Vous êtes sur la page 1sur 48

Control System Design-II

Dr. Abdul Qayyum Khan


Department of Electrical Engineering
Pakistan Institute of Engineering and Applied Sciences
P.O. Nilore Islamabad Pakistan
url: www.pieas.edu.pk/aqayyum
Email: aqkhan@pieas.edu.pk

Solving TimeTime-Invariant state equationequation- A review


Consider
x& = Ax + Bu,

(1)

x ( t ) = ???

Let us consider a homogenous state equation


x& = ax
(2)
Let
x ( t ) = b 0 + b1 t + b 2 t 2 + L + b k t k + L

(3)

x& ( t ) = b 1 + 2b 2 t + kb k t k 1 + L

(4)

subsitituting (3) and (4) into (2)


b 1 + 2b 2 t + kb k t k 1 + L = a (b 0 + b 1 t + b 2 t 2 + L + b k t k + L )
we have
b 1 = ab 0
1 2
a b 0 , Similarly
2
1
1
1 k
b 3 = a 3 b 0 = a 3 b 0 , LL b k =
a b0
6
3
k

2b 2 = ab 1 = a 2 b 0 b 2 =

Control System Design-II by Dr. A.Q. Khan

(5)

As

1
x ( t ) = 1 + at + at 2 + L b 0
2

At t = 0, equation (3) im plies


b 0 = x ( 0 ) , therefore

1
x ( t ) = 1 + at + a 2 t 2 + L x ( 0 ) = e at x ( 0 )
2

N ow consider
x& = Ax
(6)
Analogy w ith scalar, w e have
x ( t ) = b 0 + b1 t + b 2 t 2 + L + b k t k + L

(7)

x& ( t ) = b 1 + 2b 2 t + kb k t k 1 + L

(8)

proceeding in sim ilar fashion, w e

1
x ( t ) = 1 + At + A 2 t 2 + L x ( 0 ) = e At x ( 0 ) = (t)x ( 0 )
2

w here ( t ) is state transition m atrix.


3

Control System Design-II by Dr. A.Q. Khan

(9)

At

 Referring to equation (9), the term e is of particular

interest.
 It is referred to as matrix exponential and analogously, it can be

represented as

At

Ak t k
=
k = 0 k!

 This matrix exponential for an

n n converges absolutely for

all finite time.


 Some important properties
d At
e = Ae At
dx
e( A + B) t = e At e Bt

if AB = BA

e(

if AB BA

A + B) t

e At eBt

e At e At = I

Control System Design-II by Dr. A.Q. Khan

Laplace Transform Approach


C onsider equation (6) and taking L aplace T ransform
sX (s) - x (0) = A X ( s ) X ( s ) = ( sI A )

x (0)

1
x ( t ) = L 1 ( sI A ) x ( 0 ) x ( t ) = e A t x ( 0 )

or

x ( t ) = ( t ) x (0 )
w here
( t ) : n n m atrix and is the unique solution
: S tate transition m atr ix. It has all the inform ation about
the free m otion of the system define by x& = A x
(0 ) = I
A lso note
5

-1 ( t ) = e -A t = ( t )

Control System Design-II by Dr. A.Q. Khan

 Case-I: If 1,2,Ln be distinct eigenvalues of A, then ( t )

will contain exponentials e t ,e t ,Le t


 Case-II: If the matrix A is diagonal with 1,2,Ln, then
1

e1t

(t) =
M

0
e2 t
O
0

L 0
O M

n t
L e
L

 Case-III: if there is multiplicity of eigenvalues 1,1, 1,2,Ln

, then ( t ) will contain terms like te t ,t2e t in addition to


1

e1t , e2 t ,L en t

Control System Design-II by Dr. A.Q. Khan

 Properties of ( t )

As

( t ) = e At

( 0 ) = e A(0) = I

( t ) = e = (e

( t1 + t 2 ) = e

( t ) = ( nt )
( t 2 t1 ) ( t1 t 0 ) = ( t 2 t 0 )

