1

Introduction to State Space

Methods
Dr. Ed. Palmer
School of Electrical & Electronic Systems
Engineering
 2

Contents
 What is a state?
 State space representation
– The state equations - continuous time
 examples
 Matlab simulation
 Discrete time state equations
 Canonical forms
 3

Contents
 Solution of the State Equations
 Pulse transfer function matrix
– eigenvalues
 4

What is a State?
 A variable which is a member of a set of
states which provide a complete description
of system behaviour
 may or may not have physical significance
 organised into a state vector x
 Provided the value of x at t=0 is known the
state vector completely describes the system
 5

State Space Representations

 If the state vector is n dimensional then the
states are described in an n dimensional
space called the state space
 There also exists a vector of output
variables y which may be m dimensional
 Define an r dimensional vector u called the
input vector
 6

State Equations
 General state equations
x (t )  f  x(t ), u(t ), t  , y (t )  g(x(t ), u(t ), t )
f and g are vector functions which may be nonlinear
and time varying
 7

State Equations
 Linear time varying state equations

x (t )  G (t )x(t )  H (t )u(t ),
y (t )  C(t )x(t )  D(t )u(t )

The matrices G,H,C,D are in general time varying

 8

State Equations
 Linear time invariant state equations

x (t )  Gx(t )  Hu(t ),
y (t )  Cx(t )  Du(t )

The matrices G,H,C,D are in general time varying

 9

Example 1
 Consider a mass on a spring. The equation
of motion is
k
Mx   kx  x  x
M
Define state vector x   x x 
T

 0 1
 x  Gx, G    k 
 m 0 
Define output
y  x  Cx, C  [1 0]
Example 2 - Parallel LC 10

Circuit
 Consider parallel LC circuit supplied by
variable current source

iL v
u L C
 11

Example 2 Parallel LC Circuit

Define state vector
x   iL v 
T

diL diL 1
L v  v
dt dt L
dv dv u iL
u  iL  C   
dt dt C C
 12

Example 2 Parallel LC Circuit

diL 1 dv u iL
 v  
dt L dt C C
 diL   1 0 
 dt   0  
L i L  1 u
       
dv 1 v
   0   C 

 dt   C 
 13

Matlab Simulation
 Simple technique based on ode23.m routine
 example to say illustrate motion of spring
mass system
MATLAB Simulation using 14

ode23.m
Define state vector
 x
x 
 x 
x(1)  position, x(2)  velocity
x (1)  x(2)
k
x (2)   x(1)
m
MATLAB Simulation using 15

ode23.m
Define MATLAB function to compute state
derivatives deriv.m

function xdot  deriv(t, x)

xdot  [x(2);-k * x(1)/m];
MATLAB Simulation using 16

ode23.m
Define MATLAB program to simulate
system over finite time
% define times
tinit  0;
tfinal  20;
%Define initial conditions
x0  [1;0]; % give x and initial extension of 1.0
MATLAB Simulation using 17

ode23.m
Define MATLAB program to simulate
system over finite time
% Now perform simulation
[t, x]  ode23(' deriv' , tinit, tfinal, x0);
%output is anN by 2 array of position and
velocity values
 18

MATLAB Exercise
 Try to simulate the Van Der Pol Oscillator
 
x  x  1 x  x  0
2

 x
Take as a state vector x   
 x 
 
x (1)  x1 1  x 2   x 2 
2

x  2   x1
 19

Discrete State Space Equations

x(k  1 )  Gx(k)  Hu(k)

y (k )  Cx(k )  Du(k )

Note that if the system is time-varying then

these matrices may be functions of sample times.
 20

Block Diagrams
D

x(k) +
x(k+1) y(k)
u(k) +
H z-1I C
+
+
Heavy lines represent
G
vector flows
 21

Canonical Forms
 Any state space representation is not unique
– for example one can re-arrange states to get a
different set of state equations
» a different canonical form
 22

Canonical Form Example

 Consider the spring-mass example - Form
another canonical form by re-arranging
states
original state equations
d  x  x
x   , x   
dt  x   x 
 x 
Define a new state vector x1   
 x
 23

