0 1 0 x1 0
ẋ 2 = 0 0 1 x2 + 0 u
ẋ3 −2 −5 −7 x3 1

[]
x1
y= [ 1 0 0 ] x 2
x3

Transform the system to a new set of state variables z , where the new variables are related to the
original variables x as follows:

}
z 1=2 x1
z 2=3 x 1 +2 x 2 (a)
z 3=x 1+ 4 x 2 +5 x 3

Solution

[ ] []
0 1 0 0
A= 0 0 1 , B= 0 , C=[ 1 0 0 ] D=0
−2 −5 −7 1

26
From (a)

[ ] [ ][ ]
ż 1 2 0 0 x 1
ż 2 = 3 2 0 x 2 (b)
ż3 1 4 5 x3

From (3)
−1
z=P x (c)

Comparing (b) with (c) gives:

[ ]
2 0 0
−1
P =3 2 0
1 4 5

[ ]
| | | | | |
T
2 0 3 0 3 2
+ − +
4 5 1 5 1 4
−| | | | | |
0 0
+
2 0
−
2 0

[ ]
4 5 1 5 1 4 T
10 −15 10

−1 −1 adjP
| | | | | |
+0
2
0
0
−2
3
0
0
+2
3
0
2
0
0
10 −8
0 4
P=( P ) = = =

| | |24 05|−0|31 05|+0|31 42|

|P| 2 0 0 2
3 2 0
1 4 5

[ ]
10 0 0
−15 10 0

[ ]
10 −8 4 0.5 0 0
¿ = −0.75 0.5 0
2(10)−0+0
0.5 −0.4 0.2

[ ][ ][ ]
2 0 0 0 1 0 0.5 0 0
−1
P AP= 3 2 0 0 0 1 −0.75 0.5 0
1 4 5 −2 −5 −7 0.5 −0.4 0.2

[ ][ ][ ]
0 2 0 0.5 0 0 −1.5 1 0
¿ 0 3 2 −0.75 0.5 0 = −1.25 0.7 0.4
−10 −24 −31 0.5 −0.4 0.2 −2.5 0.4 −6.2

[ ][ ] [ ]
2 0 0 0 0
−1
P B= 3 2 0 0 = 0
1 4 5 1 5

27
 [ ]
0.5 0 0
CP=[ 1 0 0 ] −0.75 0.5 0 =[ 0.5 0 0 ]
0.5 −0.4 0.2

Hence

[ ] []
−1.5 1 0 0
ż= −1.25 0.7 0.4 z + 0 u
−2.5 0.4 −6.2 5

y= [ 0.5 0 0 ] z

State Space in diagonal form

A system in state space can be represented in diagonal form so that all system state are separated
or uncoupled as illustrated in the example below:

From the example, it can be seen that every state equation is a function of each state variable.
Therefore each differential equation can be solve independently

Consider the state space model

ẋ= Ax+ Bu (1)

y=Cx + Du (2)

The state space model in (1) and (2) can be transform into diagonal form as follows:
−1 −1
ż=P APz+ P Bu (4)

y=CPz + Du (5)

In general, there are several methods of finding the transformation matrix P. It can be formed by
the use of the eigenvectors of A, that is

P=V

Note that the matrix P−1 AP is a diagonal matrix whose diagonal elements are the eigenvalues of
A.

Example 1

28
Consider a system represented in the state space by the following equations:

ẋ=
[−31 −31 ] x +[12] u
y= [ 2 3 ] x

Represent the system in diagonal form.

Solution

A=
[−31 −31 ], B=[ 12], C=[ 2 3 ]

Eigenvalues

|λI − A|=0

|[ 0 λ] [ 1
|λI − A|= λ 0 − −3 1
−3
=
]| |
λ +3
−1
−1
|
λ+3
=0

( λ+ 3 ) ( λ+3 )−1=0
2
λ +6 λ+9−1=0
2
λ +6 λ+8=0
( λ+ 2 )( λ +4 )=0

Hence λ 1=−2 and λ 2=−4

Eigenvectors

( λI − A ) x=0

For λ=−2

[[ 0 −2 ][
−2 0 − −3 1
1 −3 ]][ xx ]=[−11 −11 ][ xx ]=[ 00]
1

2
a

x a−x b=0 (1)

−x a + x b=0 (2)

From (1) or (2)

x b=x a

Let

29
 x a=1

Then

x b=1

Therefore the first eigenvector is:

[ 11]
v 1=

For λ=−4

[[ −40 −40 ]−[−31 −31 ] ][ xx ]=[−1

a

b
−1 −1 ] [ x ] [ 0 ]
−1 x = 0 a

−x a−x b =0 (1)
−x a−x b =0 (2)

From (1) or (2)

x b=−x a

Let

x a=1

Then

x b=−1

Therefore the second eigenvector is:

v 1=
[−11 ]
The eigenvector of A is therefore:

V = [ v 1 v 2 ]=
[ 11 −11 ]
Transformation matrix

30
 P=V =
[ 1 1
1 −1 ]
[ −1 −1
] [ −1 1 ] 0.5 0.5
−1 −1
−1
P =
adjP −1 1
= = =
[ 0.5 −0.5]
|P|
|1 −1|
1 1 −2

P
−1
AP=
[ 0.5
0.5 −0.5 ] [ 1 −3 ] [1 −1 ] [−2 2 ] [ 1 −1 ] [ 0 −4 ]
0.5 −3 1 1 1
=
−1 −1 1 1
=
−2 0

−1
P B=
[0.5 0.5 1
[
0.5 −0.5 2 −0.5 ]
] ]=
[ 1.5

CP=[ 2 3 ]
[11 −11 ]= [5 −1]
Hence

ż=
[−20 −40 ] z +[−0.5
1.5
]u
y= [ 5 −1 ] z

Example 1 can be solved using the following Matlab code:

A=[-3 1;1 -3]

B=[1;2]
C=[2 3]
D=0
G=ss(A,B,C,D)
canon(G,'modal')

Example 2

Diagonalize the following system

[ ][]
−5 −5 4 −1
ẋ= 2 0 −2 x+ 2 r
0 −2 −1 −2

y= [−1 1 2 ] x

Solution

31
 [ ] []
−5 −5 4 −1
A= 2 0 −2 , B= 2 , C=[ −1 1 2 ]
0 −2 −1 −2

Eigenvalues

|λI − A|=0

|[ ] [ ]| | |
λ 0 0 −5 −5 4 λ+5 5 −4
|λI − A|= 0 λ 0 − 2 0 −2 = −2 λ 2 =0
0 0 λ 0 −2 −1 0 2 λ+1

| | |
( λ+ 5 ) λ 2 −5 −2 2 ±4 −2 λ =0
2 λ+1 0 λ +1 0 2 | | |
( λ+ 5 ) ( λ2 + λ−4 )−5 (−2 λ−2 )−4 (−4 )=0
( λ 3+ λ2−4 λ+ 5 λ2 +5 λ−20 ) + ( 10 λ+10 )+16=0
3 2
λ + 6 λ +11 λ+ 6=0
( λ+ 1 )( λ +2 ) ( λ+3 )=0

Hence λ 1=−1, λ 2=−2 and λ 3=−3

Eigenvectors

( λI − A ) x=0

For λ=−1

[ ][ ][ ] [ ][ ] [ ]
−1 0 0 −5 −5 4 xa 4 5 −4 x a 0
0 −1 0 − 2 0 −2 =
x b −2 −1 2 x b = 0
0 0 −1 0 −2 −1 xc 0 2 0 xc 0

4 x a+ 5 x b −4 x c =0 (1)
−2 x a−x b +2 x c =0 (2)
2 x b=0 (3)

From (3)

x b=0 (4)

Substituting (4) in (1) and (2) gives:

4 x a−4 x c=0 (5)

−2 x a+ 2 x c =0 (6)

32
From (5) or (6)

x a=x c

Let

x a=1

Then

x c =1

Therefore the first eigenvector is:

[]
1
v 1= 0
1

For λ=−2

[ ][ ][ ] [ ][ ] [ ]
−2 0 0 −5 −5 4 xa 3 5 −4 x a 0
0 −2 0 − 2 0 −2 x b = −2 −2 2 x b = 0
0 0 −2 0 −2 −1 xc 0 2 −1 x c 0

3 x a +5 x b −4 x c =0 (1)
−2 x a−2 x b+ 2 x c =0 (2)
2 x b−x c =0 (3)

From (3)

x c =2 x b (4)

Let

x b=1 (5)

Then

x c =2 (6)

Substituting (5) and (6) in (1) or (2) gives:

x a=1 (7)

Therefore the second eigenvector is:

33
 []
1
v 2= 1
2

For λ=−3

[ ][ ][ ] [ ][ ] [ ]
−3 0 0 −5 −5 4 xa 2 5 −4 x a 0
0 −3 0 − 2 0 −2 x b = −2 −3 2 xb = 0
0 0 −3 0 −2 −1 xc 0 2 −2 x c 0

2 x a+ 5 x b −4 x c =0 (1)
−2 x a−3 x b+ 2 x c =0 (2)
2 x b−2 x c =0 (3)

From (3)

x c =x b (4)

Let

x b=1 (5)

Then

x c =1 (6)

Substituting (5) and (6) in (1) or (2) gives:

x a=−0.5 (7)