At

 An Example

Control System Design-II by Dr. A.Q. Khan

At 1

A ( t1 + t 2 )

= ( t )

=e

A ( t1 ) A ( t 2 )

or 1 ( t ) = ( t )

= ( t1 ) ( t 2 )

0
For x& = Ax, where A=
2
Obtain ( t ) and -1 ( t )
Solution : As
Note that

e At = ( t ) = L-1

s
sI A =
2

( sI A )

1
s + 3

s + 3
( sI A ) =
( s + 1)( s + 2 ) 2
Using partial fraction expansion
1

1
3

1
s

2
1

s + 3 1
1
s+2
-1
-1 s + 1
(t) = L
=
L

2 s
+
+
s
1
s
2
(
)(
)


2 2

s + 1 s + 2
2e -t e 2 t
= -t
2 t
2e 2e
8

e -t e 2 t
,
-2t
t
2e e

Control System Design-II by Dr. A.Q. Khan

2e t e 2 t
(t ) = t
2t
2e 2e

1
1

s + 1 s + 2

2
1

s + 2 s + 1
et e2t

2e 2t e t

Solution to Homogenous State equations


Consider

x& = Ax + Bu (1)

Now
x& Ax = Bu

x n 1

n 1
u

nn
A
B np

Convolution integral

( multiply by e

e At ( x& Ax ) = e At Bu

At

both side )

Integrating both sides


t

e ( x& Ax )d = e
A

At

0
At

x ( t ) x (0) = e
0

d A
e x ( ) d = e A Bu ( )d
Bu ( )d
d
0
0

Bu ( )d x ( t ) = e x ( 0 ) + e
At

0
t

x ( t ) = ( t ) x ( 0 ) + ( t ) Bu ( )d (2)
9

0
Control System Design-II by Dr. A.Q. Khan

A (t )

Bu ( )d

Laplace Transform Approach


Consider
x& = Ax + Bu
The Laplace transform of the above system is
sX ( s ) x ( 0 ) = AX ( s ) + BU ( s ) ( sI A ) X ( s ) = x ( 0 ) + BU ( s )
X ( s ) = ( sI A )

( x ( 0 ) + BU ( s ) )
t

x ( t ) = e x (0) + e
At

A ( t )

Bu ( ) d

For initial time t 0 = t i 0, then


x (t) = e

A ( t ti )

x (0) + e

A ( t )

ti

10

Control System Design-II by Dr. A.Q. Khan

Bu ( ) d

An Example
Consider
x& = Ax + Bu,
where
0 1
0
0
0
A=
,
B
,
x
0
=
,
u
t
step
I/P=
=
=
( ) ()

1
-2
-3
0



1
Find x ( t ) .
Solution :
As
2e -t e 2 t
( t ) = -t
2 t
2e 2e

e -t e 2t

2e -2t e t

x ( t ) = e A ( t ) Bu ( ) d
0

1 t 1 2 t
e + e
x (t) = 2
2

t
2t
e e

11

Control System Design-II by Dr. A.Q. Khan

since x ( 0 ) = 0,

for t<0
for t 0

Computation of

At

e =?: Some Further results

 Laplace Transform method:

( t ) = e = L
At

-1

{(sI A) }
1

Discussed in
Previous slides

 Method of Diagonalization of A:

if D = P 1 A P th en
e At = P e Dt P 1
 Caley-Hamilton Theorem and minimal polynomial theorem
 Minimal polynomials of A involves distinct roots
 Minimal polynomials of A involves repeated roots

12

Control System Design-II by Dr. A.Q. Khan

Cayley-Hamilton Theorem
 Very useful theorem in linear algebra
 after the names of two mathematicians: Arthur Cayley and

William Hamilton
 It states, Every square matrix over commutative ring (Real or
complex) satisfies its own characteristic equation
 For example if A is an n n square matrix with
characteristics polynomial
( ) = n + a1 n 1 + L + a n 1 + a n = 0 ---------- (1)
Then
( A ) = A n + a1A n 1 + L + a n 1A + a n I = 0 -------- (2)
13

Control System Design-II by Dr. A.Q. Khan

 While the CH-Theorem shows that ( ) = 0 , there may be a

polynomial ( ) having lower degree than ( ) , for which


( A ) = 0 . The monic polynomial of lowest degree for which
( A) = 0 is referred to as minimal polynomial of A.
 The minimal polynomial plays important role in the

computation of polynomials of nn matrix.


