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 4

Norms of Signals

Many of the goals of controller design can be expressed in terms of the size of

various signals, e.g., tracking error signals should be made \small", while the

actuator signals should not be \too large". In this chapter we explore some of the

ways this notion of the size of a signal can be made precise, using norms, which

generalize the notion of Euclidean length.

4.1

Definition

There are many ways to describe the size of a signal or to express the idea that a

signal is small or large. For example, the fraction of time that the magnitude of a

signal exceeds some given threshold can serve as a measure of the size of the signal

we could de ne \small" to mean that the threshold is exceeded less than 1% of the

time. Among the many methods to measure the size of a signal, those that satisfy

certain geometric properties have proven especially useful. These measures of size

are called norms.

The geometric properties that norms satisfy are expressed in the framework of

a vector space roughly speaking, we have a notion of how to add two signals, and

how to multiply or scale a signal by a scalar (see the Notes and References).

Suppose is a vector space and :

+

. is a norm

De

nition

4.1:

V

V

!

R

f1g

on if it satises

V

Nonnegativity: ( ) 0,

v

Homogeneity: for ( )

, ( ) =

( ),

v

<

1

v

j

j

v

Triangle inequality: ( + )

( ) + ( ),

v

w

v

w

for all

and

.

2

R

v

w

2

V

69

index-79_1.png

index-79_2.png

index-79_3.png

index-79_4.png

70