Canonical Form Example

 x  x 
x   , x1   
 x   x
0 1 
x  Tx1 , T   
1 0
d
x  Gx  Hu   Tx1   GTx1  Hu
dt
 24

Canonical Form Example

d
x  Gx  Hu   Tx1   GTx1  Hu
dt
Tx 1  GTx1  Hu
x1  T 1GTx1  T 1 Hu
 25

Canonical Form Example

1 1
x 1  T GTx 1  T Hu
 x  G ˆx Hˆu
1 1

ˆ 1 ˆ
G  T GT, H  T H 1

Similarly the output equation

y  Cx  CTx  C ˆx , Cˆ  CT
1 1
 26

Canonical Form Example

1 1
x 1  T GTx 1  T Hu
 x  G ˆx Hˆu
1 1

ˆ 1 ˆ
G  T GT, H  T H 1

Similarly the output equation

y  Cx  CTx  C ˆx , Cˆ  CT
1 1

Similar derivation for discrete time systems

Canonical Forms for Discrete 27

Time Systems
 Four main forms
– Controllable Canonical Form
– Observable Canonical Form
– Diagonal Canonical Form
– Jordan Canonical Form
Canonical Forms for Discrete 28

Time Systems
 Based on basic difference equation for
discrete time system
y (k )  a1 y (k  1)  a2 y (k  2)  ...  an y (k  n)
 b0u (k )  b1u (k  1)  ...  bnu (k  n)
 29

Controllable Canonical Form

 x1 (k  1)   x1 (k ) 
 x (k  1)   0 1 0 ... 0 
x ( k ) 
 2   0 0 1 ... 0  
2 
.   . 
  .  
.   0 0 0 ... 1 
 . 
 xn 1 (k  1)   xn 1 (k )
   an  an 1  an  2 ...  a1   
 xn (k  1)   xn (k ) 
0 
0 
 
. 
  u (k )
. 
0 
 
1 
 30

Controllable Canonical Form

 x1 (k ) 
 x (k ) 
 2 
. 
y (k )   bn  anb0 bn 1  an 1b0 ...b1  a1b0   
. 
 xn 1 (k )
 
 xn (k ) 
 b0u (k )
See Problem A-5-1 for derivation
 31

Observable Canonical Form

 x1 (k  1)   x1 (k ) 
 x (k  1)  0 0 ..0 0  an   
 1 0 x ( k )
 2  an 1   
2
..0 0
.   . 
   . . ... . .  
.  0 0 ..1 0  a2  
 . 
 xn 1 (k  1)   xn 1 (k )
  0 0 ..0 1  a1   
 xn (k  1)   xn (k ) 
bn  anb0 
b  a b 
 n 1 n 1 0 
. 
 u (k )
. 
b2  a2b0 
 
b1  a1b0 
 32

Observable Canonical Form

 x1 (k ) 
 x (k ) 
 2 
. 
y (k )   0 0 0 ... 0 1    b0u (k )
. 
. 
 
 xn (k )
 33

Diagonal Canonical Form

 Where system has distinct poles
 x1 (k  1)   x1 (k )  1
 x (k  1)   p1 0 ... 0 0 
x ( k )  1
 2  0 p2 ... 0 0  
2  
.   .  . 
  . . . . .     u (k )
.  .  .  .
. . . . 
.   .  . 
   0 0 ... 0 pn    
 xn (k  1)  xn (k ) 1
 34

Diagonal Canonical Form

 Where system has distinct poles
 x1 (k ) 
 x (k ) 
 2 
. 
y (k )   c1 c2 ... cn     b0u (k )
. 
. 
 
 xn (k )
 35

Jordan Canonical Form

 Assume a multiple pole of order m exists at
z = p1
0 
 x (k  1)   x1 (k )   
  .  0 
.    . 
 xm (k  1)   J 012   xm (k )   
      1 u (k )
 xm 1 (k  1) 0 21 D   xm 1 (k )
1 
.  .   
    . 
 xn (k  1)   xn (k )   
1 
 36

Jordan Canonical Form

 p` 1 0 ... 0 
 0 p 1 ... 0 
 1 
J. . . . .
 