 CH-Theorem together with minimal polynomials helps in the
computation of function of matrix, e.g. exp(At), Cos(At) etc.

14

Control System Design-II by Dr. A.Q. Khan

Computation of

( A ) = e At

 Case-1: Minimal polynomials of A involves only

distinct roots
 Let us assume that the minimal polynomial of A is m. Then ( A )

can be obtained by solving the following determinant equation


1
1
M

Sylvesters interpolation
formula

1
2
M

1 m m 1
1
A

12
22
M

L 1m 1
L 2 m 1
M
M

m m 1 M m m 1
A2
L A m 1

e1t
e2 t
M =0

------------ (3)

em t
e At

 Expanding this determinant equation about the last column, we

obtain
e At = 0 ( t ) I + 1 ( t ) A + 2 ( t ) A 2 + L + m 1 ( t ) A m 1 ----- (4)
15

Control System Design-II by Dr. A.Q. Khan

and
e1t = 0 ( t ) + 1 ( t ) 1 + 2 ( t ) 12 + L + m 1 ( t ) 1m 1
e2 t = 0 ( t ) + 1 ( t ) 2 + 2 ( t ) 2 2 + L + m 1 ( t ) 2 m 1
M

------ (5)

em t = 0 ( t ) + 1 ( t ) m + 2 ( t ) m 2 + L + m 1 ( t ) m m 1

's can be solved using the equations (5). Then e A t can be


obtained by substituting the computed 's
If A is an nn matrix and has distinct eigenvalues, then the
number of the numbers of k ( t ) to be determined is m=n
If A has multiple eigenvalues, but its minimal has only simple
roots then m < n.

16

Control System Design-II by Dr. A.Q. Khan

An Example
Find e At
0 1

A=
0 2
Solution :
The eigenvalues are 1 = 0 , 2 = 2 .From the theory just discussed, we have
1

1 2
I A

e 1t
e 2 t = 0
e At

substituting 1 = 0 , 2 = 2 in the above determinant


1 0
1
1 2 e 2t = 0
I

e At

Expanding the determinant, we obtain


2e At + A + 2I Ae 2 t = 0 e At =
e At

17

1
=

1
1 e 2 t )
(

e 2t

Control System Design-II by Dr. A.Q. Khan

1
A + 2I Ae 2 t )
(
2

Alternative :
As
e At = 0 (t )I + 1 (t ) A ,
where
0 (t ) + 1 (t )1 = e 1t
0 (t ) + 1 (t ) 2 = e 2 t
Since 1 = 0 , 2 = 2 ; Then
0 (t ) = 1
0 (t ) 2 1 (t ) = e 2 t 1 (t ) =

1
1 e 2t )
(
2

Therefore
e At
18

1
1
2t
= I + (1 e ) A =

2
0

Control System Design-II by Dr. A.Q. Khan

1
2t
1 e )
(

2t
e

Computation of

( A ) = e At

 Case-2: Minimal polynomials of A involves multiple

roots
 Let us assume that the minimal polynomial of A involves three

equal roots ( 1 = 2 = 3 ) and the remaining ( 4 , 5 ,Lm ) are


distinct. Then using Sylvester's criterion, ( A ) can be obtained
as
m 3
1

21

2
( m 1) 1m 2

1
M

12
M

L
L

1m 1
M

e1t
M

1 m
1 A

m2
A2

m3
A3

L
L

m m 1
A m 1

em t
e At

1
M

2
1
3
1

Control System Design-II by Dr. A.Q. Khan

t 2 1t
e
2
te1t

31

19

( m 1)( m 2 )

=0

(6)