Therefore the third eigenvector is:

[ ]
−0.5
v 3= 1
1

The eigenvector matrix of A is therefore:

[ ]
1 1 −0.5
V = [ v 1 v 2 v 3 ]= 0 1 1
1 2 1

Transformation matrix

34
 [ ]
1 1 −0.5
P=V = 0 1 1
1 2 1

[ ]
| | | | | |
T
1 1 0 1 0 1
−
2 1 1 1 1 2
−|1 −0.5
| | | | |
1 −0.5
−
1 1

[ ]
2 1 1 1 1 2 T
−1 1 −1

adjP
−1
|1 −0.5
1 1 | | | | |
− 1 −0.5
0 1
1
0
1
1
−2 1.5 −1
1.5 −1 1
P = = =

| | |12 11|−|01 11|−0.5|01 12|

|P| 1 1 −0.5
0 1 1
1 2 1

[ ][ ]
T
−1 1 −1 −1 −2 1.5
−2 1.5 −1 1 1.5 −1

[ ]
1.5 −1 1 −1 −1 1 −2 −4 3
¿ = = 2 3 −2
−1+1+ 0.5 0.5
−2 −2 2

[ ][ ][
][ ][ ][ ]
−2 −4 3 −5 −5 4 1 1 −0.5 2 4 −3 1 1 −0.5 −1 0 0
−1
P AP= 2 3 −2 2 0 −2 0 1 1 = −4 −6 4 0 1 1 = 0 −2 0
−2 −2 2 0 −2 −1 1 2 1 6 6 −6 1 2 1 0 0 −3

[ ][ ] [ ]
−2 −4 3 −1 −12
−1
P B= 2 3 −2 2 = 8
−2 −2 2 −2 −6

[ ]
1 1 −0.5
CP=[ −1 1 2 ] 0 1 1 =[ 1 4 3.5 ]
1 2 1

Hence

[ ][ ]
−1 0 0 −12
ż= 0 −2 0 z+ 8 u
0 0 −3 −6

y= [ 1 4 3.5 ] z

Time Domain Solution of State Equations

The process of determining the time domain solution of the state equation is similar to that of
solving a simple first order differential equation.

35
Consider a simple first order differential equation given by:

ẋ=ax (1)

Taking Laplace transform of (1) gives:

sX ( s )−x (0)=aX ( s) sX ( s )−aX (s )=x (0)

(s−a) X (s)=x (0)

x (0)
X ( s) = (2)
(s−a)

Taking the inverse Laplace transform of (2) gives the solution of the differential equation in (1)
as:
at
x (t )=e x (0) (3)

Consider the state equation:

ẋ (t)= Ax (t)+ Bu(t) (4)

Homogenous system

A homogenous system is a system without input that is unforced system.

For homogenous system u(t )=0 hence the state equation in (1) becomes:

ẋ (t)= Ax (t) (5)

Similar to (3) the solution of (5) can be written as:

At
x (t )=e x ( 0 )

¿ Φ (t)x ( 0 ) (6)
At
Φ ( t )=e is called the state-transition matrix, since it performs a transformation on x (0),
taking x from the initial state, x ( 0 ) , to the state x (t) at any time, t .

Each term of Φ ( t ) is the sum of exponentials generated by the system's poles (eigenvalues).

Φ ( t ) has the following properties:

i. Φ ( 0 )=I
ii. Φ̇ ( 0 )= A

36
The state-transition matrix Φ ( t ) can be obtained using two techniques:

1. Classical technique
2. Laplace technique

Classical technique

The following example illustrates the classical technique:

Example 1

Obtain the state-transition matrix Φ ( t ) of the following system using classical method:

[][
ẋ1
=
0 1 x1
ẋ2 −2 −3 x 2 ][ ]
Solution

A=
[−20 −31 ]
Eigenvalues

|λI − A|=0

|[ ] [
|λI − A|= λ 0 − 0
0 λ
1
−2 −3 ]| |
=
λ
2
−1
|
λ+ 3
=0

λ ( λ +3 ) +2=0
2
λ +3 λ +2=0
( λ+ 1 )( λ +2 )=0

Hence λ 1=−1 and λ 2=−2

Since each term of the state-transition matrix is the sum of responses generated by the poles
(eigenvalues), we assume a state-transition matrix of the:

[ ]
−t −2 t −t −2 t
k 1 e +k 2 e k3 e + k4 e
Φ ( t )= −t −2 t −t −2 t (1)
k 5 e +k 6 e k7 e + k8 e

Using property (i) that is

Φ ( 0 )=I

[ k 1 +k 2 k 3 + k 4
k 5 +k 6 k 7+ k 8
=
][ ]
1 0
0 1

37
Hence

k 1+ k 2=1 (2)

k 3+ k 4 =0 ⇒ k 3 ¿−k 4 (3)

k 5+ k 6=0 ⇒ k 5 ¿−k 6 (4)

k 7+ k 8=1 ⇒ k 7 ¿ 1−k 8 (5)

Using property (ii) that is

Φ̇ ( 0 )= A

[ ][ ]
−t −2t −t −2 t
−k 1 e −2 k 2 e −k 3 e −2k 4 e
= 0 1
−t
−k 5 e −2 k 6 e
−2t −t
−k 7 e −2k 8 e
−2 t
−2 −3

−k 1−2 k 2=0⇒ k 1 ¿−2 k 2 (6)

−k 3−2 k 4=1 (7)

−k 5−2 k 6=−2 (8)

−k 7−2 k 8=−3 (9)

Adding (2) and (6) gives:

k 2=−1

Hence

k 1=2

Adding (3) and (7) gives:

k 4=−1

Hence

k 3=1

Adding (4) and (8) gives:

38
 k 6=2

Hence

k 5=−2

Adding (5) and (9) gives:

k 8=2

Hence

k 7=−1

Therefore

[ ]
−t −2 t −t −2 t
2 e −e e −e
Φ ( t )= −t −2t
−2 e +2 e −e−t + 2 e−2 t

Laplace technique

Consider the homogenous state equation:

ẋ= Ax (1)

Taking the Laplace transform of (1) gives:

sX ( s )−x ( 0 )= AX ( s)

sX ( s )− AX ( s )=x ( 0 )

(sI− A) X ( s )=x ( 0 )
−1
X ( s ) =( sI− A) x ( 0 ) (2)

X ( s ) =Φ ( s ) x ( 0 ) (3)

Where
−1
Φ ( s ) =(sI− A) (4)

Taking inverse Laplace transform gives:

−1 −1 −1
L [ X ( s ) ]=L [ ( sI− A ) ]x ( 0 )
−1 −1
x (t)=L [ ( sI− A ) ]x ( 0 )

39
 −1
x (t)=L [Φ ( s ) ]x ( 0 )
−1
Φ ( t )=L [Φ ( s ) ]

Example 2

Obtain the state-transition matrix Φ ( t ) of the following system using Laplace method:

[][
ẋ1
=
0 1 x1
ẋ2 −2 −3 x 2 ][ ]
Solution

[−20 −31 ]
A=

( sI− A ) =[ s 0 ]−[ 0 ] [ 2 s+3 ]

1 s −1
=
0 s −2 −3

[ ] [ ]
[ ]
s+3 1 s+3 1 s+3 1
adj ( sI− A ) −2 s −2 s 2 2
= s +3 s +2 s + 3 s+ 2
−1
( sI− A ) = =
|sI −A|
|2 s +3| −2
2
s −1 s +3 s +2 s
2 2
s +3 s +2 s + 3 s+ 2

[ ]
s +3 1
(s+1)(s+2) (s+1)(s+2)
¿
−2 s
(s+1)(s+2) (s+1)(s+2)

Using partial fraction expansion:

s+ 3 A B
= +
( s+1)( s+2) (s +1) ( s+2)

( s+2 ) A + ( s+1 ) B=s+3

( A+ B ) s + ( 2 A + B )=s+3

Comparing coefficients gives:

A+ B=1⇒ B=1− A (1)

2 A + B=3 (2)

Equation (2) – (1) gives:

40
 A=2

Hence

B=1−2=−1

1 A B
= +
( s+1)( s+2) (s +1) ( s+2)

( s+2 ) A + ( s+1 ) B=1

( A+ B ) s + ( 2 A + B )=1

Comparing coefficients gives:

A+ B=0⇒ B=− A (3)

2 A + B=1 (4)

Equation (4) – (3) gives:

A=1

Hence

B=−1

−2 A B
= +
( s+1)( s+2) (s +1) ( s+2)

( s+2 ) A + ( s+1 ) B=−2

( A+ B ) s + ( 2 A + B )=−2

Comparing coefficients gives:

A+ B=0⇒ B=− A (5)

2 A + B=−2 (6)

Equation (6) – (5) gives:

A=−2

Hence

B=2

41
 s A B
= +
( s+1)( s+2) (s +1) ( s+2)

( s+2 ) A + ( s+1 ) B=s

( A+ B ) s + ( 2 A + B )=s

Comparing coefficients gives:

A+ B=1⇒ B=1− A (7)

2 A + B=0 (8)

Equation (8) – (7) gives:

A=−1

Hence

B=2

Therefore

[ ]
2 1 1 1
− −
( sI− A )−1= (s+1) (s +2) (s +1) ( s+ 2)
−2 1 −1 2
+ +
( s+1 ) (s+2) ( s+1 ) (s +2)

[ ]
−t −2t −t −2 t
2 e −e e −e
Φ ( t )=L [ Φ ( s ) ] =L [ ( sI− A ) ]=
−1 −1 −1
−t −2 t
−2 e +2 e −e−t +2 e−2 t

Non-homogenous system

A non-homogenous system is a system with input that is forced system.