. . . ... . 
 0 0 0 0 p1 
J : M  M  mth order Jordan block
D : ( N  M )  ( N  M )  diag pm 1... pn 
 37

Jordan Canonical Form

O12  M  ( N  M ) matrix of zeros
O 21  ( N  M )  M matrix of zeros
 38

Jordan Canonical Form

y (k )   c1 c2 ... cn  x(k)  b0u (k )

 39

Example
 Represent the following system in the
controllable, observable and diagonal
canonical forms
Y ( z) z 1
 2
U ( z ) z  1.3 z  0.4
 40

Example
Controllable Canonical Form
 x1 (k  1)   0 1   x1 (k )  0
 x (k  1)   0.4  1.3  x (k )  1 u (k )
 2    2   
 x1 (k ) 
y (k )  [1 1] 
 x2 ( k ) 
 41

Example
Observable Canonical Form
 x1 (k  1)  0  0.4  x1 (k )  1
 x (k  1)  1  1.3   x (k )  1u (k )
 2    2   
 x1 (k ) 
y (k )  [0 1] 
 x2 ( k ) 
 42

Example
Diagonal Canonical Form
 x1 (k  1)   0.5 0   x1 (k )  1
 x (k  1)   0      u (k )
 0.8  x2 (k ) 1
 2  
5  2  x1 (k ) 
y (k )  [ ] 
3 3  x2 ( k ) 
 43

Example
Diagonal Canonical Form
This arises since the transfer function can be expanded
as a sum of partial fractions as
5 2
Y(z)
 3  3
U(z) z  0.5 z  0.8
Solving Discrete Time State 44

Space Equations
 Consider the following set of state space
equations
x(k  1)  Gx(k )  Hu(k )
y (k )  Cx(k )  Du(k )

x(1)  Gx(0)  Hu(0)

x(2)  Gx(1)  Hu(1)  G 2 x(0)  GHu(0)  Hu(1)
Solving Discrete Time State 45

Space Equations
x(2)  Gx(1)  Hu (1)  G 2 x(0)  GHu (0)  Hu(1)
x(3)  Gx(2)  Hu (2)  G x(0)  G Hu (0)  GHu (1)
3 2

 Hu (2)
In general k 1
x(k )  G k x(0)   G k  j 1Hu ( j ) k  1,2,3...
j 0

&
k 1
y (k )  CG x(0)  C G k  j 1Hu ( j )  Du(k )
k

j 0
 46

State Transition Matrix

 Note that it is possible to write x(k+1) in
the form:
x(k  1)  Ψ (k )x(0)
Ψ (k  1)  GΨ (k ), Ψ(0)  I N
I N  N  N identity matrix
 47

State Transition Matrix

 The free motion of the system is defined by
the state transition matrix
k 1
x(k )  Ψ (k )x(0)   Ψ (k  j  1)Hu ( j )
j 0
k 1
x(k )  Ψ (k )x(0)   Ψ ( j )Hu(k  j  1)
j 0
 48

State Transition Matrix

 Output Equation
k 1
y (k )  CΨ (k )x(0)  C Ψ (k  j  1)Hu( j )  Du(k )
j 0
k 1
 CΨ (k )x(0)  C Ψ ( j )Hu(k  j  1)  Du(k )
j 0
Z Transform Approach to 49

Solution of State Equations

State equations
x(k  1 )  Gx(k)  Hu(k)
zX( z )  zx(0)  GX( z )  HU ( z )
 zI  G  X( z )  zx(0)  HU( z )
Z Transform Approach to 50

Solution of State Equations

 zI  G  X( z )  zx(0)  HU( z )
 X( z )   zI  G   zx(0)  HU ( z ) 
1

X( z )   zI  G  zx(0)   zI  G  HU ( z )
1 1

x(k )  Z  zI  G  zx(0)   zI  G  HU ( z )
1 1 1
Z Transform Approach to 51

Solution of State Equations

x( k )  Z zI  G  zx(0)   zI  G  HU( z)
1 1 1

x(k )  Z  zI  G  zx(0)  Z  zI  G  HU( z )

