Sei sulla pagina 1di 6

1 Alternate State-Space Representations

As ststed earlier, there are an in…nite number of possible state space realiza-
tions (models) for any given system. Some of these representations (canonical
forms) are more useful than others, they are:

1. Controllable Canonical Form

2. Observable Canonical Form

3. Diagonal Canonical Form

4. Jordan Canonical Form

Given any particular representation, it is possible to transform to any of


the other representations. These forms have various advantages, that simplify
conversion to transfer function models and allow us to "see" the properties
of controllability and observability (topics that have been covered in the last
2 lectures).
Consider the system de…ned by
(n) (n 1) (n) (n 1)
y + a1 y + : : : + an 1 y_ + an y = b0 u + b1 y + : : : + bn 1 u_ + bn u

where u is the input and y is the output. Taking the Laplace transform
of both sides we get:

Y (s)(sn + a1 sn 1
+ : : : + an 1 s + an ) = U (s)(b0 sn + b1 sn 1
+ : : : + bn 1 s + bn )

whcih yields the transfer function:

Y (s) b 0 sn + b 1 sn 1 + : : : + b n 1 s + b n
= n (1)
U (s) s + a1 s n 1 + : : : + an 1 s + an

We will now consider the various canonical state space forms for a system
with the transfer function of equantion 1.

1
1.1 Controllable Canonical Form
The controllable canonical form has the following layout:
2 3 2 32 3 2 3
x_ 1 0 1 0 0 x1 0
6 x_ 2 7 6 0 0 1 0 7 6 x2 7 607
7 6 7 6
6 7 6 7
6 .. 7 6 .. .
. .
. . . .
. 7 6 .. 7 6 .. 7
6 . 7 = 6 . . . . . 7 6 7 + 6 7u
6 7 6 7 6 . 7 6.7
4x_ n 1 5 4 0 0 0 0 1 5 4xn 1 5 405
x_ n an an 1 an 2 a1 xn 1
2 3
x1
6 x2 7
6 7
6 .. 7
y = b n a b
n 0 b n 1 a b
n 1 0 b 1 a 1 0 6 . 7 + b0 u
b
6 7
4xn 1 5
xn

Controllable canonical form is useful for pole placement controller design


techniques (this will be shown later).

Example 1 Consider the system given by


Y (s) s+3
= 2
U (s) s + 3s + 2
Obtain a state space representation in controllable canonical form.

Controllable Canonical, by inspection, a1 = 3; a2 = 2 and b0 = 0; b1 =


1; b2 = 3:

x_ 1 (t) 0 1 x1 (t) 0
= + u(t)
x_ 2 (t) 2 3 x2 (t) 1
x1 (t)
y(t) = 3 1
x2 (t)

Remark 2 Matlab (or Octave) uses the controllable canonical form by de-
fault when converting from a transfer function representation.

>>num = [1 3];
>>den = [1 3 2];
>>sys = tf2ss(num,den) %create an LTI SS object from transfer function

2
>>a =

0 1
-2 -3

b =

0
1

c =

3 1

d = 0

1.2 Observable Canonical Form


The observable canonical form has the following layout:
2 3 2 32 3 2 3
x_ 1 0 0 0 an x1 b n an b 0
6 x_ 2 7 61 0 0 an 1 7 6 7 6 7
6 7 6 7 6 x2 7 6bn 1 an 1 b0 7
6 .. 7 = 6 .. .. .. .. 7 6 . 7 + 6 .. 7u
4.5 4. . . . 5 4 .. 5 4 . 5
x_ n 0 0 1 a1 xn b 1 a1 b 0
2 3
x1
6 x2 7
6 7
6 7
y = 0 0 0 1 6 ... 7 + b0 u
6 7
4xn 1 5
xn

Note that the A matrix for observable canonical form is the transpose of
the A matrix for controllable canonical form.

Example 3 Consider the system given of example 1. Obtain state space


representation in observable canonical form.

Observable Canonical form, by inspection, a1 = 3; a2 = 2 and b0 = 0; b1 =

3
1; b2 = 3:

x_ 1 (t) 0 2 x1 (t) 3
= + u(t)
x_ 2 (t) 1 3 x2 (t) 1
x1 (t)
y(t) = 0 1
x2 (t)

Remark 4 Matlab provides conversion to canonical observable (i.e. com-


panion) form with the command canon.

>>csys = canon(sys,’companion’)
>>a =

0 -2
1 -3

b =

3
1

c =

0 1

d = 0

1.3 Diagonal Canonical Form


Diagonal canonical form can be written providing the denominator poly-
nomila of equation 1 can be factored into distinct roots. In this case, we
have:
Y (s) b 0 sn + b 1 sn 1 + : : : + b n 1 s + b n
= +
U (s) (s + p1 )(s + p2 ) (s + pn )
expanding using partial fractions gives
c1 c2 cn
= b0 + + + +
s + p1 s + p2 s + pn

4
Then the diagonal canonical form is given by
2 3 2 32 3 2 3
x_ 1 p1 0 x1 1
6 x_ 2 7 6 p 7 6 7 6 7
6 7 6 2 7 6 2 7 617
x
6 .. 7 = 6 .. 7 6 .. 7 + 6 .. 7 u
4.5 4 . 5 4 . 5 4.5
x_ n 0 pn xn 1
2 3
x1
6 x2 7
6 7
y = c1 c2 cn 6 .. 7 + b0 u
4.5
xn

Example 5 Convert the transfer function of example 1 to diagonal canonical


form:
Y (s) s+3 s+3
= 2 =
U (s) s + 3s + 2 (s + 1)(s + 2)
so p1 = 1 and p2 = 2,

x_ 1 (t) 1 0 x1 (t) 1
= + u(t)
x_ 2 (t) 0 2 x2 (t) 1
x1 (t)
y(t) = 2 1
x2 (t)

Remark 6 Matlab can convert to diagonal canonical (i.e. modal) form as


well:

>>csys = canon(sys,’modal’)
>>a =

-1 0
0 -2

b =

1
1

c =

5
2 -1

d = 0

1.4 Jordan Canonical Form


Consider the case where the polynomial equation of the denominator of equa-
tion 1 involves multiple roots. The diagonal canonical form can be rewritten
as a modi…ed diagonal form called the Jordan form. Say for example, that
the we have a denominator with three identical roots, p1 = p2 = p3 , then the
factored form of the transfre function is
Y (s) b 0 sn + b 1 sn 1 + : : : + b n 1 s + b n
=
U (s) (s + p1 )3 (s + p4 )(s + p5 ) (s + pn )

partial fraction expansion yields


c1 c2 c3 c4 c5 cn
= b0 + 3
+ 2
+ + + + +
(s + p1 ) (s + p1 ) (s + p1 ) s + p4 s + p5 s + pn

the state space form is then


2 3 2 32 3 2 3
x_ 1 p1 1 0 0 0 x1 0
6 x_ 2 7 6 .. .. 7 6 x2 7 607
6 7 6 0 p1 1 . . 76 7 6 7
6 x_ 3 7 6 76 7 6 7
6 7 6 0 0 p1 0 0 7 6 x3 7 617
6 x_ 4 7 = 6 76 7 + 6 7u
6 7 6 0 0 p4 0 7 6 x4 7 617
6 .. 7 6 . .. .. 7 6 .. 7 6 .. 7
4.5 4 .. . . 5 4 . 5 4.5
x_ n 0 0 0 pn xn 1
2 3
x1
6 x2 7
6 7
y = c1 c2 cn 6 .. 7 + b0 u
4.5
xn

Potrebbero piacerti anche