 Equation (6) can be expanded in similar fashion as before


e At = 0 ( t ) I + 1 ( t ) A + 2 ( t ) A 2 + L + m 1 ( t ) A m 1

(7)

and

( m 1)( m 2 )
t 2 1t
e = 2 ( t ) + 3 ( t ) 1 + L +
m 1 ( t ) 1m 3
2
2
te1t = 1 ( t ) + 2 2 ( t ) 1 + 3 3 ( t ) 12 + L + m 1 ( t ) 1m 1
e1t = 1 ( t ) + 2 ( t ) 1 + 3 ( t ) 12 + L + m 1 ( t ) 1m 1
e2 t = 1 ( t ) + 2 ( t ) 2 + 3 ( t ) 2 2 + L + m 1 ( t ) 2 m 1
M
em t = 0 ( t ) + 1 ( t ) m + 2 ( t ) m 2 + L + m 1 ( t ) m m 1

Extend the above formulation to other cases; e.g; two or


more sets of multiple roots. Try at least two examples
20

Control System Design-II by Dr. A.Q. Khan

Example: Find e At
2 1 4

A = 0 2 0

0 3 1

Solution :
The eigenvalues are 1 = 2 = 2 , 3 = 1 .From the theory just discussed, we have
e At = 0 (t )I + 1 (t ) A + 2 (t ) A 2 and
te 1t = 1 (t ) + 2 2 (t )1 te 2 t = 1 (t ) + 4 2 (t )
e 2t = 0 (t ) + 1 (t )2 + 2 (t )2 2 e 2 t = 0 (t ) + 21 (t ) + 4 2 (t )
e 3t = 0 (t ) + 1 (t )3 + 2 (t )3 2 e t = 0 (t ) + 1 (t ) + 2 (t )
Solving the above set of equations, we have
0 (t ) = 4e t 3e 2 t + 2te 2 t ,

1 (t ) = 4e t + 4e 2 t 3te 2 t

2 (t ) = e t e 2 t + 3te 2 t
Finally
e At

21

e 2t

2
= 0 (t )I + 1 (t ) A + 2 (t ) A = 0
0

Control System Design-II by Dr. A.Q. Khan

12e t 12e 2 t + 13te 2 t


e 2t
3e t + 3e 2 t

4e t + 4e 2 t

Controllability and Observability


Consider a dynam ical system described by
x ( t ) = A ( t ) x ( t ) + B ( t ) u ( t ),
x (t 0 ) = x 0
y (t ) = C (t ) x (t )
where
x (t ) n : State, x ( t 0 ) n : State at initial tim e t 0 , u ( t ) p : Control input,
y ( t ) m : output/m easurem ent vector. The solution x ( t ) is g iven as
t

x ( t ) = ( t,t 0 ) x ( t 0 ) + ( t, )B ( ) u ( ) d
t0

where (t,t 0 ) is the transition m atrix and it is the solution of


(t,t 0 ) =A ( t ) ( t,t 0 ), ( t 0 ,t 0 ) =In
In this course, we shall study LTI system s of the following form
x ( t ) = Ax ( t ) + Bu (t ),
x (t 0 ) = x 0
y ( t ) = Cx ( t )
The transition m atrix is
(t, t 0 ) =e
22

A ( t-t 0 )

Control System Design-II by Dr. A.Q. Khan

 A continuous time LTI system is said to be


reachable at time t 0 if it is possible by means of

any unconstrained control input vector to transfer


the system from initial state to any other state in a
finite interval of time.
 A continuous time LTI system is said to be
controllable (to the zero state) at time t 0 if it is
possible by means of any unconstrained control
input vector to transfer the system from initial
state x (t 0 ) to origin.
 State Controllability
 Output Controllability

23

A system is said to controllable at time t 0 if it is possible by means of


an unconstrained control vector to transfer from any initial state x (t 0 )
toControl
any other
statebyinDr.aA.Q.
finite
System Design-II
Khan interval of time

T h e o r e m : A n e c e s s a r y a n d s u ffic ie n t c o n d itio n s
fo r s y s te m r e a c h a b ility [c o n tr o lla b ility ] a t tim e is
th a t W ( ,t ) is p o s itiv e d e fin ite fo r s o m e t
w h e r e W (.,.) is a n n n G r a m m ia n m a tr ix d e fin e d b y
t

W ( , t ) =

( , )B ( ) B T ( ) T (t, ) d

 Uniform Reachability/Uniform Controllability

Strong notion
 It refers to the fact that the property is independent of initial
time.