The non-homogenous state equation is given by:

ẋ (t)= Ax (t)+ Bu(t) (1)

Equation (1) can be written as:

ẋ (t)− Ax (t)=Bu(t )

Multiplying both sides by e− At gives:

− At − At
e [ ẋ ( t )− Ax ( t ) ]=e Bu(t )

42
 d − At −At
[e x ( t ) ]=e Bu(t)
dt

Integrating both sides gives:

t
t
[ e− Aτ x ( τ ) ]|0=∫ e− Aτ Bu (τ ) dτ
0

t
x ( t )−x ( 0 )=∫ e
− At −Aτ
e Bu(τ )dτ
0

Multiplying both sides by e At gives:

t
At
x (t )−e x ( 0 )=e
At
∫ e− Aτ Bu(τ)dτ
0

t
x (t )=e x ( 0 ) +∫ e
At A (t −τ)
Bu(τ )dτ
0

t
x (t )=Φ ( t ) x ( 0 )+∫ Φ ( t−τ ) Bu (τ )dτ (2)
0

Equation (2) is the time domain solution of non-homogenous state equation:

Where
At
Φ ( t )=e
A (t−τ )
Φ ( t−τ )=e

Example 1

Consider the system below:

[ ][ ẋ1 (t )
=
0 1 x 1 (t) 0
+ u(t)
ẋ2 (t ) −8 −6 x 2 (t) 1 ][ ] [ ]
y ( t ) =[ 1 0 ]
[ ]
x 1 (t)
x 2 (t)

x ( 0 )=
[ 10]
u ( t )=1

43
 i. Determine the state transition matrices, Φ ( t ) using Laplace method.
ii. Solve for x (t) and y (t ).
iii. Use Matlab to plot the response y (t )

Solution

A=
[−80 −61 ], B=[ 01], C=[ 1 0 ], x ( 0)=[ 10]
( sI− A ) =[ s 0 ]−[ 0
0 s −8 −6 ] [ 8 s+6 ]
1 s −1
=

[ ] [
[
]
]
s+6 1 s+ 6 1 s+6 1
(
adj sI− A ) −8 s −8 s 2 2
= s + 6 s+ 8 s + 6 s +8
−1
( sI− A ) = =
|sI −A|
|8 s +6| −8
2
s −1 s +6 s+ 8 s
2 2
s + 6 s+ 8 s + 6 s +8

[ ]
s+6 1
(s+ 2)(s+ 4) ( s+2)( s+ 4)
¿
−8 s
(s+ 2)(s+ 4) ( s+2)( s+ 4)

Using partial fraction expansion:

s+6 A B
= +
( s+2)( s+ 4) (s+2) (s+ 4)

( s+ 4 ) A + ( s+2 ) B=s+6

( A+ B ) s + ( 4 A +2 B )=s +6

Comparing coefficients gives:

A+ B=1⇒ B=1− A (1)

4 A +2 B=6 ⇒ 2 A + B=3 (2)

Equation (2) – (1) gives:

A=2

Hence

B=1−2=−1

44
 1 A B
= +
( s+2)( s+ 4) (s+2) (s+ 4)

( s+ 4 ) A + ( s+2 ) B=1

( A+ B ) s + ( 4 A +2 B )=1

Comparing coefficients gives:

A+ B=0⇒ B=− A (3)

4 A +2 B=1 (4)

Equation (4) – 2 x equation (3) gives:

A=0.5

Hence

B=−0.5

−8 A B
= +
( s+2)( s+ 4) (s+2) (s+ 4)

( s+ 4 ) A + ( s+2 ) B=−8

( A+ B ) s + ( 4 A +2 B )=−8

Comparing coefficients gives:

A+ B=0⇒ B=− A (5)

4 A +2 B=−8 ⇒ 2 A + B=−4 (6)

Equation (6) – (5) gives:

A=−4

Hence

B=4

s A B
= +
( s+2)( s+ 4) (s+2) (s+ 4)

( s+ 4 ) A + ( s+2 ) B=s

( A+ B ) s + ( 4 A +2 B )=s

45
Comparing coefficients gives:

A+ B=1⇒ B=1− A (7)

4 A +2 B=0 ⇒ 2 A + B=0 (8)

Equation (8) – (7) gives:

A=−1

Hence

B=2

Therefore

[ ]
2 1 0.5 0.5
− −
( sI− A )−1= (s+2) (s+ 4) (s +2) (s +4)
−4 4 −1 2
+ +
( s+2 ) (s+ 4) ( s+2 ) (s+ 4)

[ ]
−2t −4t −2 t −4 t
2 e −e 0.5 e −0.5 e
Φ ( t )=L
−1
[ Φ ( s ) ]=L−1 [ ( sI− A )−1 ]= −2 t −4 t
−4 e + 4 e −e−2t +2 e−4 t
t
x (t )=Φ ( t ) x ( 0 )+∫ Φ ( t−τ ) Bu ( τ ) dτ
0

[ ][ ] [ ][ ]
−2 t −4 t −2 t −4 t t −2(t −τ) −4 (t −τ ) −2(t−τ ) −4(t −τ)
2 e −e 0.5 e −0.5 e 1 2e −e 0.5 e −0.5 e 0
¿ −2 t −4 t +∫ dτ
−4 e + 4 e −e−2t +2 e−4 t 0 0 −4 e−2 (t−τ )
+4 e
−4(t −τ)
−e
−2 (t −τ)
+2 e
−4 (t−τ )
1

[ ] ∫[ ]
−2 t −4 t t −2 (t −τ ) −4 (t−τ )
2 e −e 0.5 e −0.5 e
¿ −2 t −4 t + −2(t−τ ) −4 (t −τ )
dτ
−4 e + 4 e 0 −e +2 e

[ ][ ]|
t
−2 t −4 t −2 (t −τ) −4 (t −τ )
2 e −e 0.25 e −0.125 e
¿ −2 t −4 t + −2 ( t −τ ) −4 (t −τ )
−4 e + 4 e −0.5 e + 0.5 e 0

[ ] [[ ][ ]]
−2 t −4 t −2(t−t ) −4 (t−t ) −2(t −0 ) −4 (t−0 )
2 e −e 0.25 e −0.125 e 0.25 e −0.125 e
¿ −2 t −4 t + −2 ( t−t ) −4 (t−t ) − −2 (t −0) −4 (t−0)
−4 e + 4 e −0.5 e +0.5 e −0.5 e +0.5 e

[ ] [[ ][ ]]
−2 t −4 t −2t −4 t
2 e −e 0.25−0.125 0.25 e −0.125 e
¿ −2 t −4 t + −
−4 e + 4 e −0.5+0.5 −0.5 e−2 t +0.5 e−4 t

46
 [ ][ ]
−2 t −4 t −2 t −4t
2 e −e 0.125−0.25 e +0.125 e
¿ −2 t −4 t +
−4 e + 4 e 0.5 e−2 t−0.5 e−4 t

[ ]
−2t −4 t
0.125+ 1.75 e −0.875 e
¿
−3.5 e−2 t +3.5 e− 4 t

y ( t ) =C x 1 (t)

[ ]
−2 t −4 t
0.125+1.75 e −0.875 e
y ( t ) =[ 1 0 ]
−3.5 e−2 t +3.5 e−4 t
−2 t −4 t
¿ 0.125+1.75 e −0.875 e

The output can be plotted using the following Matlab code:

t=(0:0.1:30);
y=0.125+1.75*exp(-2*t)-0.875*exp(-4*t);
plot(t,y)
grid
xlabel('t')
ylabel('y(t)')

Running the code the following result will be obtained:

0.9

0.8

0.7

0.6
y(t)

0.5

0.4

0.3

0.2

0.1
0 5 10 15 20 25 30
t

Example 2

For state & output equation and initial state vector shown below, using classical technique:

47
 a. determine the state transition matrices, φ(t) using classical technique.
b. Solve for x(t).
c. Solve for output y(t)

[ ][
ẋ1 (t )
ẋ2 (t )
=
0
−1 −3
2 x 1 (t) 0
][ ] [ ]
+ u(t )
x 2 (t) 1

y ( t ) =[ 1 0 ]
[ ]
x 1 (t)
x 2 (t)

x ( 0 )=
[ 00]
u ( t )=1

Solution

A=
[−10 −32 ], B=[ 01], C=[ 1 0 ], D=0, x ( 0 )= [ 00], u ( t )=1
a.

|λI − A|=0

|[ ] [
|λI − A|= λ 0 − 0
0 λ
2
−1 −3
=
λ
1 ]| | −2
|
λ+ 3
=0

λ ( λ +3 ) +2=0
2
λ +3 λ +2=0
( λ+ 1 )( λ +2 )=0

Hence λ 1=−1 and λ 2=−2

Since each term of the state-transition matrix is the sum of responses generated by the poles
(eigenvalues), we assume a state-transition matrix of the:

[ ]
−t −2 t −t −2 t
k 1 e +k 2 e k3 e + k4 e
Φ ( t )= −t −2 t −t −2 t (1)
k 5 e +k 6 e k7 e + k8 e

Using property (i) that is

Φ ( 0 )=I

[ k 1 +k 2 k 3 + k 4
k 5 +k 6 k 7+ k 8
=
1 0
0 1 ][ ]
48
Hence

k 1+ k 2=1 (2)

k 3+ k 4 =0 ⇒ k 3 ¿−k 4 (3)

k 5+ k 6=0 ⇒ k 5 ¿−k 6 (4)

k 7+ k 8=1 ⇒ k 7 ¿ 1−k 8 (5)

Using property (ii) that is

Φ̇ ( 0 )= A

[ ][ ]
−t −2t −t −2 t
−k 1 e −2 k 2 e −k 3 e −2k 4 e
= 0 2
−t
−k 5 e −2 k 6 e
−2t −t
−k 7 e −2k 8 e
−2 t
−1 −3

−k 1−2 k 2=0⇒ k 1 ¿−2 k 2 (6)

−k 3−2 k 4=2 (7)

−k 5−2 k 6=−1 (8)

−k 7−2 k 8=−3 (9)

Adding (2) and (6) gives:

k 2=−1

Hence

k 1=2

Adding (3) and (7) gives:

k 4=−2

Hence

k 3=2

Adding (4) and (8) gives:

49
 k 6=1

Hence

k 5=−1

Adding (5) and (9) gives:

k 8=2

Hence

k 7=−1

Therefore

[ ]
−t −2 t −t −2 t
2e −e 2 e −2 e
Φ ( t )= −t −2 t
−e +e −e−t +2 e−2 t

b.
t
x (t )=Φ ( t ) x ( 0 )+∫ Φ ( t−τ ) Bu ( τ ) dτ
0

[ ][ ] [ ][ ]
−t −2 t −t −2 t t −(t −τ) −2(t−τ ) −(t −τ ) −2(t−τ )
2 e −e 2e −2 e 0 2e −e 2e −2 e 0
¿ −t −2 t −t −2 t +∫ −(t −τ ) −2(t−τ ) −(t−τ ) −2(t−τ )
dτ
−e +e −e +2 e 0 0 −e +e −e +2 e 1

[ ]
t −(t −τ) −2(t−τ )
2e −2 e
¿∫ −(t−τ ) −2(t−τ )
dτ
0 −e +2 e

[ ]|
t
− ( t −τ ) −2(t −τ)
2e −e
¿
−e − ( t−τ )
+e−2(t −τ) 0

[[ ][ ]]
− ( t−t ) −2 (t −t) ( ) −2(t−0 )
2e −e 2 e− t −0 −e
¿ − ( t −t ) −2 (t −t) − − ( t−0 ) −2(t−0 )
−e +e −e +e

[[ ][ ]]
−t −2 t
2−1 2e −e
¿ −
−1+ 1 −e−t +e−2 t

[ ]
−t −2 t
1−2 e +e
¿
e−t−e−2t

c.
y ( t ) =C x 1 (t)

50
 [ ]
−t −2t
1−2 e + e
y ( t ) =[ 1 0 ]
e−t −e−2 t
−t −2t
¿ 1−2 e + e

Controllability and observability

Controllability

It is possible to have a system in which the state variables have no connection with the input.
Such systems are said to be uncontrollable, because no input can be used to affect their
behaviour.

For example consider the following system:

[ ] [ ][ ] [ ]
ẋ1
ẋ2
=
1 0 x1 0
+ u
0 2 x2 1

ẋ 1=x 1

ẋ 2=2 x 2+u

From the above equations, it can be seen that x 2can be changed by u, but x 1 cannot be effected
by changing u. Hence the state x 1 is uncontrollable. Therefore the system is uncontrollable.

Consider the following system:

[ ] [ ][ ] [ ]
ẋ1
ẋ2
=
1 3 x1 0
+ u
0 2 x2 1

ẋ 1=x 1+ 3 x 2

ẋ 2=2 x 2+u

From the above equations, it can be seen that x 2can be changed by u directly and x 1 can be
controlled by u indirectly through x 2, hence the states x 1 and x 2 are controllable. Therefore the
system is controllable.

Test for state controllability

Kalman’s Test

Consider a system:

ẋ= Ax+ Bu

y=Cx

51
The system is completely controllable if and only if the controllability matrix
2 n−1
Q=[B AB A B … … … A B] has a full rank (rank of n). Where n is the number of states.

Example 1

Determine the controllability of the following systems:

1. [ ] [ ][ ] [ ]
ẋ1
=
2 1 x1 1
+ u
ẋ2 −1 2 x 2 1

2. [ ] [ ][ ] [ ]
ẋ1
ẋ2
=
2 0 x1 1
+ u
1 1 x2 1

Solution

1.
A=
[−12 12], B=[ 11]
Q= [ B AB ] =
[[ 11] [−12 12][ 11]]=[ 11 31]
detQ=|
1 1|
1 3
=1−3=−2≠ 0

Since detQ ≠0 , the matix Q has full rank, hence the system is completely controllable.

2.
A=
[ ] []
2 0
1 1
, B=
1
1

Q= [ B AB ] =
[[ 11] [ 21 01][ 11]]=[ 11 22]
detQ=|
1 2|
1 2
=2−2=0

Since detQ=0 , the matix Q has no full rank, hence the system is not completely
controllable.

Gilbert’s Test

This test is for systems that are represented in diagonal form.

Consider the system:

−1 −1
ż=P APz+ P Bu

52
 y=CPz

Recall

[ ]
λ1 0 ⋯ 0
−1 0 λ2 ⋯ 0
P AP=
⋮ ⋯ ⋱ ⋮
0 ⋯ ⋯ λn

The system is said to be completely state controllable if diagonal elements of P−1 AP are distinct
and no row of P−1 B has all zero elements.

Example 2

Determine whether or not the following systems are completely controllable

1. [][
ż2 ][ ] [ ]
ż1 −1 0 z 1 2
= + u
0 −2 z 2 3

2. [][
ẋ1 −1
ẋ2
=
0 −5 ][ x ] [ 0 ]
0 x 1

2
+ u
4

Solution

1. The matrix [−10 −20 ] has distinct diagonal elements and there is no zero row in [ 23]
matrix. Hence the system is completely state controllable.

2. The matrix [
−1 0
0 −5 ]
has distinct diagonal elements, but matrix
4
0 []
has a zero row.

Hence the system is not completely state controllable.

Observability

It is possible to have a system in which some state variables do not affect any of the outputs. In
this case, the systems are said to be unobservable, since some of the state variables cannot be
observed from the output of the system.

For example consider the following system:

[ ] [ ][ ] [ ]
ẋ1
ẋ2
=
1 0 x1 0
+ u
0 2 x2 1

[ ]
[]
y= 0 1
x1
x2

53
 ẋ 1=x 1

ẋ 2=2 x 2+u

y=x 2

From the above equations, it can be concluded that the system is unobservable, since the state x 1
does not affect the output y nor does it affect the state x 2 (which is coupled to the output)

Consider the following system:

[ ] [ ][ ] [ ]
ẋ1
ẋ2
=
1 3 x1 0
+ u
0 2 x2 1

[ ]
[]
y= 0 1
x1
x2

ẋ 1=x 1+ 3 x 2

ẋ 2=2 x 2+u

y=x 2

From the above equations, it can be concluded that the system is observable, since the state x 1
affected the output y indctly through x 2 it affect the state x 2 (which is coupled to the output)

From the above equations, it can be seen that x 2can affect y directly and x 1 can affect y
indirectly through x 2, hence the states x 1 and x 2 are observable. Therefore the system is
observable.

Test for state Observability

Kalman’s Test

Consider a system:

ẋ= Ax+ Bu

y=Cx

54
 [ ]
C
CA
2
The system is completely observable if and only if the observability matrix O= CA has a
⋮
n−1
CA
full rank (rank of n ). Where n is the number of states.

Example 1

Determine the observability of the following systems:

1. [ ] [ ][ ] [ ]
ẋ1
=
2 1 x1 1
ẋ2 −1 2 x 2 1
x
+ u , y= [ 1 1 ] 1
x2 []
2. [ ] [ ][ ] [ ]
ẋ1
ẋ2
=
2 0 x1 1 x
+ u , y= [ 0 1 1
0 1 x2 1 x2
]
[]
Solution

1.

A=
[−12 12], C=[ 1 1 ]

[ ]
1 1
O=
C
[ ]
CA
=
[ 1 1] 2 1 =
−1 2
1 1
[
1 3 ] [ ]
detO= |11 31|=1−3=−2≠ 0
Since detO ≠ 0 , the matix O has full rank, hence the system is completely observable.

3.
A=
[ 20 01], C=[ 0 1]

[ ]
0 1
O= C
[ ]
CA
=
[0 1] 2 0 =
0 1
0 1
0 1 [ ] [ ]
detO= |11 22|=0−0=0
55
 Since detO=0 , the matix O is not of full rank, hence the system is not completely
observable.

Gilbert’s Test

This test is for systems that are represented in diagonal form.

