Global Global

edition edition

edition
Global
Nonlinear Control
For these Global Editions, the editorial team at Pearson has
collaborated with educators across the world to address
a wide range of subjects and requirements, equipping
students with the best possible learning tools. This Global
Edition preserves the cutting-edge approach and pedagogy
of the original, but also features alterations, customization
and adaptation from the North American version.
Nonlinear Control
Hassan K. Khalil

Khalil
This is a special edition of an established
title widely used by colleges and universities
throughout the world. Pearson published this
exclusive edition for the benefit of students
outside the United States and Canada. If you
purchased this book within the United States
or Canada you should be aware that it has
been imported without the approval of the
Publisher or Author.

Pearson Global Edition

KHALIL_1292060506_mech.indd 1 31/07/14 5:23 pm

Nonlinear Control
Global Edition
 Nonlinear Control
Global Edition

Hassan K. Khalil
Department of Electrical and Computer Engineering
Michigan State University
Vice President and Editorial Assistant Project Editor,
Director, ECS: Marcia J. Horton Global Editions: Amrita Kar
Executive Editor: Andrew Gilfillan Art Director: Jayne Conte
Marketing Manager: Tim Galligan Cover Designer: Bruce Kenselaar
Managing Editor: Scott Disanno Cover Image:  c HunThomas/Shutterstock
Project Manager: Irwin Zucker Full-Service Project Management/
Head of Learning Asset Acquisition, Composition: SPi Global
Global Editions: Laura Dent
Assistant Acquisitions Editor,
Global Editions: Aditee Agarwal
Credits and acknowledgments borrowed from other sources and reproduced, with permission, in
this textbook appear on appropriate page within text.

Pearson Education Limited

Edinburgh Gate
Harlow
Essex CM20 2JE
England

and Associated Companies throughout the world

Visit us on the World Wide Web at: www.pearsonglobaleditions.com


c Pearson Education Limited 2015

The rights of Hassan K. Khalil to be identified as the author of this work have been asserted by
him in accordance with the Copyright, Designs and Patents Act 1988.

Authorized adaptation from the United States edition, entitled Nonlinear Control, 1st Edition,
ISBN 978-0-133-49926-1, by Hassan K. Khalil, published by Pearson Education  c 2015.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system,
or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or
otherwise, without either the prior written permission of the publisher or a license permitting
restricted copying in the United Kingdom issued by the Copyright Licensing Agency Ltd, Saffron
House, 6–10 Kirby Street, London EC1N 8TS.

All trademarks used herein are the property of their respective owners. The use of any
trademark in this text does not vest in the author or publisher any trademark ownership rights
in such trademarks, nor does the use of such trademarks imply any affiliation with or
endorsement of this book by such owners.

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

10 9 8 7 6 5 4 3 2 1

Typeset by SPi Global

Printed and bound by Courier Westford

ISBN 10: 1-292-06050-6
ISBN 13: 978-1-292-06050-7
 To my parents
Mohamed and Fat-hia

and my grandchildren
Maryam, Tariq, Aya, and Tessneem
Contents

Preface 11

1 Introduction 13
1.1 Nonlinear Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.2 Nonlinear Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.3 Overview of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 Two-Dimensional Systems 27
2.1 Qualitative Behavior of Linear Systems . . . . . . . . . . . . . . . . . . 29
2.2 Qualitative Behavior Near Equilibrium Points . . . . . . . . . . . . . . 33
2.3 Multiple Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.4 Limit Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.5 Numerical Construction of Phase Portraits . . . . . . . . . . . . . . . . 43
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3 Stability of Equilibrium Points 49

3.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3 Lyapunov’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 The Invariance Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.5 Exponential Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.6 Region of Attraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
3.7 Converse Lyapunov Theorems . . . . . . . . . . . . . . . . . . . . . . . 80
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82

4 Time-Varying and Perturbed Systems 87

4.1 Time-Varying Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.2 Perturbed Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
4.3 Boundedness and Ultimate Boundedness . . . . . . . . . . . . . . . . . 97
4.4 Input-to-State Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

7
8 CONTENTS

5 Passivity 115
5.1 Memoryless Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.2 State Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
5.3 Positive Real Transfer Functions . . . . . . . . . . . . . . . . . . . . . 124
5.4 Connection with Stability . . . . . . . . . . . . . . . . . . . . . . . . . 127
5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130