24

Control System Design-II by Dr. A.Q. Khan

Theorem: For LTI systems given in before, the pair (A,B) is


said to be completely state controllable if and only if

rank B AB A n1B = n
Pr oof : The solution is
t

x ( t) = (t) x (0) + (t,)B () u () d, t 0 = 0


0

Apply the definition of state controllability,


i.e., x ( t1 ) = 0, therefore
t1

0= (t1 ) x (0) + ( t1 )B () u () d
0
t1

x (0) = ()B () u () d
0

u sin g () = e At = a0 ()I + a1 () A + a2 () A 2 + + +an1 () A n1


n1

25

= ak () A k

Control System Design-II by


k =Dr.
0 A.Q. Khan

Now
t1 n1

x (0) =

t1

n1

a () A B ()u() d = A B a ()u() d
k

0 k =0

k =0

t1

Let k = ak ()u () d

Controllability matrix

1

n1
2
k
n

1
x (0) = A Bk = B AB A B =
k =0

----- (1)

 For system to be completely state controllable, equation

must holds. This requires that matrix M must have n linearly


independent row/column vectors.
Rank (M)= n
 For u , the dim of M is nnp. The Rank (M) should be n
for CSS.
p

26

Control System Design-II by Dr. A.Q. Khan

Example-I
x 1 1 1 x1 1
=
+ u


x 2 0 1 x 2 0
1 1
Rank (M) = 1 2
M = [B AB ] =
0 0

Hence the system is not completely state controllable

Example-II
x 1 1 1 x1 0
=
+ u


x 2 0 1 x 2 1
0 1
Rank (M) = 2
M = [B AB ] =
1 1

The system is completely state controllable


27

Control System Design-II by Dr. A.Q. Khan

Controllability conditions in s-plane


 Necessary and sufficient conditions: No cancellation

of poles and zeros


 The system losses its controllability during
cancellation of poles and zeros
An Example
X (s)

(s + 2.5)
=
U(s) (s + 2.5)(s 1)

This system is not CSC.


Check it in state space??

28

Control System Design-II by Dr. A.Q. Khan

An Example
x&1 = 3x1 + x2 + u
x&2 = 2 x1 + 1.5 x2 + 4u
Controllability ????

29

Control System Design-II by Dr. A.Q. Khan

A n E x a m p le
x& 1 = 3 x 1 + x 2 + u
x& 2 = 2 x 1 + 1 .5 x 2 + 4 u
y = x1
C o n tro lla b ility ? ? ? ?
S o lu tio n :

A p p a ra n tly Y E S

L e t u s p ro c e e d a s fo llo w s
C h e c k th e c o n tro lla b iliy m a trix . H e re
3
A =
2
M = [B
30

1
1
,B =

1 .5
4
1 1
AB ] =
, R ank (M

4 4

Control System Design-II by Dr. A.Q. Khan

)=1

An Exam ple
x&1 = 3 x1 + x 2 + u
x& 2 = 2 x1 + 1.5 x 2 + 4 u
C ontrollability ????
Apparantly Y ES
Solution:
Let us proceed as follow s
C heck the controllabiliy m atrix. H ere
Interesting
3 1
1
A=
,B =

2 1.5
4
1 1
, R ank ( M ) = 1
M = [ B AB ] =

4 4
31

Control System Design-II by Dr. A.Q. Khan

F ro m th e s ta te -s p a c e m o d e l, w e h a v e
&&
x 1 + 1 . 5 x& 1 2 . 5 x 1 = u& + 2 . 5 u L

(s
Y
U

) (s ) = (s

+ 1 .5 s 2 .5 Y

(s )
(s )