Consider the system:

−1 −1
ż=P APz+ P Bu

y=CPz

The system is said to be completely state observable if diagonal elements of P−1 AP are distinct
and none of the columns of CP has all zero elements.

Example 2

Comment on the observability of the following systems:

1. [ ] [ ][ ] [ ]
ż1 −1 0 z 1 2
ż2
= + u,
0 −2 z 2 3
y= [ 2 3 ]
[]
z1
z2

2. [ ẋ ] 0 −5][ x ]+[ 40] u,

ẋ
1

2
=[
−1 0 x 1

2
y= [ 0 2 ]
[]
z1
z2

Solution

1. The system is completely state observable, because all the diagonal elements of P−1 AP
are distinct and non of the columns of CP are zeros
2. The system is not completely state observable, because a column in CP is zero even
though all the diagonal elements of P−1 AP are distinct.

Stability in the sense of Lyapunov

The stability of LTI (linear time invariant) systems can be determined using classical methods
(i.e. Routh Hurwitz, Nyquest, etc). However, when the system is LTV (linear time variant) or
nonlinear the above criteria may not be successfully applied. The stability of such systems can be
analyzed using Lyapunov methods.

Lyapunov First Method

This method consists of all procedures of stability analysis in which the solution of the
differential equation is required. For nonlinear system, the nonlinearity is linearized around a
given point. If the roots of the linear model have negative real parts, then the original nonlinear
system is stable. If any of the roots is zero, then no conclusion can be drawn.

56
Lyapunov Second Method (Direct Method)

A system is said to be asymptotically stable in the sense of Lyapunov if a positive definite scalar
V (x ) (Lyapunov function) can be found whose first derivative with respect to time is negative
dV (x )
along the trajectory of the system (i.e is negative definite).
dt

Mathematically the Lyapunov direct method can be illustrated as:

{V ( x ) >0 for x ≠ 0
V ( x ) =0 for x=0
positive definite

{V̇ ( x ) <0 for x ≠ 0

V̇ ( x ) =0 for x=0
negative definite

The following example illustrates the Lyapunov direct method

Example 1

Consider the system:

[][
ẋ1
=
0 1 x1
ẋ2 −1 −2 x 2 ][ ]
ẋ 1=x 1

ẋ 2=−x 1−2 x 2

The first step is to find the Lyapunov function V ( x )

Let
2 2
V ( x )=x 1+ x 2

V̇ ( x )=2 x1 ẋ 1+2 x 2 ẋ 2

¿ 2 x1 (x 1 )+ 2 x 2 (−x 1−2 x 2)
2 2 2 2
¿ 2 x1 −2 x 1−4 x 2=−4 x 2

Therefore V ( x ) is positive definite and V̇ ( x ) is negative definite, hence it can be concluded that
the system is asymptotically stable in the sense of Lyapunov.

The task of applying Lyapunov second method is determining the suitable scalar function V ( x ).
Even if V ( x ) is found, the first derivative may not be negative definite.

57
Consider the system:

ẋ= Ax

Let
T
V ( x )=x Px

Where P is a positive definite real and symmetrical matrix.

T T
V̇ ( x )= ẋ Px+ x P ẋ
T T
¿( Ax) Px + x P (Ax )
T T T
¿ x A Px+ x P Ax
T T
¿ x ( A P+ P A) x (1)

It is require that V̇ ( x ) should be negative definite, so let

T
V̇ ( x )=−x Qx (2)

Where Q is a positive definite real and symmetrical matrix.

Equating (1) and (2) gives:

T
A P+ P A=−Q (3)

For simplicity let Q be an identity matrix, hence

T
A P+ P A=−I (4)

The matrix P can be found by solving (4) and the positive definiteness can be checked using
Sylvester’s criterion.

Sylvester’s criterion states that a matrix is positive definite if the determinants of the successive
matrix are positive.

Example 1

Given that [][

ẋ1
=
0 1 x1
ẋ2 −1 −2 x 2 ][ ]
Determine the stability of the system using Lyapunov second method

Solution

58
A=
[−10 −21 ], T
A =
[ 01 −1
−2 ] I=
[ 0 1]
, [p p ]
1 0 P= p p 1

2
2

T
A P+ P A=−I

[ 01 −2 ][ p p ]+[ pp pp ][−10 −21 ]=−[ 10 01]

−1 p p 1

2
2

3
1

2
2

[ p −2 p p −2 p ] [− p p −2 p ]=[−10 −10 ]
1
−p 2 −p
2
+
2
− p p −2 p
3

3
2

3
1

2
2

[ −2 p 2
p 1−2 p 2− p3
p1−2 p2− p3 −1 0
2 p 2−4 p 3
=
0 −1 ][ ]
−2 p 2=−1 ⟹ p 2=0.5 (1)

p1−2 p2− p3=0 (2)

2 p 2−4 p 3=−1 (3)

Substituting (1) in (3) gives:

2 ( 0.5 )−4 p3=−1⟹ p3=0.5 (4)

Substituting (1) and (4) in (2) gives:

p1=2 p2 + p3=2 ( 0.5 ) +0.5=1.5

Hence

P=
[ p1
p2
p2
p3
=
][
1.5 0.5
0.5 0.5 ]
The positive definiteness oh matrix P can be checked using Sylvester’s criterion.

Determinant of first minor ¿ 1.5 which is positive.

Determinant of second minor ¿ [1.5

0.5 0.5 ]
0.5
=0.75−0.25=0.5 which is positive.

P is positive definite, since all the determinants of the successive matrix are all positive, hence
the system is stable

Example 2

59
 Given that [][
ẋ1
ẋ2
=
0 1 x1
−3 4 x2 ][ ]
Determine the stability of the system using Lyapunov second method

Solution

A=
[ 0 1
−3 4 ]
,
T
A =
[ 0 −3
1 4 ] [ ]
I=
1 0 P= p 1
0 1
,
p2 [ p2
p3 ]
T
A P+ P A=−I

[ 0 −3 p1
1 4 p2 ][ ] [ ][
p2 p 1
+
p3 p 2
p2 0 1
p3 −3 4
=− ] [ ]
1 0
0 1

[ −3 p2
p 1 + 4 p2
−3 p 3
+
][
−3 p 2
p 2+ 4 p3 −3 p3
p1 +4 p2 −1 0
p2 +4 p3
=
0 −1][ ]
[ −6 p2
p 1+ 4 p2−3 p3
p 1+ 4 p2−3 p3 −1 0
2 p2 +8 p 3
=
0 −1 ][ ]
−6 p2=−1 ⟹ p2=0.167 (1)

p1 +4 p 2−3 p 3=0 (2)

2 p 2+ 8 p 3=−1 (3)

Substituting (1) in (3) gives:

−1.333
2 ( 0.167 ) +8 p3=−1⟹ p3= =−0.167 (4)
8

Substituting (1) and (4) in (2) gives:

p1=−4 p2 +3 p 3=−4 ( 0.167 ) +3 (−0.167 )=−7 ( 0.167 )=−1.167

Hence

P=
[ p1
p2 p3
=
][
p2 −1.167 0.167
0.167 −0.167 ]
The positive definiteness oh matrix P can be checked using Sylvester’s criterion.

Determinant of first minor ¿−1.167 which is negative.

60
Determinant of second minor ¿ [
−1.167 0.167
0.167 −0.167 ]
=0.195−0.028=0.167 which is positive.

P is not positive definite, since not all the determinants of the successive matrix are all positive,
hence the system is unstable

State Feedback Controller Design

Consider the state space model

ẋ= Ax+ Bu (1)

y=Cx (2)

Equations (1) and (2) can be represented as state diagram as shown below:

In state feedback control, each state variable x i is fed back through a gain k i to form the input to
the system u as illustrated in the figure below:

u=r−k 1 x 1−k 2 x 2−…−k n x n

[]
x1
x2
¿ r + [ −k 1−k 2 −…−k n ]
⋮
xn

61
 ¿ r −Kx (3)

Where r is the reference input, K is the controller gain vector and x is the state vector.

Substituting (3) in (1) gives:

ẋ= Ax+ B( r−Kx)

¿ Ax−BKx+ Br

¿ ( A−BK ) x+ Br

¿ Ac x+ Br

Where Ac = ( A−BK )is the closed loop state matrix.

The closed loop characteristics equation is:

ϕ ( s )=|sI− A c|=0

¿|sI−( A−BK )|=0

Pole placement (pole assignment) design

Pole placement (assignment) is the process of moving the poles of a system to a desired location
so that a desired response is obtained. The poles are moved to a specified location by means of
state feedback.

Let the desired poles be p1 , p2 , … … , p n

The desired characteristic equation can be formed as:

D ( s )=( s− p1 ) ( s− p2 ) … … ..(s− pn )

For pole placement the the closed loop characteristics equation is the same as the desired
characteristic equation, that is:

ϕ ( s )=D( s)

Example 1

Given that

[][
ẋ1 −1 0 x 1 1
ẋ2
= + u
0 −4 x 2 1 ][ ] [ ]

62
 y= [ 1 1 ]
[]
x1
x2

Design a state feedback controller so that the poles are at (-2, -3).

