SYSTEM & CONTROL

CONTENTS
Controllability
Range Space and Null Space
Reachability
Observability
Kalman Decomposition
Lyapunov Stability
Lyapunov Equation and solution, Sylvester Equation and its solution
Pole-Placement
I.
II.

Bass- Gura Method

Ackermans Formulae

Reduced Order Observer

LQR and LQG
LQG as a general H2 Problem

CONTROLLABILITY :
Consider a system with n-dimensional equation
= +
The system is controllable (or in short, pair(A, B) ) if for any initial state 0 and any final
state , there exits an input () such that (0) is transferred to () in finite time.
THEOREM :
For the above system, the following statements are equivalent
1. The n-dimensional pair (A, B) is controllable.
2. The (n n) matrix
()=

is non-singular for any > 0 .

()
()

3. The n n p controllability matrix

C = [ 2 . . . 1 ]
has rank n (full row rank).
4. The n ( n + p ) matrix | | has full row rank at every eigen value of A .
5. If in addition, all eigenvalues of A have negative real parts, then the unique solution of
+ =
is positive definite. The solution is called the controllability Gramian
and can be expressed as
=

Note:
Row rank means the left range space dimension.

 Proof :
34, Assume p is a eigenvector of A corresponding to eigen value . Then
= [0 ]
So for any first part is always zero, but assume that is also zero. Then system is
uncontrollable.
From (3) Rank of [ 2 . . .] is n. That means there is no v such that ;
[ 2 . . . ] = [0 0 0]
but take v = p then [0 2 ] [0 2 ] [0 0 0].
So system is un-controllable.
Assume system is uncontrollable. So there must exist a such that [ ] will lose rank.
Next: Any matrix (A, B) can be decomposed into such that ;

B
Where controllable and , uncontrollable
0
0
parts.
Now [ ]

0
Take 2 is a left eigen vector of and =
, [ ] =[0 0]
2
13, (t)= +
Take Laplace on both sides
0 = + ()
= 1 0 + 1
Take Laplace Inverse on both sides
= 0 +
= 0+

[
0

+ + 2 /2!+]

1 ()
2 ()

= 0 [ 2 . . . ] 1 () /2!
.
.
.

 = [ 2 . . . ]

.
.
.

C
The region in which () can be present is decided by C . . in the range space of C .

NOTE :
Range : By linear transformation the space we can reach is called range space
of a matrix.
Linear Vector Space : Assume , ( )
If two operations (i) +
and (ii) , (scalar , ) are valid.
then the set is called space. So set space.

 Example :
1 1 1
M=
2 2 2
What is the Range ?
, with any x will give component = 0. So all the vector will be in plane. So
1
plane is spanned by
2
0
Null or kernel of a matrix : Set of independent vectors for which =
is null or
0
0
1
kernel of a matrix. y = 1 , 0
1
1
Similarly we have Left null space and left range space.
1
Row space= Range of ( ) => 1
1
2
Left Null = Null of ( ) =>
1

If the matrix is of here n=2, m=3

NOTE :
1. Range and left-null space are orthogonal
2. Analogously Row space and null space are orthogonal
3. Dimension of a space :
Number of independent vectors required to span a space is the dimension and
the vectors are called basis vectors.
3. Rank of a matrix : dimension of range space = dimension of row space.
4. In the example : Rank=1
5. Dimension of range+ Dimension of left-null space=n
6. Dimension of row space+ Dimension of null space=m

Proof :
x t = [B AB A2 B . . . ]

. d
0
.
.
Since (i.e., any where), the range space dimension should be n or the matrix C
have
rank n. End of proof.
3 2 Follow directly from above because

=
=
=

0

)
(
)(
0

+ + 2 /2! + ] [ + + 2 /2! + ] dt

From the above, [ + + 2 /2! + ] has rank n.

So there is no vector such that [ + + 2 /2! + ] 0
So, =

for any 0

[ + + 2 /2! + ][ + + 2 /2! + ] dt 0

 35, =
=

)
(
)(
0

is linear combination of (> 0). Since system is stable.

is finite.
Consider +

= 0 [
+ ]

= 0
[ ]

= [ ]0
= 0- 0 0 =

So is a solution of + + = 0

Uniqueness Say another solution exists 2

+ + = 0

(1)

2 + 2 + = 0

(2)

Subtract equation (2) from (1);

( 2 ) + ( 2 ) = 0
[(

So

2 ) + ( 2

) ]

( 2 )
+ (

[ ( 2 ) ] = 0

[ ( 2 ) ] constant

At t = 0, 2 = constant
t = , 2 = 0 = 2

) =

REACHABILITY :
If the input of a system = + can take sates (solution) from zero to any non-zero value
in finite time, then the system is called Reachable. The difference between Reachability and
Controllability is explained next. Take the following example in discrete domain,
1 1
1
=
,=
0 0
0
1 1
= =
0 0
So rank = 1 System is Uncontrollable
Now see even if the system is not controllable it can be taken to zero in finite time (Step)

0
0 = , () =
0
( + 1) = A () + ()
0
1 1
1

=
+
u
0
0 0
0
= ( + ), in one step we can reach 0 state.
Note: Reachability is a typical property seen in discrete time system, for continuous time system
Reachability and controllability are same