6 Input-Output Stability 133

6.1 L Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.2 L Stability of State Models . . . . . . . . . . . . . . . . . . . . . . . . 139
6.3 L2 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

7 Stability of Feedback Systems 153

7.1 Passivity Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
7.2 The Small-Gain Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.3 Absolute Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.3.1 Circle Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.3.2 Popov Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 176
7.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180

8 Special Nonlinear Forms 183

8.1 Normal Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
8.2 Controller Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
8.3 Observer Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

9 State Feedback Stabilization 209

9.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
9.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
9.3 Feedback Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . 213
9.4 Partial Feedback Linearization . . . . . . . . . . . . . . . . . . . . . . 219
9.5 Backstepping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
9.6 Passivity-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 229
9.7 Control Lyapunov Functions . . . . . . . . . . . . . . . . . . . . . . . . 234
9.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239

10 Robust State Feedback Stabilization 243

10.1 Sliding Mode Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
10.2 Lyapunov Redesign . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
10.3 High-Gain Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
10.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
CONTENTS 9

11 Nonlinear Observers 275

11.1 Local Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
11.2 The Extended Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . 278
11.3 Global Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
11.4 High-Gain Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
11.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

12 Output Feedback Stabilization 293

12.1 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
12.2 Passivity-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 295
12.3 Observer-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 298
12.4 High-Gain Observers and the Separation Principle . . . . . . . . . . . . 300
12.5 Robust Stabilization of Minimum Phase Systems . . . . . . . . . . . . 308
12.5.1 Relative Degree One . . . . . . . . . . . . . . . . . . . . . . . 308
12.5.2 Relative Degree Higher Than One . . . . . . . . . . . . . . . 310
12.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315

13 Tracking and Regulation 319

13.1 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
13.2 Robust Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
13.3 Transition Between Set Points . . . . . . . . . . . . . . . . . . . . . . 326
13.4 Robust Regulation via Integral Action . . . . . . . . . . . . . . . . . . 330
13.5 Output Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334
13.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

A Examples 341
A.1 Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
A.2 Mass–Spring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343
A.3 Tunnel-Diode Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
A.4 Negative-Resistance Oscillator . . . . . . . . . . . . . . . . . . . . . . 347
A.5 DC-to-DC Power Converter . . . . . . . . . . . . . . . . . . . . . . . . 349
A.6 Biochemical Reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
A.7 DC Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
A.8 Magnetic Levitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353
A.9 Electrostatic Microactuator . . . . . . . . . . . . . . . . . . . . . . . . 354
A.10 Robot Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
A.11 Inverted Pendulum on a Cart . . . . . . . . . . . . . . . . . . . . . . . 357
A.12 Translational Oscillator with Rotating Actuator . . . . . . . . . . . . . 359

B Mathematical Review 361

10 CONTENTS

C Composite Lyapunov Functions 367

C.1 Cascade Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
C.2 Interconnected Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 369
C.3 Singularly Perturbed Systems . . . . . . . . . . . . . . . . . . . . . . . 371

D Proofs 375

Bibliography 381

Symbols 392

Index 394
Preface
This book emerges from my earlier book Nonlinear Systems, but it is not a fourth
edition of it nor a replacement for it. Its mission and organization are different from
Nonlinear Systems. While Nonlinear Systems was intended as a reference and a text
on nonlinear system analysis and its application to control, this book is intended as
a text for a first course on nonlinear control that can be taught in one semester (forty
lectures). The writing style is intended to make it accessible to a wider audience
without compromising the rigor, which is a characteristic of Nonlinear Systems.
Proofs are included only when they are needed to understand the material; otherwise
references are given. In a few cases when it is not convenient to find the proofs in
the literature, they are included in the Appendix. With the size of this book about
half that of Nonlinear Systems, naturally many topics had to be removed. This is
not a reflection on the importance of these topics; rather it is my judgement of what
should be presented in a first course. Instructors who used Nonlinear Systems may
disagree with my decision to exclude certain topics; to them I can only say that those
topics are still available in Nonlinear Systems and can be integrated into the course.
An electronic solution manual is available to instructors from the publisher, not
the author. The instructors will also have access to Simulink models of selected
exercises. The Instructor Resource Center (IRC) for this book (www.pearsonglobal
editions.com/khalil) contains the solution manual, the Simulink models of selected
examples and the pdf slides of the course. To gain access to the IRC, please contact
your local Pearson sales representative.
The book was typeset using LATEX. Computations were done using MATLAB
and Simulink. The figures were generated using MATLAB or the graphics tool
of LATEX.
I am indebted to many colleagues, students, and readers of Nonlinear Systems,
and reviewers of this manuscript whose feedback was a great help in writing this
book. I am grateful to Michigan State University for an environment that allowed
me to write the book, and to the National Science Foundation for supporting my
research on nonlinear feedback control.
Hassan Khalil
Pearson would like to thank and acknowledge Lalu Seban (National Institute
of Technology, Silchar) and Zhiyun Lin (Zhejiang University) for their contribu-
tions to the Global Edition, and Sunanda Khosla (writer), Ratna Ghosh (Jadavpur
University), and Nikhil Marriwala (Kurukshetra University) for reviewing the Global
Edition.

