Lecture 06 Analysis (II)
Controllability and Observability
6.1 Controllability and Observability
6.2 Kalman Canonical Decomposition
6.3 Pole-zero Cancellation in Transfer Function
6.4 Minimum Realization
Modern Contral Systems 1
Motivation1
x1 2 1 x1 1
x 0 1 x 0u (t )
2 2
x1
y 1 0
x2
x2 (0) x1 (0)
uncontrollable s s
u x2 s 1 x2 1 x1 s 1 x1 1 y
1 2
1
controllable
Modern Contral Systems 2
Controllability and Observability
Plant: x Ax Bu , x R n
y Cx Du
Definition of Controllability
A system is said to be (state) controllable at time t 0 , if
there exists a finite t1 t0 such for any x(t0 ) and any x1 ,
there exist an input u[t0 ,t1 ] that will transfer the state x(t0 )
to the state x1 at time t1 , otherwise the system is said to
be uncontrollable at time t 0 .
Modern Contral Systems 3
Controllability Matrix
A, B Controllab le rank (U ) n,
U B AB A2 B An1 B
det(U ) 0 if u R
Controllab ility Matrix U B AB A2 B An1B
Example: An Uncontrollable System
U (s ) 1 s -1 x1
1 0 1
x x u 1
0 3 0 1
Y (s)
y 1 2x s 3 x2
-1 2
※ State x2 is uncontrollable.
Modern Contral Systems 4
Proof of controllability matrix
xk 1 Axk Bu k
xk 2 Axk 1 Bu k 1
xk 2 A( Axk Bu k ) Bu k 1 A2 xk ABu k Bu k 1
xk n An xk An 1 Bu k An 2 Bu k 1 ABu k ( n 2) Bu k ( n 1)
xk n An xk An 1 Bu k An 2 Bu k 1 ABu k ( n 2 ) Bu k ( n 1)
uk
xk n An xk An 1 B AB B
uk ( n 2)
uk ( n 1)
Initial condition
Modern Contral Systems 5
Motivation2
x1 2 0 x1 3
x 0 1 x 1u (t )
2 2
x1
y 1 0
x2 x2 (0) x1 (0)
s s
u 1 x2 s 1 x2 x1 s 1 x1 1 y
1 2
3
observable
unobservable
Modern Contral Systems 6
Definition of Observability
A system is said to be (completely state) observable at
time t 0 , if there exists a finite t1 t0 such that for any x(t0 )
at time t 0 , the knowledge of the input u[t t ] and the
0, 1
output y[t t ] over the time interval [t0 , t1 ] suffices to
0, 1
determine the state x0 , otherwise the system is said to be
unobservable at t 0 .
Modern Contral Systems 7
Observability Matrix
A, C Observable rank (V ) n det(V ) 0 if y R
C
CA
Observabil ity Matrix V CA
2
CAn 1
Example: An Unobservable System 4
0 1 0
x x u U (s ) 1 s -1 x2 s -1 x1 Y (s)
0 2 1
y 0 4x 2
※ State x1 is unobservable.
Modern Contral Systems 8
Proof of observability matrix
xk 1 Axk Bu k
yk Cxk Du k (1)
yk 1 Cxk 1 Du k 1
yk 1 C ( Axk Bu k ) Du k 1 CAxk CBu k Du k 1 (2)
yk n 1 CAn 1 xk CAn 2 Bu k CAn 3 Bu k 1 CBu k ( n 2 ) Du k ( n 1) (n)
C
CA
(1), (2), (n) xk
n 1
CA
yk Du k yk 1 CBu k Du k 1 CABu k ( n 3) CBu k ( n 2 ) Du k ( n 1)
Inputs & outputs
Modern Contral Systems 9
Example
0 1 0
x Ax Bu , x R n A , B , C 0 1
Plant: 1 0 1
y Cx Du
0 1
Controllab ility Matr ix V B AB
1 0
C 0 1
Obervabili ty Matrix N
CA 1 0
rank (V ) rank ( N ) 2
Hence the system is both controllable and observable.
Modern Contral Systems 10
Theorem I
xc (t ) Ac xc (t ) B c u(t )
Controllable canonical form Controllable
Theorem II
xo (t ) Ao xo (t ) B o u (t )
y (t ) Co xo (t )
Observable canonical form Observable
Modern Contral Systems 11
s2
example T (s)
( s 1)( s 2)
0 1 0
xc xc u
Controllable canonical form 2 3 1
y 2 1xc
0 1
U B AB
1 3
rank [U ] 2 n
C 2
V
1
rank [V ] 1 n
CA 2 1
0 2 2
xo xo u
Observable canonical form 1 3 1
y 0 1xo
2 2
U B AB rank [U ] 1 n
1 1
C 0 1
V
rank [V ] 2 n
CA 1 3 Modern Contral Systems 12
Linear system
Theorem III x (t ) Jx(t ) Bu (t ) 1. Analysis
Jordan form
y (t ) Cx(t ) Du (t )
Jordan block
J1 B1
J J2
B B2 Least row
has no zero
row
J 3 B3
C C1 C2 C3
First column has no zero column
Modern Contral Systems 13
Example
2 1 0 b11
x 0 2 0 x b12 u
0 0 1 1
y c11 c12 3x
If b12 0 uncontrollable
If c11 0 unobservable
Modern Contral Systems 14
0 0 0
0
1 1 1 0 b11
1 0 1 0 b12
1 0 0 1 b13
x 1 x 1 1 2 u
2 1 0 1 0
b21
2 1 0 0 1
2
1 1 2 0 0 2 0
y 1 0 1 2 0 1 1 x
1 0 2 3 0 2 2
C11 C12 C13 C21
Modern Contral Systems 15
b11 b12 b13L.I . b21L.I . controllable
C11 C12 C13L.I . C21L.I . observable
In the previous example
b11 b12 b13L.I . b21L.I . controllable
C11 C12 C13L.I . C21L.D. unobservable
Modern Contral Systems 16
Example
2 1 2 1 1
2 2 1 1
2 1 1 1 L.I.
x 2 x 3 2 1u
1 1 1 0 0
1 1 0 1
1 1 0 0
L.I.