1 1 1 1

 G  Z  zI  G  z 
k 1 1

 G HU( j ) Z  zI  G  HU( z )
k 1
k  j 1 1 1

j 0
 52

Example
 Obtain the state transition matrix if

 0 1 1
G  , H   , C  [1 0]
 0.16  1 1
u (k )  1 k  0
1 
x ( 0)    For all k...
 1
 53

Example
Ψ(k )  G  Z k 1
 zI  G  z
1

1
 z 1 
 zI  G   
1

0.16 z  1
 z 1 1 
  z  0.2 z  0.8  z  0.2 z  0.8 
 
  0.16 z 
  z  0.2 z  0.8  z  0.2 z  0.8 
 54

Example

 4 1 5 5 
 3  3 3  3 
 zI  G    z 0.08.2
1 z  0.8
0.8
z  0.2 z  0.8 
1 4 
 3 3 3  3 
 z  0.2 z  0.8 z  0.2 z  0.8 
 55

Example

Ψ ( k )  G k  Z 1  z I  G  z
1

 4 1 5 5 k 
 3   0.2   3   0.8   0.2    0.8 
k k k

 3 3
 0.8 0.8 1 4 
   0.2    0.8 k
k
  0.2    0.8 
k k

 3 3 3 3 
 56

Example

To obtain x(k ) we compute X( z ) and then take

the inverse z - transform
See Example 5 - 2 Ogata
Pulse Transfer Function 57

Matrix
 Consider the state space representation of
a system with n states, r inputs, and m
outputs.
x(k  1)  Gx(k )  Hu(k )
y (k )  Cx(k )  Du(k )
Pulse Transfer Function 58

Matrix
x(k  1)  Gx(k )  Hu (k )
y (k )  Cx(k )  Du(k )
Taking z - transforms
zX( z )  zx(0)  GX( z )  HU ( z )
Y( z )  CX( z )  DU( z )
Pulse Transfer Function 59

Matrix
zX( z )  zx(0)  GX( z )  HU( z )
Y( z )  CX( z )  DU( z )

X( z )   zI  G  HU( z )
1

 
Y( z )  C zI  G  H  D U( z )  F( z )U( z )
1
Pulse Transfer Function 60

Matrix

 
Y( z )  C zI  G  H  D U( z )  F( z )U( z )
1

F( z )  C zI  G  H  D
1

 pulse transfer function matrix

Pulse Transfer Function 61

Matrix
 
Y( z )  C zI  G  H  D U( z )  F ( z )U( z )
1

Now for any matrix A

1 adj  A 
A  ,
det  A 
adj  A   adjoint of A
det  A   determinant of A
Pulse Transfer Function 62

Matrix
Cadj zI  G  H
 F( z )  D
det  zI  G 
 Poles of system are given as roots of
det  zI  G   0
i.e. characteristic equation is det zI  G   0
Pulse Transfer Function 63

Matrix
The roots of
det  zI  G   0
are the eigenvalues of G
i.e. the poles of a system are the eigenvalues of the
state matrix G
 64

Similarity Transformations
 If a new state vector is defined by a
similarity transform such that
x  Px̂, P non - singular, M  M

then the eigenvalues of the new system

are unchanged
 65

Similarity Transformations
Under such a transformation
ˆ  P -1GP
G
ˆ  P 1H
H
ˆ  CP
C
ˆ D
D
 66

Similarity Transformations

ˆ zI  G
Fˆ ( z )  C ˆ  1
ˆ D
H ˆ

 CP zI  P GP  P H  D
1 1 1
 67

Similarity Transformations
ˆF( z )  CP zP  GP  1 H  D

 1
 C zPP  GPP 
1 1
HD
 C zI  G  H  D
1

 F( z )
Note that the characteristic equation is unchanged
 68

Next Week
 Review of Matrix Theory - ref.:Appendix A
Ogata
 Definitions
 Determinants
 Inversion
 Multiplication
 Matrix Calculus
 69

Next Week
 Vectors - linear dependence and
independence
 Vector spaces, basis vectors
 Orthogonal and Unitary transformations
 Matrix and vector norms
 Eigenvalues and Eigenvectors
 Quadratic forms
 70

Next Week
 Pseudoinverses