Solution

The desired characteristics equation is:

D ( s )=( s−−2 ) ( s−−3 )=0

( s+2 ) ( s +3 ) =0
2
s +5 s +6=0

The closed loop characteristics equation is:

ϕ ( s )=|sI− ( A−BK )|=0

¿|sI− A+ BK |=0

¿
|[ ] [ s 0
0 s
−
−1 0 1
][]
+ [ k 1 k 2 ] =0
0 −4 1 |
¿
[| s+01 0
][k k
+ 1 2 =0
s+ 4 k 1 k 2 ]|
|
¿
s+1+ k 1
k1
k2
s+ 4+k 2
=0
|
(s+1+ k 1) ( s +4 +k 2) −k 1 k 2=0

2
s +4 s +k 2 s +s +4 +k 2+ k 1 s+ 4 k 1+ k 1 k 2−k 1 k 2=0
2
s + ( 5+ k 1+ k 2 ) s +4 +4 k 1 +k 2=0

For pole placement

ϕ ( s )=D( s)
2 2
s +5 s +6=s + ( 5+k 1+ k 2 ) s+ 4+ 4 k 1+ k 2

Therefore

5+ k 1+ k 2=5 ⟹ k 1 +k 2 =0 (1)

63
 4 + 4 k 1 +k 2 =6 ⟹ 4 k 1+ k 2=2 (2)

Subtracting (1) from (2) gives:

3 k 1=2⟹ k 1 =2/3

Hence

k 2=−2/3

Therefore the controller feedback gain is

K= [ 2 /3 −2/3 ]

Note that pole placement method can only be applied to completely state controllable system.
Therefore before designing a pole placement controller for a particular system, the controllability
of the system must be confirmed.

Example 2

For the plant below, design a controller to yield a10% overshoot and a settling time of 5 second.

[][ ẋ1
=
0 1 x1 1
][ ] [ ]
+ u
ẋ2 −1 −2 x 2 1

y= [ 1 0 ]
[]
x1
x2

Solution

A=
[−12 12], B=[ 11], C=[ 1 0 ], T s=5 s, %OS=10 %

4
T s= =5⟹ ζ ω n=0.8 (1)
ζ ωn

OS=e−ζπ /√ 1−ζ =0.1

2

Taking ln of both sides gives:

−ζπ
=ln ( 0.1 ) =−2.3
√1−ζ 2
Squaring both sides gives:

64
 2 2
ζ π
=5.3 ⟹ ζ π =5.3 ( 1−ζ ) ⟹ ζ π + 5.3 ζ =5.3
2 2 2 2 2 2
2
1−ζ

Hence

ζ=
√ 2
5.3
π +5.3
=0.59 (2)

Substituting (2) in (1) gives:

0.8 0.8
ω n= = =1.36
ζ 0.59

The general second order (desired) characteristics equation is given by:

2 2
D ( s )=s +2 ζ ω n s+ ωn=0
2 2
¿ s +2(0.59)(1.36)s +(1.36) =0
2
¿ s +1.6 s+1.9=0

The closed loop characteristics equation is:

ϕ ( s )=|sI− ( A−BK )|=0

¿|sI− A+ BK |=0

|[ ] [
¿
s 0
0 s
−
0 1 1
][]
+ [ k k ] =0
−1 −2 1 1 2 |
[| 1s
¿
−1
][
k k
+ 1 2 =0
s +2 k 1 k 2 ]|
¿
| s +k 1 k 2−1
k 1+1 s +2+k 2
=0
|
( s+ k 1 )( s +2+k 2 ) −(k 1+1)(k 2−1)=0
2
s +2 s +k 2 s +k 1 s +2 k 1 +k 1 k 2−k 1 k 2+ k 1−k 2+ 1=0
2
s + ( 2+ k 1+ k 2 ) s +1+3 k 1−k 2=0

For pole placement

ϕ ( s )=D( s)

65
Therefore

2+k 1+ k 2=1.6 ⟹ k 1 +k 2=−0.4 (3)

1+3 k 1−k 2=1.9 ⟹ 3 k 1−k 2=0.9 (4)

Adding (3) and (4) gives:

4 k 1=0.5 ⟹ k 1=0.125

Hence

k 2=−0.4−0.125=−0.525

Therefore the controller feedback gain is

K= [ 0.125 −0.525 ]

Full-order state observers

Controller design upon relies on the state variables for feedback through the gain K . However,
sometimes due to system configuration or cost, the state variables (or some of the state variables)
cannot be accessed (measured). These inaccessible states can be estimated using an observer.

An observer (estimator) is used to calculate state variables that are not accessible from the plant
as illustrated in the block diagram below.

66
The Observer is a model of the plant as illustrated in the figure below:

Where

y is plant output

^y is estimated output

^x is estimated state vector

From the above diagram the observer model can be written as:
^ẋ= A ^x +Bu+ L( y−^y ) (1)

67
 ^y =C ^x (2)

The plant model is given as:

ẋ= Ax+ Bu (3)

y=Cx (4)

The error between the estimated model (observer) and the actual model (plant) is:

¿
^ẋ− ẋ= A ( x^ −x ) + L( y −^y ) (5)

Substituting (2) and (4) in (5) gives:

^ẋ− ẋ= A ( x^ −x ) + L(Cx−C x^ )

^ẋ− ẋ= A ( x^ −x )−LC ( x^ −x )

^ẋ− ẋ=( A−LC ) ( x^ −x )

γ̇ =( A− LC)γ (6)

The eigenvalues of ( A−LC) which are the poles of the observer can be found from the
characteristics equation:

ϕ ( s )=|sI− ( A−LC )|=0 (7)

Similar to pole placement controller design, the characteristic equation obtained the will be
equated with the desired characteristic equation for the observer to obtain the observer matrix L.

Note that observer design can only be applied to observable system. Therefore before designing
a observer for a particular system, the observability of the system must be confirmed.

Example 1

Given that: [][

ẋ1
ẋ2
=
0 1 x1 0
+ u
0 −1 x 2 1 ][ ] [ ]
y= [ 1 0 ]
[]
x1
x2

Design a full order state observer for the plant with poles at [-5 -6]

68
A=
[ 00 −11 ], C=[ 1 0 ], L=
[]
l1
l2

ϕ ( s )=|sI− ( A−LC )|=0

¿|sI− A+ LC|=0

|[ ] [ ] [ ] |
l
¿ s 0 − 0 1 + 1 [ 1 0 ] =0 k
0 s 0 −1 l 2

|[ ] [ ]|
l 0
¿ s −1 + 1
0 s+1 l 2 0
=0

|
¿
s+l 1 −1
l2 |
s+ 1
=0

( s+l1 ) ( s +1 ) +l2=0
2
s + ( l 1 +1 ) s +l 1+l 2=0

The desired characteristics equation is:

D ( s )=( s−−5 ) ( s−−6 )=0

( s+5 ) ( s+6 )=0

2
s +11 s+30=0

For observer pole placement

ϕ ( s )=D( s)
2 2
s + ( l 1 +1 ) s +l 1+l 2=s +11 s+ 30

Therefore

( l1 +1 )=11⟹ l1=10
l 1 +l 2=30 ⟹l 2=20

Hence

L=
[][ ]
l1
=
10
l 2 20

69
Optimal quadratic Regulator

In optimal control, the controller design is based on a quadratic cost function (performance
index) which is given by:
∞
J=∫ ( x Qx+ u Ru ) dt
T T
(1)
0

Where Q and R are symmetrical positive definite matrices.

The control law is given by:

u=−Kx

Where K=R−1 BT P

P can be obtained by solving the algebraic Ricatti equation which is given by:
T −1 T
A P+ PA+ Q−PB R B P=0

Where P is symmetrical positive definite matrix

The objective in optimal control is to select K that minimizes the performance index J . Since the
plant is linear and the performance index is quadratic, the problem of determining K to minimize
J is called linear quadratic regulator (LQR).

Example 1

Consider the plant:

[ ] [ ][ ] [ ]
ẋ1
ẋ2
=
0 1 x1 0
+ u
0 0 x2 1

Design an LQR with performance index J=∫ ( x Qx+ u Ru ) dt

T T

Where Q is an identity matrix and R=1.69.

Solution

A=
[ ]
0 1
0 0
, B=
[] [ ]
0
1
, Q=
1 0
0 1
, R=1.69, P=
[ p1
p2
p2
p3 ]
T −1 T
A P+ PA+ Q−PB R B P=0

70
 [ p1
][ ] [
p2 0 p
[0 1] 1
p2
]
[ ][ ][ ][ ] [ ] [ 00 00]
0 0 p1 p2 p 1 p2 0 1 1 0 p2 p3 1 p2 p3
+ + − =
1 0 p2 p3 p 2 p3 0 0 0 1 1.69

[] p2
[ p2 p3 ]

[ ][ ][ ] [00 00]
0 0 0 p1 1 0 p3
+ + − =
p1 p2 0 p2 0 1 1.69

[ ][ ][ ]
2
1 p1 0.59 p 2 0.59 p2 p 3
− =0 0
p 1 2 p 2+1 0.59 p2 p3 0.59 p3
2
0 0

[ ][ ]
2
1−0.59 p 2 p1−0.59 p2 p 3
=0 0
p 1−0.59 p2 p3 2 p2 +1−0.59 p 0 0 2
3

2 2
1−0.59 p2 =0 ⟹ p2 =1.69⟹ p2=1.3 (1)

p1−0.59 p2 p 3=0 (2)

2
2 p 2+ 1−0.59 p3 (3)

Substituting (1) in (3) gives:

2 2
2 ( 1.3 ) +1−0.59 p3 ⟹ p 3=6.1 ⟹ p3 =2.47 (4)

Substituting (1) and (4) in (2) gives:

p1−0.59 ( 1.3 )( 2.47 )=0 ⟹ p 1=1.89 (5)

Hence

P=
[ p1
p2
p2
p3
=
][
1.89 1.3
1.3 2.47 ]
K=R B P=
−1 T
[ 1.3
[ 0 1 ] 1.89 1.3
2.47 ] = [ 1.3 2.47 ]
=[ 0.77 1.46 ]
1.69 1.69

The control law is given by:

u=−Kx= [ 0.77 1.46 ]

[]
x1
x2

71
Example 2

A multivariable system is given by the following set of differential equation:

da
−2 b=u1+ 2u 2
dt

db
−a=0
dt
∞

Design a linear quadratic regulator with the performance index J=∫ ( x Qx+ 2u ) dt . Where Q is
T 2

identity matrix.