11
Chapter 1

Introduction

The chapter starts in Section 1.1 with a definition of the class of nonlinear state
models that will be used throughout the book. It briefly discusses three notions
associated with these models: existence and uniqueness of solutions, change of
variables, and equilibrium points. Section 1.2 explains why nonlinear tools are
needed in the analysis and design of nonlinear systems. Section 1.3 is an overview
of the next twelve chapters.

1.1 Nonlinear Models

We shall deal with dynamical systems, modeled by a finite number of coupled first-
order ordinary differential equations:

ẋ1 = f1 (t, x1 , . . . , xn , u1 , . . . , um )
ẋ2 = f2 (t, x1 , . . . , xn , u1 , . . . , um )
.. ..
. .
ẋn = fn (t, x1 , . . . , xn , u1 , . . . , um )

where ẋi denotes the derivative of xi with respect to the time variable t and u1 ,
u2 , . . ., um are input variables. We call x1 , x2 , . . ., xn the state variables. They
represent the memory that the dynamical system has of its past. We usually use

13
14 CHAPTER 1. INTRODUCTION

vector notation to write these equations in a compact form. Define

⎡ ⎤ ⎡ ⎤
x1 f1 (t, x, u)
⎢ ⎥ ⎡ ⎤ ⎢ ⎥
⎢ ⎥ u1 ⎢ ⎥
⎢ x2 ⎥ ⎢ ⎥ ⎢ f2 (t, x, u) ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢ u 2 ⎥ ⎢ ⎥
⎢ .. ⎥ ⎢ ⎥ ⎢ .. ⎥
⎢ ⎥
x = ⎢ . ⎥, u = ⎢ ⎥ , f (t, x, u) = ⎢ . ⎥
⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢
..
⎥ ⎢ ⎥
⎢ ⎥ ⎢ . ⎥ ⎢ ⎥
⎢ . ⎥ ⎣ ⎦ ⎢ .. ⎥
⎢ .. ⎥ ⎢ . ⎥
⎢ ⎥ ⎢ ⎥
⎣ ⎦ um ⎣ ⎦
xn fn (t, x, u)

and rewrite the n first-order differential equations as one n-dimensional first-order

vector differential equation
ẋ = f (t, x, u) (1.1)
We call (1.1) the state equation and refer to x as the state and u as the input.
Sometimes another equation,
y = h(t, x, u) (1.2)
is associated with (1.1), thereby defining a q-dimensional output vector y that com-
prises variables of particular interest, like variables that can be physically measured
or variables that are required to behave in a specified manner. We call (1.2) the
output equation and refer to equations (1.1) and (1.2) together as the state-space
model, or simply the state model. Several examples of nonlinear state models are
given in Appendix A and in exercises at the end of this chapter. For linear systems,
the state model (1.1)–(1.2) takes the special form

ẋ = A(t)x + B(t)u
y = C(t)x + D(t)u

Sometimes we consider a special case of (1.1) without explicit presence of an

input u, the so-called unforced state equation:

ẋ = f (t, x) (1.3)

This case arises if there is no external input that affects the behavior of the system,
or if the input has been specified as a function of time, u = γ(t), a feedback function
of the state, u = γ(x), or both, u = γ(t, x). Substituting u = γ in (1.1) eliminates
u and yields an unforced state equation.
In dealing with equation (1.3), we shall typically require the function f (t, x) to
be piecewise continuous in t and locally Lipschitz in x over the domain of interest.
For a fixed x, the function f (t, x) is piecewise continuous in t on an interval J ⊂ R
if for every bounded subinterval J0 ⊂ J, f is continuous in t for all t ∈ J0 , except,
possibly, at a finite number of points where f may have finite-jump discontinuities.
1.1. NONLINEAR MODELS 15