(s

(s
2

+ 2 .5

+ 1 .5 s 2 .5

+ 2 . 5 )U

(s

(s

(s )

+ 2 .5

+ 2 .5

)(s

1)

Remarks
The system can not be controllable if it is not reducible to controllable
canonical form
 B should not be in the eigen-space of A. Alternatively
 B should not be in the null space of A
Conclusion
Actuator position is very important

32

Control System Design-II by Dr. A.Q. Khan

A n o th e r E x a m p l e
x& 1 = x 1 + x 2 + 4 u 1 + 2 u 2
x& 2 = x 2
x& 3 = 2 x 3 + 3 u 1
y = x1
C o n tro lla b ility ? ? ? ?

33

Control System Design-II by Dr. A.Q. Khan

Output Controllability
 This can also be defined in the manner similar to state

controllability.
 The system is said to be completely output controllable if it is
possible to construct an unconstrained control vector u(t)
that will transfer the output y(t) to origin from any initial
condition.
Let us consider
x (t) = Ax (t) + Bu (t),

x (t 0 ) = x0

y (t) = Cx (t)

The output controllability matrix

OM = CB CAB CA n1B

should have a rank m


34

Control System Design-II by Dr. A.Q. Khan

Stabilizability
 For partially controllable systems, if the controllable modes

are stable and the unstable modes are controllable, the


system is said to be stabilizable.

Remember: Controllability for continuous-time LTI system


implies Reachability. However for discrete-time LTI system this
does not hold.

35

Control System Design-II by Dr. A.Q. Khan

Observability and Reconstructability


 Definition: The system is said to be observable, if every state x(t0) can

be determined from the observation y(t) over finite interval of time


t0t t1.
 In completely observable system, every transition in state
affects output (or every element of the output vector in case
of multiple output systems)
 The difficulty, in practice, arise with the use of state feedback
control when all the system states are not available for
feedback.
 The observability concept is useful in solving the problem of
reconstructing the un-measurable states of the system from
the measurable states in shortest possible length of time.

36

Control System Design-II by Dr. A.Q. Khan

Consider

x (t) = Ax (t),

x (t 0 ) = x 0

y (t) = Cx (t)
 Definition: Let y(t;t0,x,u) be the output of the above system to the

initial state x0.The present state x0 is unobservable if the (future)


output y(t;t0,x,0)=0 for all t t0
 Definition: The present state x0 is unconstructible if the (past) output
y ( ; ,x, 0) = 0

The LTI system shown above or the pair (A,C) is


completely observable if and only if it is completely state
reconstructable

37

Control System Design-II by Dr. A.Q. Khan

Complete Observability and Reconstructability


Consider
x (t) = Ax (t),

x (t 0 ) = x 0

y (t) = Cx (t)

The above system (the pair (C,A)) is said to be observable if


and only if the following observability matrix is full rank
OM = C AC

38

Control System Design-II by Dr. A.Q. Khan

n1

(A )

Consider the system describe before, the output is


y (t ) = Ce At x (0 )
Recall that e

At

n 1

(t ) A k , therefore

k =0

n 1

y (t ) = C k (t ) A k x (0 ) = 0 (t )Cx (0 ) + 1 (t )CAx (0 ) + + n 1 (t )CA n 1x (0 )


k =0

CA

= 0 (t ) 1 (t ) n 1 (t ) = (OM )

n 1
CA
OM nmn = Observability matrix
For Complete observability, the observability matrix s hould be Full Rank .
Sometimes in literature


n 1
OM = C A C ( A ) C

is used as observability matrix


39

Control System Design-II by Dr. A.Q. Khan

An Example:
x = Ax + Bu,

y = Cx where
1
0
1
, B = ,C = [1 0 ]
A=
2 1
1
Here
1 1

Rank (OM ) = 2
OM = C A C =
0 1
Hence the system is completely
completely state
state observable
observable

Now Find the obserbabiliy for the following ???