Solution

ȧ−2 b=u 1+2 u2

ḃ−a=0 } (1)

Let

x 1=a
x 2=b } (2)

ẋ 1=ȧ
ẋ 2=ḃ } (3)

Substituting (2) and (3) in (1) gives:

ẋ 1−2 x2 =u1+ 2u 2

ẋ 1=2 x2 +u1 +2 u2 (4)

ẋ 2−x 1=0

ẋ 2=x 1 (5)

Put (4) and (5) in matrix form:

[ ] [ ] [ ] [ ][ ]
ẋ1
ẋ2
=
0 2 x 1 1 2 u1
+
1 0 x 2 0 0 u2

72
A=
[ 01 20], T
A =
[ 02 10], B=
[ 10 20 ], T
B =
[ 12 00], Q=[ 10 01], R=2, P=
[ p1
p2
p2
p3 ]
T −1 T
A P+ PA+ Q−PB R B P=0

[ p1
] [ ][ ][
p2 1 2 1 0 p1 p2
]
[ ][ ][ ][ ] [ ] [ 00 00 ]
0 1 p1 p2 p 1 p2 0 2 1 0 p2 p3 0 0 2 0 p2 p3
+ + − =
2 0 p2 p3 p 2 p3 1 0 0 1 2

[ p 1 2 p 1 p1
][
p2
]
[ ][ ][ ] [ 00 00 ]
p2 p3 p 2 p1 1 0 p 2 2 p 2 2 p1 2 p2
+ 2 + − =
2 p1 2 p2 p 3 2 p 2 0 1 2

[ ][
2
5 p1 5 p1 p2

[ ]
2

]
2 p2 +1 p 3+ 2 p1 5 p1 p2 5 p2 0 0
− =
p 3+ 2 p1 4 p2 +1 2 0 0

[ ][ ]
2
2 p2 +1−2.5 p1 p3 +2 p1−2.5 p1 p 2
=0 0
p 3+ 2 p1 −2.5 p 1 p2 4 p2 +1−2.5 p
2
2
0 0

2
2 p 2+ 1−2.5 p1=0 (6)

p3 +2 p 1−2.5 p 1 p2=0 (7)

2
4 p 2+ 1−2.5 p2=0 (8)

From (8)
2
5 p2−8 p2−2=0

Therefore

p2=1.82 or −2.2 (9)

Substituting (9) in (6) gives:

2
2 ( 1.82 )+ 1=2.5 p1

Hence

p1=1.36 (10)

Substituting (9) and (10) in (7) gives:

73
 p3=2.5 ( 1.36 )( 1.82 )−2 ( 1.36 )=3.47

Hence

P=
[ p1
p2
p2
p3 ][
=
1.36 1.82
1.82 3.47 ]
[ ][
1 0 1.36 1.82 1.36 1.82
][ ]
−1 T
K=R B P=
2 0 1.82 3.47
2
=
2.72 3.64
2
=
0.68 0.91
1.36 1.82 [ ]
The control law is given by:

u=−Kx= [ 0.68
1.36 ][ ]
0.91 x 1
1.82 x 2

Controller Design for Steady-State Error (via Integral Control)

Recall: Steady state error ( e ss ) is the difference between input and output for a prescribed test
input as t → ∞

The steady-state error for a STEP input is given by:

74
With state feedback controller, u=−Kx , sometimes we encounter the problem of steady-state
error i.e.

steady−state value of output ≠ setpoint

This problem can be overcome by using the configuration in the figure below:

A feedback path from the output has been added to form the error, e , which is fed forward to the
controlled plant via an integrator. The integrator increases the system type and reduces the
previous finite error to zero.

We will now derive the form of the state equations for the system in the above figure and then
use that form to design a controller. Thus, we will be able to design a system for zero steady-
state error for a step input as well as design the desired transient response.

An additional state variable, x N , has been added at the output of the leftmost integrator. The error
is the derivative of this variable.

Now, from the above figure:

x N =∫ (r−C x )dt (1)

ẋ= A x +B u (2)

y=C x (3)

Differentiating (1) with respect to time gives:

ẋ N =−Cx+ r (4)

Equations (2), (3) and (4) can be written as augmented vectors and matrices. Hence

75
 [ ][ ẋ
=
A 0 x
+
ẋ N −C 0 x N 0][ ] [ ] [ ]
B 0
u+ r
1
(5)

y= [ C 0 ]
[ ]
x
xN
(6)

But

u=−K x + K e x N =[−K K e]
[ ]
x
xN
(7)

Substituting (7) in (5) and (6) gives:

[ ][
ẋ
=
A 0 x
+
ẋ N −C 0 x N 0
B
][ ] [ ]
[−K Ke ]
[ ][]
x 0
+ r
xN 1

[ ][
ẋ
=
A
ẋ N −C ]0 [ x ]+[−BK
0 x
N 0
BKe x
0 ][ ] [ ]
0
+ r
xN 1

[ ][
ẋ
ẋ N
=
A−BK
−C 0 ][ ] [ ]
B Ke x 0
+ r
xN 1
(8)

The characteristics equation is:

ϕ ( s )=|sI− Ă|=0

Where

Ă=
[ A−BK
−C
B Ke
0 ] (9)

Thus, the system type has been increased, and we can use the characteristic equation (9) to
design K and K e to yield the desired transient response. Realize, we now have an additional pole
to place. The effect on the transient response of any closed-loop zeros in the final design must
also be taken into consideration. One possible assumption is that the closed-loop zeros will be
the same as those of the open-loop plant. This assumption, which of course must be checked,
suggests placing higher-order poles at the closed-loop zero locations. Let us demonstrate with an
example.

Example 1

Consider the plant:

76
 ẋ=
[0 1
−3 −5
0
x+ u
1 ] []
y= [ 1 0 ] x

a. Design a controller without integral control to yield a 10% overshoot and a settling time
of 0.5 second. Evaluate the steady-state error for a unit step input.
b. Repeat the design of Part a using integral control. Evaluate the steady-state error for a
unit step input.

Solution

a. A= [−30 −51 ], B=[ 01], C=[ 1 0 ], T s=0.5 s, %OS=10 %

4
T s= =0.5 ⟹ ζ ωn=8 (1)
ζ ωn

OS=e−ζπ /√ 1−ζ =0.1

2

Taking ln of both sides gives:

−ζπ
=ln ( 0.1 ) =−2.3
√1−ζ 2
Squaring both sides gives:
2 2
ζ π
=5.3 ⟹ ζ π =5.3 ( 1−ζ ) ⟹ ζ π + 5.3 ζ =5.3
2 2 2 2 2 2
2
1−ζ

Hence

ζ=
√ 2
5.3
π +5.3
=0.59 (2)

Substituting (2) in (1) gives:

8 8
ω n= = =13.56
ζ 0.59

The general second order (desired) characteristics equation is given by:

2 2
D ( s )=s +2 ζ ω n s+ ωn=0
2 2
¿ s +2(0.59)(13.56) s +(13.56) =0

77
 2
¿ s +16 s +183.9=0

The closed loop characteristics equation is:

ϕ ( s )=|sI− ( A−BK )|=0

¿|sI− A+ BK |=0

|[ ] [
¿
s 0
0 s
−
0 1 0
+ [ k k ] =0
−3 −5 1 1 2 ][] |
[| 1+3s
¿
−1
+
][
0 0
s+5 k 1 k 2
=0
]|
¿
| s −1 =0
k 1+ 3 s+5+ k 2 |
( s ) ( s+5+ k 2 )−(k 1 +3)(−1)=0
2
s +5 s +k 2 s +k 1 +3=0
2
s + ( 5+ k 2 ) s+ k 1 +3=0

For pole placement

ϕ ( s )=D( s)

Therefore

5+ k 2=16 ⟹ k 2=11

k 1+3=183.9 ⟹ k 1=180.9

Therefore the controller feedback gain is

K= [ 180.9 11 ]

The controlled plant can be represented as:

ẋ=( A−BK ) x +Br =

[[0 1
−3 −5
0
1 ][] 0
− [ 180.9 11] x + r
1 ] []
¿
[[ 0 1
−3 −5
−
0
][ 0
180.9 11
0
x+ r
1 ]] [ ]
78
 ¿
[ 0 1
−183.9 −16
0
x+ r
1 ] []
y=Cx=[ 1 0 ] x

The steady-state error for a step input is:

[[ ]] [01 ]=0.995
−1
−1 0 1
e ( ∞ )=1+C ( A−BK ) B=1+ [ 1 0 ]
−183.9 −16

b. The transfer function of plant is:

Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|

G ( s )= =
U (s) |sI− A|

(sI− A)=
[ 0s 0s ]−[−30 −51 ]=[3s −1
s+5 ]
[ s−3+5
adj ( sI −A )=
1
s]

|sI− A|=|s −1 |=s ( s +5 )−3 (−1 )=s +5 s+3 2

3 s +5

Therefore the transfer function can be obtained as:

G ( s )=
Y ( s)
=
[ −3 s ][ 1]
[ 1 0 ] s +5 1 0
U (s) 2
s +5 s+ 3

¿
[ s+5 1 ] 0
2
[ 1]
s +5 s+3

1
¿ 2
s + 5 s+3
From the transfer function it can be seen that the system has no finite zero.
From
2
D ( s )=s +16 s+183.9=0
The desired closed loop poles are located at -8.0000 +10.9499i and -8.0000 -10.9499i.