This allows for cases where f (t, x) depends on an input u(t) that may experience
step changes with time. A function f (t, x), defined for t ∈ J ⊂ R, is locally
Lipschitz in x at a point x0 if there is a neighborhood N (x0 , r) of x0 , defined by
N (x0 , r) = {x − x0  < r}, and a positive constant L such that f (t, x) satisfies the
Lipschitz condition
f (t, x) − f (t, y) ≤ Lx − y (1.4)

for all t ∈ J and all x, y ∈ N (x0 , r), where

√ 
x = xT x = x21 + · · · + x2n

A function f (t, x) is locally Lipschitz in x on a domain (open and connected set)

D ⊂ Rn if it is locally Lipschitz at every point x0 ∈ D. It is Lipschitz on a set
W if it satisfies (1.4) for all points in W , with the same Lipschitz constant L. A
locally Lipschitz function on a domain D is not necessarily Lipschitz on D, since
the Lipschitz condition may not hold uniformly (with the same constant L) for all
points in D. However, a locally Lipschitz function on a domain D is Lipschitz on
every compact (closed and bounded) subset of D. A function f (t, x) is globally
Lipschitz if it is Lipschitz on Rn .
When n = 1 and f depends only on x, the Lipschitz condition can be written as

|f (y) − f (x)|
≤L
|y − x|

which implies that on a plot of f (x) versus x, a straight line joining any two points
of f (x) cannot have a slope whose absolute value is greater than L. Therefore,
any function f (x) that has infinite slope at some point is not locally Lipschitz at
that point. For example, any discontinuous function is not locally Lipschitz at the
points of discontinuity. As another example, the function f (x) = x1/3 is not locally
Lipschitz at x = 0 since f  (x) = (1/3)x−2/3 → ∞ as x → 0. On the other hand, if
f  (x) is continuous at a point x0 then f (x) is locally Lipschitz at the same point
because continuity of f  (x) ensures that |f  (x)| is bounded by a constant k in a
neighborhood of x0 ; which implies that f (x) satisfies the Lipschitz condition (1.4)
over the same neighborhood with L = k.
More generally, if for t in an interval J ⊂ R and x in a domain D ⊂ Rn ,
the function f (t, x) and its partial derivatives ∂fi /∂xj are continuous, then f (t, x)
is locally Lipschitz in x on D.1 If f (t, x) and its partial derivatives ∂fi /∂xj are
continuous for all x ∈ Rn , then f (t, x) is globally Lipschitz in x if and only if
the partial derivatives ∂fi /∂xj are globally bounded, uniformly in t, that is, their
absolute values are bounded for all t ∈ J and x ∈ Rn by constants independent of
(t, x).2
1 See [74, Lemma 3.2] for the proof of this statement.
2 See [74, Lemma 3.3] for the proof of this statement.
16 CHAPTER 1. INTRODUCTION

Example 1.1 The function

−x1 + x1 x2
f (x) =
x2 − x1 x2
is continuously differentiable on R2 . Hence, it is locally Lipschitz on R2 . It is not
globally Lipschitz since ∂f1 /∂x2 and ∂f2 /∂x1 are not uniformly bounded on R2 . On
any compact subset of R2 , f is Lipschitz. Suppose we are interested in calculating
a Lipschitz constant over the set W = {|x1 | ≤ a, |x2 | ≤ a}. Then,
|f1 (x) − f1 (y)| ≤ |x1 − y1 | + |x1 x2 − y1 y2 |
|f2 (x) − f2 (y)| ≤ |x2 − y2 | + |x1 x2 − y1 y2 |
Using the inequalities
|x1 x2 − y1 y2 | = |x1 (x2 − y2 ) + y2 (x1 − y1 )| ≤ a|x2 − y2 | + a|x1 − y1 |
|x1 − y1 | |x2 − y2 | ≤ 12 |x1 − y1 |2 + 12 |x2 − y2 |2
we obtain
f (x) − f (y)2 = |f1 (x) − f1 (y)|2 + |f2 (x) − f2 (y)|2 ≤ (1 + 2a)2 x − y2
Therefore, f is Lipschitz on W with the Lipschitz constant L = 1 + 2a.
Example 1.2 The function

x2
f (x) =
−sat(x1 + x2 )
is not continuously differentiable on R2 . Using the fact that the saturation function
sat(·) satisfies |sat(η) − sat(ξ)| ≤ |η − ξ|, we obtain
f (x) − f (y)2 ≤ (x2 − y2 )2 + (x1 + x2 − y1 − y2 )2
≤ (x1 − y1 )2 + 2(x1 − y1 )(x2 − y2 ) + 2(x2 − y2 )2
Using the inequality