0
0
1
0


x = 0
0
1 x + 0 u,
y = [ 4 5 1] x


6 11 6
1

40

Control System Design-II by Dr. A.Q. Khan

 Analogous to Controllability, the necessary and sufficient

condition for complete observability is that no cancellation


occur in the transfer function or transfer matrix.
 If there is some cancellation of Poles/Zeros, the
observability will be lost.
An Example
0
0
1
0


x = 0
0
1 x + 0 u


6 11 6
1


y = [ 4 5 1] x
Rank of the observability matrix


2
= Rank (OM ) = C A C ( A ) C = 2

41

Control System Design-II by Dr. A.Q. Khan

 Let us see what happens to Poles and zeros cancellation.

Consider the output equation


y = 4 x1 + 5 x1 + x1
Y (s )
Y (s ) = (4 + 5s + s ) X1 (s ) =
= (s + 1)(s + 4)
X1 (s )
2

X1 (s )
Also Computing
from state equations, we have
U (s )
X1 (s )
1
1
= 3
=
U (s ) s + 6s 2 + 11s + 6 (s + 1)(s + 2)(s + 3)
Y (s ) Y (s ) X1 (s )
(s + 1)(s + 4)
=
=
U (s ) X1 (s ) U (s ) (s + 1)(s + 2)(s + 3)

(s + 4)
=
(s + 2)(s + 3)
42

Control System Design-II by Dr. A.Q. Khan

Poles and Zero Cancellation


occurs

 Let us see what happens to Poles and zeros cancellation.

Consider the output equation


y = 4 x1 + 5 x1 + x1
Y (s )
Y (s ) = (4 + 5s + s ) X1 (s ) =
= (s + 1)(s + 4)
X1 (s )
2

X1 (s )
Also Computing
from state equations, we have
U (s )
X1 (s )
1
1
= 3
=
U (s ) s + 6s 2 + 11s + 6 (s + 1)(s + 2)(s + 3)
Y (s ) Y (s ) X1 (s )
(s + 1)(s + 4)
=
=
U (s ) X1 (s ) U (s ) (s + 1)(s + 2)(s + 3)

(s + 4)
=
(s + 2)(s + 3)
43

Control System Design-II by Dr. A.Q. Khan

Poles and Zero Cancellation


occurs

A CSC system may not necessarily be CSO and vice versa


Illusrativde Example

1
0
0
x& =
x + u

0.4 1.3
1
y = [ 0.8 1] x
1
0
M =
; Rank ( M ) = 2 CSC

1 1.3
0.8 0.4
OM =
; Rank ( OM ) = 1 NOT CSO

1 0.5
Show if there is Poles Zero Cancellation or not??????

44

Control System Design-II by Dr. A.Q. Khan

T ry th is o n e
2
x& = 0
0
y = [1

45

Control System Design-II by Dr. A.Q. Khan

0
2
0 x
3
1
1 1] x
0

Duality properties
Symmetry properties exists
Observability (Reconstructability)
Controllability (Reachability)
The Linear time-invariant system
x = Ax + Bu

y = Cx
is the dual of the system (and vice versa)
z = A1z + B1v
y * = C1x*
where A1 = AT ,B1 = CT ,C1 = BT

46

The principle of duality is useful in that it allows the state


reconstruction problem can be casted as dual control problem and
vice versa.
Control System Design-II by Dr. A.Q. Khan

Controllability and Observability indices


Controllability index
Consider the controlability matrix
M = B AB A n1B
If there exists a least integer q such that
Rank( B AB
AB A q1B ) = n
1 q n
Then this least integer refferred to as is called controlability index

47

Control System Design-II by Dr. A.Q. Khan

Controllability and Observability indices


Observability index
Consider the observability matrix
*
* *
*
* n1
OM = C C A C ( A )

If there exists a least integer v, such that


*

*
* *
*
* n1
Rank( C C A C ( A ) ) = n
1 v n

Then this least integer refferred to as is called observability index


*

Very powerful concept


Quite often used in minimum order observer design

48

Control System Design-II by Dr. A.Q. Khan

Vous aimerez peut-être aussi