Linear Controller Design: Limits of Performance by Stephen Boyd and Craig Barratt - HTML preview

PLEASE NOTE: This is an HTML preview only and some elements such as links or page numbers may be incorrect.
Download the book in PDF, ePub, Kindle for a complete version.

CHAPTER 5 NORMS OF SYSTEMS

( + )

jH

j

!

j

!

(a)

( + )

jH

j

!

j

(0 + )

jH

j

!

j

!

(1 + )

jH

j

!

j

kH

k

1

0

6

kH

k

1

;1

= 1

(b)

The magnitude of a transfer function is shown in (a). The

Figure

5.5

H

gain of the system is the peak magnitude of along the line = .

L

H

s

j

!

2

The exponentially weighted ( = 1)

gain of the system is the peak

a

;

L

2

magnitude of along the line = 1 + , as shown in (b).

H

s

j

!

index-112_1.png

index-112_2.png

index-112_3.png

index-112_4.png

index-112_5.png

index-112_6.png

index-112_7.png

index-112_8.png

5.2 NORMS OF SISO LTI SYSTEMS

103

a

1

+

r

1

+

s

s

(a)

(b)

A realization of (

) can be formed by taking each inte-

Figure

5.6

H

s

;

a

grator in a realization of ( ), shown in (a), and adding a feedback of , as

H

s

a

shown in (b).

5.2.8

Hankel Norm

The Hankel norm of a transfer function is a measure of the e ect of its past input on

its future output, or the amount of energy that can be stored in and then retrieved

from the system. It is given by

hankel

kH

k

(

1 2

)

= sup Z

Z

=

1

( )2

T

( )2

1 ( ) = 0

0 :

z

t

dt

0 w t dt

w

t

t

>

T

T

We can think of in this de nition as an excitation that acts over the time period

w

0

the response or ring of the system after the excitation has stopped is

t

T

( ) for

. An example of a past excitation and the resulting ring is shown in

z

t

t

>

T

gure 5.7.

It is useful to think of the map from the excitation ( ( ) for 0

) to

w

t

t

T

ring ( ( ) for

) as consisting of two parts: rst, the mapping of the excitation

z

t

t

T

into the state of the system at = and then, the mapping from the state of the

t

T

system at = (which \summarizes" the total e ect on the future output that the

t

T

excitation can have) into the output for

. This interpretation will come up

t

T

again when we describe a method for computing

hankel.

kH

k

5.2.9

Example 1: Comparing Two Transfer Functions

In this section we will consider various norms of the two transfer functions

(a)

44 1 3 + 334 2 + 1034 + 390

13 ( ) =

;

:

s

s

s

H

s

6 + 20 5 + 155 4 + 586 3 + 1115 2 + 1034 + 390

(5.17)

s

s

s

s

s

s

(b)

220 3 + 222 2 + 19015 + 7245

13 ( ) =

;

s

s

s

(5.18)

H

s

6 + 29 1 5 + 297 4 + 1805 3 + 9882 2 + 19015 + 7245:

s

:

s

s

s

s

s

These transfer functions are the I/O transfer functions ( ) achieved by the con-

T

trollers (a) and (b) in the standard plant example of section 2.4. The step

K

K

index-113_1.png

index-113_2.png

index-113_3.png

index-113_4.png

index-113_5.png

index-113_6.png

index-113_7.png

index-113_8.png

index-113_9.png

index-113_10.png

index-113_11.png

index-113_12.png

index-113_13.png

index-113_14.png

index-113_15.png

index-113_16.png

index-113_17.png

index-113_18.png

index-113_19.png

index-113_20.png

index-113_21.png

index-113_22.png

index-113_23.png

index-113_24.png

index-113_25.png

index-113_26.png

index-113_27.png

index-113_28.png

index-113_29.png

index-113_30.png

index-113_31.png

index-113_32.png

index-113_33.png

index-113_34.png

index-113_35.png

index-113_36.png

index-113_37.png

index-113_38.png

index-113_39.png

index-113_40.png

index-113_41.png

index-113_42.png

index-113_43.png

index-113_44.png

104