T




a 1 1 a 1 1 a 2
2
a + 2ab + 2b = 2
≤ λmax
×
b 1 2 b 1 2 b
we conclude that
√
f (x) − f (y) ≤ 2.618 x − y, ∀ x, y ∈ R2
Here we have used a property of positive semidefinite symmetric matrices; that is,
xT P x ≤ λmax (P ) xT x, for all x ∈ Rn , where λmax (·) is the maximum eigenvalue of
P . A more conservative (larger) Lipschitz constant will be obtained if we use the
more conservative inequality
a2 + 2ab + 2b2 ≤ 2a2 + 3b2 ≤ 3(a2 + b2 )
√
resulting in a Lipschitz constant L = 3.
1.1. NONLINEAR MODELS 17

The local Lipschitz property of f (t, x) ensures local existence and uniqueness of
the solution of the state equation (1.3), as stated in the following lemma.3
Lemma 1.1 Let f (t, x) be piecewise continuous in t and locally Lipschitz in x at
x0 , for all t ∈ [t0 , t1 ]. Then, there is δ > 0 such that the state equation ẋ = f (t, x),
with x(t0 ) = x0 , has a unique solution over [t0 , t0 + δ]. 3
Without the local Lipschitz condition, we cannot ensure uniqueness of the solu-
tion. For example, the state equation ẋ = x1/3 , whose right-hand side function is
continuous but not locally Lipschitz at x = 0, has x(t) = (2t/3)3/2 and x(t) ≡ 0 as
two different solutions when the initial state is x(0) = 0.
Lemma 1.1 is a local result because it guarantees existence and uniqueness of
the solution over an interval [t0 , t0 + δ], but this interval might not include a given
interval [t0 , t1 ]. Indeed the solution may cease to exist after some time.
Example 1.3 In the one-dimensional system ẋ = −x2 , the function f (x) = −x2
is locally Lipschitz for all x. Yet, when we solve the equation with x(0) = −1, the
solution x(t) = 1/(t − 1) tends to −∞ as t → 1.
The phrase “finite escape time” is used to describe the phenomenon that a solution
escapes to infinity at finite time. In Example 1.3, we say that the solution has a
finite escape time at t = 1.
In the forthcoming Lemmas 1.2 and 1.3,4 we shall give conditions for global
existence and uniqueness of solutions. Lemma 1.2 requires the function f to be
globally Lipschitz, while Lemma 1.3 requires f to be only locally Lipschitz, but
with an additional requirement that the solution remains bounded. Note that the
function f (x) = −x2 of Example 1.3 is locally Lipschitz for all x but not globally
Lipschitz because f  (x) = −2x is not globally bounded.
Lemma 1.2 Let f (t, x) be piecewise continuous in t and globally Lipschitz in x for
all t ∈ [t0 , t1 ]. Then, the state equation ẋ = f (t, x), with x(t0 ) = x0 , has a unique
solution over [t0 , t1 ]. 3
The global Lipschitz condition is satisfied for linear systems of the form
ẋ = A(t)x + g(t)
when A(t) ≤ L for all t ≥ t0 , but it is a restrictive condition for general nonlinear
systems. The following lemma avoids this condition.
Lemma 1.3 Let f (t, x) be piecewise continuous in t and locally Lipschitz in x for
all t ≥ t0 and all x in a domain D ⊂ Rn . Let W be a compact (closed and bounded)
subset of D, x0 ∈ W , and suppose it is known that every solution of
ẋ = f (t, x), x(t0 ) = x0
lies entirely in W . Then, there is a unique solution that is defined for all t ≥ t0 . 3
3 See [74, Theorem 3.1] for the proof of Lemma 1.1. See [56, 62, 95] for a deeper look into exis-

tence and uniqueness of solutions, and the qualitative behavior of nonlinear differential equations.
4 See [74, Theorem 3.2] and [74, Theorem 3.3] for the proofs of Lemmas 1.2 and 1.3, respectively.
18 CHAPTER 1. INTRODUCTION