2 2 1 3 1 1 1
y 1 1 1 2 0 0 0
1 1 1 1 0 0 0
L.I. L.D.
Modern Contral Systems 17
Kalman Canonical Decomposition
Diagonalization: x Ax Bu
y Cx Du
All the Eigenvalues of A are distinct, i.e. 1 2 3 n
There exists a coordinate transform (See Sec. 4.4) z Tx such that
1 0
Am T 1 AT where Am .
0 n bm1
Bm T 1B
System in z-coordinate becomes
z Am z Bmu bmn
y Cm z Cm CT cm1 cmn
Homogeneous solution of the above state equation is
z (t ) v1e 1t z1 (0) v n e nt z n (0)
If bmi 0 and cmi 0, mode i is controllab le and observable
Modern Contral Systems 18
How to construct coordinate transformation matrix for diagonalization
All the Eigenvalues of A are distinct, i.e. 1 2 3 n
Eigenvecto rs, v1, v2 , ,vn , are independen t.
The coordinate matrix for diagonalization
T [v1, v2 , ,vn ]
Consider diagonalized system z1 1 z1 bm1u
z2 2 z2 bm 2u
zn λn zn bmnu
y cm1z1 cm 2 z2 cmn zn
Modern Contral Systems 19
Transfer function is
n
cmibmi cm1bm1 c b
H ( s) mn mn
i 1 s i s 1 s n
If bmi 0 or cmi 0, mode i is uncontroll able or unobservab le,
H(s) has pole-zero cancellation.
bm1 cm 1
1
u (t )
bm 2 cm 2
y (t )
2 ∑
bmn cmn
n
Modern Contral Systems 20
Kalman Canonical Decomposition
SC O
u(t )
SCO SC O
y (t )
SCO
SCO : Controllab le, Observable Subsystem
SCO : Controllab le, Unobserva ble Subsystem
SC O : Uncontroll able, Observable Subsystem
SC O : Uncontroll able, Unobserva ble Subsystem
Modern Contral Systems 21
Kalman Canonical Decomposition: State Space Equation
xCO ACO 0 0 0 xCO BCO
x 0 ACO 0 0 xCO BCO
CO u (5.X)
xC O 0 0 AC O 0 xC O 0
xC O 0 0 0 AC O xC O 0
y CCO 0 CC O 0x
Modern Contral Systems 22
Example
1 0 0 b11
x 0 2 0 x b12 u
Plant: 0 0 3 b13
y c11 c12 c13 x
If b13 0, mode 3 is uncontrollable.
If c13 0, mode 3 is unobservable.
The same reasoning may be applied to mode 1 and 2.
Modern Contral Systems 23
Pole-zero Cancellation in Transfer Function
From Sec. 5.2, state equation
x Ax Bu
y Cx Du
may be transformed to
n
Cmibmi Cm1bm1 Cmnbmn
H ( s)
i 1 s i s 1 s n
If bmi 0 , mode i is uncontroll able and vanishes in T.F..
If cmi 0, mode i is unobservab le and vanishes in T.F..
Hence, the T.F. represents the controllable and observable
parts of the state variable equation.
Modern Contral Systems 24
Example
4 0 2
x x u
Plant: 1 2 1 1 2, 2 4
y 0 1x
Transfer Function
H ( s) C sI A B
Y ( s) 1
U ( s)
s 2 0 2
0 1
1
1
s 2s 4 1 s 4
2s 2
0 1
1
s 2s 4 s 2
1
s4
Mode "-2" vanishes in T.F..
Modern Contral Systems 25
Example 5.6
0 6 2
x x u 1 2, 2 -3
Plant: 1 1 1
y 0 1x
Transfer Function
T ( s ) C sI A B
Y ( s) 1 1
U ( s) s3
Mode "2" vanishes in T.F..
Modern Contral Systems 26
Minimum Realization
Realization:
Realize a transfer function via a state space equation.
Example 1
Realization of the T.F. T ( s)
s3
Method 1: U (s ) 1 s -1 1 Y (s)
Y ( s) 1
T ( s)
U ( s) s3 3
Method 2:
U (s ) 1 s -1 1 Y (s)
Y ( s) 1 s2
T ( s)
U ( s) s3 s2
3 1
s -1
※There is infinity number of realizations for
a given T.F. .
Modern Contral Systems 27
2
Minimum realization:
Realize a transfer function via a state space equation
with elimination of its uncontrollable and unobservable parts.
Example 5.8
5
Realization of the T.F. T ( s )
s3
U (s ) 1 s -1 5 Y (s)
Y ( s) 5
T ( s)
U ( s) s3
3
Modern Contral Systems 28