79
 Since the plant has no finite zeros, we assume no zeros for the closed-loop system. The
third pole is selected as -100, which has a real part greater than five times that of the
desired dominant second order poles. The desired third-order closed-loop system
characteristic equatio is:
2
(s+100)(s +16 s+183.9)=0
3 2 2
s +16 s +183.9 s +100 s +1600 s+ 18390=0
3 2
s +116 s +1783.9 s+18390=0 (3)

The closed loop system characteristics equation can be obtained as:

ϕ ( s )=|sI− Ă|=0

[
] [ ][] [ 01] K ]
0 1 −0 k k
Ă=
A−BK
−C [ B Ke
0
= −3 −5 1 1 2
−[ 1 0 ]
[ ]
0
e

[
[¿ −30 −51 ]−[ k0 k0 ] [ K0 ] = [−3−k
][ ][ ] ][ ]
0 1 0 0 1 0
1 2 e 1 −5−k 2 K e = −3−k 1 −5−k 2 K e
−[ 1 0 ] 0 − [1 0] 0 −1 0 0

|[ ] [ ]| | |
s 0 0 0 1 0 s −1 0
ϕ ( s )= 0 s 0 − −3−k 1 −5−k 2 K e = 3+ k 1 s+5+ k 2 −K e =0
0 0 s −1 0 0 1 0 s

ϕ ( s )=s |
s+5+ k 2 K e
0 s
−−1
1 s | |
3+k 1 −K e
+0
3+k 1 s +5+k 2
1 0
=0 | | |
ϕ ( s )=s ( s 2+5 s+ k 2 s−0 ) + ( 3 s+ k 1 s + K e ) +0=0
3 2
ϕ ( s )=s + ( 5+ k 2 ) s + ( 3+k 1¿ s+ K e ) =0 (4)

Equating (3) and (4) gives:

3 2 3 2
s + ( 5+ k 2 ) s + ( 3+k 1 ¿ s+ K e ) =s +116 s +1783.9 s +18390

Hence

5+ k 2=116⟹ k 2 =111

3+ k 1=1783.9 ⟹ k 1=1780.9

80
 K e =18390

Therefore

K= [ 1780.9 111 ], K e =18390

Substituting the controller gains in (8) gives the closed-loop integral-controlled system:

[ ][
ẋ
ẋ N
=
A−BK
−C 0 ][ ] [ ]
B Ke x 0
+ r
xN 1
(8)

[ ][ [ ][] [ 01] 18390 ][ ] [ ]

ẋ 1 0 1 − 0 [ 1780.9 111 ] x1 0
ẋ 2 = −3 −5 1 x2 + 0 r
ẋ N −[ 1 0 ] 0 xN 1

[] [ [ ][ ] [18390
[]
]] [ ]
ẋ 1 0 1 − 0 0 0 x1 0
ẋ 2 = −3 −5 1780.9 111 x2 + 0 r
ẋ N − [1 0 ] 0 xN 1

[ ][ ][ ] [ ]
ẋ 1 0 1 0 x1 0
ẋ 2 = −1783.9 −116 18390 x 2 + 0 r
ẋ N −1 0 0 xN 1

[]
x1
[ ]
y= [ C 0 ] x = [ 1 0 0 ] x 2
xN
xN

In order to check our assumption for the zero (there is no finite zero), the closed-loop transfer
function can be found as:

18390
3 2
s + 116 s +1783.9 s +18390

Since the transfer function matches our design, we have the desired transient response.

The steady-state error for a step input is:

[ ][]
−1
0 1 0 0
−1
e ( ∞ )=1+C ( A−BK ) B=1+ [ 1 0 0 ] −1783.9 −116 18390 0 =0
−1 0 0 1

Example 2

Consider the plant:

81
 ẋ=
[0 1
−7 −9
0
x+ u
1 ] []
y= [ 4 1 ] x

Design an integral controller for the plant to yield a step response with 10% overshoot, a peak
time of 2 seconds, and zero steady-state error.

Solution

a. A= [−70 −91 ], B=[ 01], C=[ 4 1 ], T p=2 s , %OS=10 %

OS=e−ζπ /√ 1−ζ =0.1

2

Taking ln of both sides gives:

−ζπ
=ln ( 0.1 ) =−2.3
√1−ζ 2
Squaring both sides gives:
2 2
ζ π
=5.3 ⟹ ζ π =5.3 ( 1−ζ ) ⟹ ζ π + 5.3 ζ =5.3
2 2 2 2 2 2
2
1−ζ

Hence

ζ=
√ 2
5.3
π +5.3
=0.59

π
T p=
ωn √ 1−ζ 2

π π
ω n= = =1.95
T p √ 1−ζ 2
2 √ 1−0.592

Hence the desired characteristics equation is:

2 2
D ( s )=s +2 ζ ω n s+ ωn=0
2 2
¿ s +2(0.59)(1.95) s +(1.95) =0
2
¿ s +2.3 s +3.8=0

The transfer function of plant is:

82
 Y ( s ) C [ adj ( sI− A ) ] B+ D|sI− A|
G ( s )= =
U (s) |sI− A|

[ 0s 0s ]−[−70 −91 ]=[ 7s s+−19]

(sI− A)=

adj ( sI −A )=
[ s−7+9 1s ]
|sI− A|=|s −1 |=s ( s+9 )−7 (−1 )=s + 9 s+7 2
7 s+9

Therefore the transfer function can be obtained as:

G ( s )=
Y ( s)
=
[ −7 s ] [ 1]
[ 4 1 ] s+ 9 1 0
U (s) 2
s +9 s+7

¿
[ 4 s+29 s +4 ] 0
2
[ 1]
s +9 s+7

s +4
¿ 2
s + 9 s +3

From the transfer function it can be seen that the system has one finite zero located at -4.
Therefore adding a pole at - 4, which corresponds to the original system’s zero location, yields
the resulting desired characteristic equation as:
2
(s+ 4)( s +2.3 s +3.8)=0
3 2 2
s +2.3 s +3.8 s +4 s + 9.2 s+ 15.2=0
3 2
s +6.3 s +13 s +15.2=0 (1)

The closed loop system characteristics equation can be obtained as:

ϕ ( s )=|sI− Ă|=0

[
] [ ][] [ 01] K ]
0 1 −0 k k
Ă=
[ A−BK
−C
B Ke
0
= −7 −9 1 1 2
−[ 4 1 ]
[ ]
0
e

83
 [
[¿ −70 −91 ]−[ k0 k0 ] [ K0 ] = [−7−k
][ ][ ] ][ ]
0 1 0 0 1 0
1 2 e 1 −9−k 2 K e = −7−k 1 −9−k 2 K e
−[ 4 1 ] 0 −[ 4 1] 0 −4 −1 0

|[ ] [ ]| | |
s 0 0 0 1 0 s −1 0
ϕ ( s )= 0 s 0 − −7−k 1 −9−k 2 K e = 7 +k 1 s+ 9+k 2 −K e =0
0 0 s −4 −1 0 4 1 s

ϕ ( s )=s |
s+ 9+k 2
1
Ke
s | |
−−1
7+ k 1 −K e
4 s
+0 | |
7+ k 1 s+ 9+k 2
4 1
=0 |
ϕ ( s )=s ( s 2+ 9 s+ k 2 s+ K e ) + ( 7 s+k 1 s+ 4 K e ) +0=0
3 2
ϕ ( s )=s + ( 9+k 2 ) s + ( 7+ k 1 + K e ¿ s +4 K e ) =0 (2)

Equating (1) and (2) gives:

3 2 3 2
s + ( 9+ k 2 ) s +(7+ k 1 + K e )s +4 K e =s +6.3 s +13 s +15.2

Hence

9+ k 2=6.3 ⟹ k 2=−2.7

7+ k 1+ K e =13 ⟹ k 1=13−7−3.8=2.2

4 K e =15.2⟹ K e =3.8

Therefore

K= [ 2.2 −2.7 ], K e =3.8

84