The trick in applying Lemma 1.3 is in checking the assumption that every so-
lution lies in a compact set without solving the state equation. We will see in
Chapter 3 that Lyapunov’s method for stability analysis provides a tool to ensure
this property. For now, let us illustrate the application of the lemma by an example.
Example 1.4 Consider the one-dimensional system

ẋ = −x3 = f (x)

The function f (x) is locally Lipschitz on R, but not globally Lipschitz because
f  (x) = −3x2 is not globally bounded. If, at any instant of time, x(t) is positive,
the derivative ẋ(t) will be negative and x(t) will be decreasing. Similarly, if x(t) is
negative, the derivative ẋ(t) will be positive and x(t) will be increasing. Therefore,
starting from any initial condition x(0) = a, the solution cannot leave the compact
set {|x| ≤ |a|}. Thus, we conclude by Lemma 1.3 that the equation has a unique
solution for all t ≥ 0.
A special case of (1.3) arises when the function f does not depend explicitly on
t; that is,
ẋ = f (x)
in which case the state equation is said to be autonomous or time invariant. The
behavior of an autonomous system is invariant to shifts in the time origin, since
changing the time variable from t to τ = t−a does not change the right-hand side of
the state equation. If the system is not autonomous, then it is called nonautonomous
or time varying.
More generally, the state model (1.1)–(1.2) is said to be time invariant if the
functions f and h do not depend explicitly on t; that is,

ẋ = f (x, u), y = h(x, u)

If either f or h depends on t, the state model is said to be time varying. A time-

invariant state model has a time-invariance property with respect to shifting the
initial time from t0 to t0 + a, provided the input waveform is applied from t0 + a
instead of t0 . In particular, let (x(t), y(t)) be the response for t ≥ t0 to initial state
x(t0 ) = x0 and input u(t) applied for t ≥ t0 , and let (x̃(t), ỹ(t)) be the response for
t ≥ t0 + a to initial state x̃(t0 + a) = x̃0 and input ũ(t) applied for t ≥ t0 + a. Now,
take x̃0 = x0 and ũ(t) = u(t − a) for t ≥ t0 + a. By changing the time variable from
t to t − a it can be seen that x̃(t) = x(t − a) and ỹ(t) = y(t − a) for t ≥ t0 + a.
Therefore, for time-invariant systems, we can, without loss of generality, take the
initial time to be t0 = 0.
A useful analysis tool is to transform the state equation from the x-coordinates
to the z-coordinates by the change of variables z = T (x). For linear systems, the
change of variables is a similarity transformation z = P x, where P is a nonsingular
matrix. For a nonlinear change of variables, z = T (x), the map T must be invertible;
that is, it must have an inverse map T −1 (·) such that x = T −1 (z) for all z ∈ T (D),
1.1. NONLINEAR MODELS 19

where D is the domain of T . Moreover, because the derivatives of z and x should

be continuous, we require both T (·) and T −1 (·) to be continuously differentiable. A
continuously differentiable map with a continuously differentiable inverse is known
as a diffeomorphism. A map T (x) is a local diffeomorphism at a point x0 if there
is a neighborhood N of x0 such that T restricted to N is a diffeomorphism on N .
It is a global diffeomorphism if it is a diffeomorphism on Rn and T (Rn ) = Rn .
The following lemma gives conditions on a map z = T (x) to be a local or global
diffeomorphism using the Jacobian matrix [∂T /∂x], which is a square matrix whose
(i, j) element is the partial derivative ∂Ti /∂xj .5
Lemma 1.4 The continuously differentiable map z = T (x) is a local diffeomor-
phism at x0 if the Jacobian matrix [∂T /∂x] is nonsingular at x0 . It is a global
diffeomorphism if and only if [∂T /∂x] is nonsingular for all x ∈ Rn and T is
proper; that is, limx→∞ T (x) = ∞. 3

Example 1.5 In Section A.4 two different models of the negative resistance oscil-
lator are given, which are related by the change of variables

−h(x1 ) − x2 /ε
z = T (x) =
x1

Assuming that h(x1 ) is continuously differentiable, the Jacobian matrix is

⎡ ∂T1 ∂T1 ⎤


⎦ = −h (x1 ) −1/ε
∂T ∂x1 ∂x2
=⎣
∂x ∂T2 ∂T2 1 0
∂x1 ∂x2

Its determinant, 1/ε, is positive for all x. Moreover, T (x) is proper because

T (x)2 = [h(x1 ) + x2 /ε]2 + x21

which shows that limx→∞ T (x) = ∞. In particular, if |x1 | → ∞, so is T (x).

If |x1 | is finite while |x2 | → ∞, so is [h(x1 ) + x2 /ε]2 and consequently T (x).

Equilibrium points are important features of the state equation. A point x∗ is

an equilibrium point of ẋ = f (t, x) if the equation has a constant solution x(t) ≡ x∗ .
For the time-invariant system ẋ = f (x), equilibrium points are the real solutions of

f (x) = 0

An equilibrium point could be isolated; that is, there are no other equilibrium points
in its vicinity, or there could be a continuum of equilibrium points. The linear
system ẋ = Ax has an isolated equilibrium point at x = 0 when A is nonsingular
or a continuum of equilibrium points in the null space of A when A is singular. It
5 The proof of the local result follows from the inverse function theorem [3, Theorem 7-5]. The

proof of the global results can be found in [117] or [150].

20 CHAPTER 1. INTRODUCTION

cannot have multiple isolated equilibrium points, for if xa and xb are two equilibrium
points, then by linearity any point on the line αxa + (1 − α)xb connecting xa and xb
will be an equilibrium point. A nonlinear state equation can have multiple isolated
equilibrium points. For example, the pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 − bx2

has equilibrium points at (x1 = nπ, x2 = 0) for n = 0, ±1, ±2, · · · .

1.2 Nonlinear Phenomena

The powerful analysis tools for linear systems are founded on the basis of the su-
perposition principle. As we move from linear to nonlinear systems, we are faced
with a more difficult situation. The superposition principle no longer holds, and
analysis involves more advanced mathematics. Because of the powerful tools we
know for linear systems, the first step in analyzing a nonlinear system is usually to
linearize it, if possible, about some nominal operating point and analyze the result-
ing linear model. This is a common practice in engineering, and it is a useful one.
However, there are two basic limitations of linearization. First, since linearization
is an approximation in the neighborhood of an operating point, it can only pre-
dict the “local” behavior of the nonlinear system in the vicinity of that point. It
cannot predict the “nonlocal” behavior far from the operating point and certainly
not the “global” behavior throughout the state space. Second, the dynamics of
a nonlinear system are much richer than the dynamics of a linear system. There
are “essentially nonlinear phenomena” that can take place only in the presence of
nonlinearity; hence, they cannot be described or predicted by linear models. The
following are examples of essentially nonlinear phenomena:

• Finite escape time. The state of an unstable linear system goes to infinity
as time approaches infinity; a nonlinear system’s state, however, can go to
infinity in finite time.

• Multiple isolated equilibria. A linear system can have only one isolated equi-
librium point; thus, it can have only one steady-state operating point that
attracts the state of the system irrespective of the initial state. A nonlinear
system can have more than one isolated equilibrium point. The state may
converge to one of several steady-state operating points, depending on the
initial state of the system.

• Limit cycles. For a linear time-invariant system to oscillate, it must have a

pair of eigenvalues on the imaginary axis, which is a nonrobust condition that
is almost impossible to maintain in the presence of perturbations. Even if we
do so, the amplitude of oscillation will be dependent on the initial state. In
real life, stable oscillation must be produced by nonlinear systems. There are
1.3. OVERVIEW OF THE BOOK 21

nonlinear systems that can go into oscillation of fixed amplitude and frequency,
irrespective of the initial state. This type of oscillation is known as limit cycles.

• Subharmonic, harmonic, or almost-periodic oscillations. A stable linear sys-

tem under a periodic input produces a periodic output of the same frequency.
A nonlinear system under periodic excitation can oscillate with frequencies
that are submultiples or multiples of the input frequency. It may even gener-
ate an almost-periodic oscillation, an example of which is the sum of periodic
oscillations with frequencies that are not multiples of each other.

• Chaos. A nonlinear system can have a more complicated steady-state behavior

that is not equilibrium, periodic oscillation, or almost-periodic oscillation.
Such behavior is usually referred to as chaos. Some of these chaotic motions
exhibit randomness, despite the deterministic nature of the system.

• Multiple modes of behavior. It is not unusual for two or more modes of be-
havior to be exhibited by the same nonlinear system. An unforced system
may have more than one limit cycle. A forced system with periodic excita-
tion may exhibit harmonic, subharmonic, or more complicated steady-state
behavior, depending upon the amplitude and frequency of the input. It may
even exhibit a discontinuous jump in the mode of behavior as the amplitude
or frequency of the excitation is smoothly changed.

In this book, we encounter only the first three of these phenomena.6 The phe-
nomenon of finite escape time has been already demonstrated in Example 1.3, while
multiple equilibria and limit cycles will be introduced in the next chapter.

1.3 Overview of the Book

Our study of nonlinear control starts with nonlinear analysis tools that will be
used in the analysis and design of nonlinear control systems. Chapter 2 introduces
phase portraits for the analysis of two-dimensional systems and illustrates some
essentially nonlinear phenomena. The next five chapters deal with stability analy-
sis of nonlinear systems. Stability of equilibrium points is defined and studied in
Chapter 3 for time-invariant systems. After presenting some preliminary results for
linear systems, linearization, and one-dimensional systems, the technique of Lya-
punov stability is introduced. It is the main tool for stability analysis of nonlinear
systems. It requires the search for a scalar function of the state, called Lyapunov
function, such that the function and its time derivative satisfy certain conditions.
The challenge in Lyapunov stability is the search for a Lyapunov function. How-
ever, by the time we reach the end of Chapter 7, the reader would have seen many
ideas and examples of how to find Lyapunov functions. Additional ideas are given
6 To read about forced oscillation, chaos, bifurcation, and other important topics, consult [52,

55, 136, 146].

22 CHAPTER 1. INTRODUCTION

in Appendix C. Chapter 4 extends Lyapunov stability to time-varying systems and

shows how it can be useful in the analysis of perturbed system. This leads into the
notion of input-to-state stability. Chapter 5 deals with a special class of systems
that dissipates energy. One point we emphasize is the connection between passivity
and Lyapunov stability. Chapter 6 looks at input-output stability and shows that
it can be established using Lyapunov functions. The tools of Chapters 5 and 6 are
used in Chapter 7 to derive stability criteria for the feedback connection of two
stable systems.
The next six chapters deal with nonlinear control. Chapter 8 presents some
special nonlinear forms that play significant roles in the design of nonlinear con-
trollers. Chapters 9 to 13 deal with nonlinear control problems, including nonlinear
observers. The nonlinear control techniques we are going to study can be catego-
rized into five different approaches to deal with nonlinearity. These are:
• Approximate nonlinearity
• Compensate for nonlinearity
• Dominate nonlinearity
• Use intrinsic properties
• Divide and conquer
Linearization is the prime example of approximating nonlinearities. Feedback lin-
earization that cancels nonlinearity is an example of nonlinearity compensation.
Robust control techniques, which are built around the classical tool of high-gain
feedback, dominate nonlinearities. Passivity-based control is an example of a tech-
nique that takes advantage of an intrinsic property of the system. Because the
complexity of a nonlinear system grows rapidly with dimension, one of the effective
ideas is to decompose the system into lower-order components, which might be eas-
ier to analyze and design, then build up back to the original system. Backstepping
is an example of this divide and conquer approach.
Four appendices at the end of the book give examples of nonlinear state models,
mathematical background, procedures for constructing composite Lyapunov func-
tions, and proofs of some results. The topics in this book overlap with topics in
some excellent textbooks, which can be consulted for further reading. The list in-
cludes [10, 53, 63, 66, 92, 118, 129, 132, 144]. The main source for the material in
this book is [74], which was prepared using many references. The reader is advised
to check the Notes and References section of [74] for a detailed account of these
references.

1.4 Exercises
1.1 A general mathematical model that describes the system with n state variables,
m input variables and p output variables is given by
ẋ1 = f1 (x1 , . . . xn , u1 , . . . um ), ẋn = fn (x1 , . . . xn , u1 , . . . um )
y1 = g1 (x1 , . . . xn , u1 , . . . um ) yp = gp (x1 , . . . xn , u1 , . . . um )
where u is the input and y is the output. Linearise the model at an equilibrium
T T
point x̄ = [x̄1 , x̄2 , . . . x̄n ] and ū = [ū1 , ū2 , . . . ūm ] . Find the state space model.