Nonlinear Control

This page intentionally left blank

 Nonlinear Control

Hassan K. Khalil
Department of Electrical and Computer Engineering
Michigan State University
Vice President and Editorial Cover Image: Reconstruction of a clepsydra
Director, ECS: Marcia J. Horton water clock invented by Ctesibius of
Executive Editor: Andrew Gilfillan Alexandria c270 BC 1857  c The Print
Marketing Manager: Tim Galligan Collector / Alamy
Managing Editor: Scott Disanno Full-Service Project Management/
Project Manager: Irwin Zucker Composition: SPI Global
Art Director: Jayne Conte Printer/Binder: Courier Westford
Cover Designer: Bruce Kenselaar Cover Printer: Courier Westford

Credits and acknowledgments borrowed from other sources and reproduced, with permission, in
this textbook appear on appropriate page within text.

Copyright c 2015 by Pearson Education, Inc., publishing as Prentice Hall, 1 Lake Street, Upper
Saddle River, NJ 07458.

All rights reserved. Manufactured in the United States of America. This publication is
protected by Copyright, and permission should be obtained from the publisher prior to any
prohibited reproduction, storage in a retrieval system, or transmission in any form or by any
means, electronic, mechanical, photocopying, recording, or likewise. To obtain permission(s) to
use material from this work, please submit a written request to Pearson Education, Inc.,
Permissions Department, imprint permissions address.

Many of the designations by manufacturers and seller to distinguish their products are claimed
as trademarks. Where those designations appear in this book, and the publisher was aware of a
trademark claim, the designations have been printed in initial caps or all caps.

MATLAB and Simulink are registered trademarks of The Math Works, Inc., 3 Apple Hill Drive,
Natick, MA 01760-2098

The author and publisher of this book have used their best efforts in preparing this book. These
efforts include the development, research, and testing of the theories and programs to determine
their effectiveness. The author and publisher make no warranty of any kind, expressed or
implied, with regard to these programs or the documentation contained in this book. The author
and publisher shall not be liable in any event for incidental or consequential damages in
connection with, or arising out of, the furnishing, performance, or use of these programs.

Library of Congress Cataloging-in-Publication Data on file

10 9 8 7 6 5 4 3 2 1

ISBN-10: 0-13-349926-X
ISBN-13: 978-0-13-349926-1
 To my parents
Mohamed and Fat-hia
and my grandchildren
Maryam, Tariq, Aya, and Tessneem
This page intentionally left blank
Contents

Preface xi

1 Introduction 1
1.1 Nonlinear Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Nonlinear Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Overview of the Book . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Two-Dimensional Systems 15
2.1 Qualitative Behavior of Linear Systems . . . . . . . . . . . . . . . . . . 17
2.2 Qualitative Behavior Near Equilibrium Points . . . . . . . . . . . . . . 21
2.3 Multiple Equilibria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.4 Limit Cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.5 Numerical Construction of Phase Portraits . . . . . . . . . . . . . . . . 31
2.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

3 Stability of Equilibrium Points 37

3.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3 Lyapunov’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4 The Invariance Principle . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.5 Exponential Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.6 Region of Attraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.7 Converse Lyapunov Theorems . . . . . . . . . . . . . . . . . . . . . . . 68
3.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70

4 Time-Varying and Perturbed Systems 75

4.1 Time-Varying Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.2 Perturbed Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.3 Boundedness and Ultimate Boundedness . . . . . . . . . . . . . . . . . 85
4.4 Input-to-State Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

vii
viii CONTENTS

5 Passivity 103
5.1 Memoryless Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 State Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.3 Positive Real Transfer Functions . . . . . . . . . . . . . . . . . . . . . 112
5.4 Connection with Stability . . . . . . . . . . . . . . . . . . . . . . . . . 115
5.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

6 Input-Output Stability 121

6.1 L Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
6.2 L Stability of State Models . . . . . . . . . . . . . . . . . . . . . . . . 127
6.3 L2 Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
6.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

7 Stability of Feedback Systems 141

7.1 Passivity Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
7.2 The Small-Gain Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 152
7.3 Absolute Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
7.3.1 Circle Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7.3.2 Popov Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

8 Special Nonlinear Forms 171

8.1 Normal Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
8.2 Controller Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
8.3 Observer Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
8.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

9 State Feedback Stabilization 197

9.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
9.2 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
9.3 Feedback Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . 201
9.4 Partial Feedback Linearization . . . . . . . . . . . . . . . . . . . . . . 207
9.5 Backstepping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
9.6 Passivity-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 217
9.7 Control Lyapunov Functions . . . . . . . . . . . . . . . . . . . . . . . . 222
9.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

10 Robust State Feedback Stabilization 231

10.1 Sliding Mode Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
10.2 Lyapunov Redesign . . . . . . . . . . . . . . . . . . . . . . . . . . . . 251
10.3 High-Gain Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
10.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
CONTENTS ix

11 Nonlinear Observers 263

11.1 Local Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
11.2 The Extended Kalman Filter . . . . . . . . . . . . . . . . . . . . . . . 266
11.3 Global Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
11.4 High-Gain Observers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
11.5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277

12 Output Feedback Stabilization 281

12.1 Linearization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
12.2 Passivity-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 283
12.3 Observer-Based Control . . . . . . . . . . . . . . . . . . . . . . . . . . 286
12.4 High-Gain Observers and the Separation Principle . . . . . . . . . . . . 288
12.5 Robust Stabilization of Minimum Phase Systems . . . . . . . . . . . . 296
12.5.1 Relative Degree One . . . . . . . . . . . . . . . . . . . . . . . 296
12.5.2 Relative Degree Higher Than One . . . . . . . . . . . . . . . 298
12.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303

13 Tracking and Regulation 307

13.1 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
13.2 Robust Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
13.3 Transition Between Set Points . . . . . . . . . . . . . . . . . . . . . . 314
13.4 Robust Regulation via Integral Action . . . . . . . . . . . . . . . . . . 318
13.5 Output Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 322
13.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

A Examples 329
A.1 Pendulum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
A.2 Mass–Spring System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
A.3 Tunnel-Diode Circuit . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
A.4 Negative-Resistance Oscillator . . . . . . . . . . . . . . . . . . . . . . 335
A.5 DC-to-DC Power Converter . . . . . . . . . . . . . . . . . . . . . . . . 337
A.6 Biochemical Reactor . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
A.7 DC Motor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
A.8 Magnetic Levitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
A.9 Electrostatic Microactuator . . . . . . . . . . . . . . . . . . . . . . . . 342
A.10 Robot Manipulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
A.11 Inverted Pendulum on a Cart . . . . . . . . . . . . . . . . . . . . . . . 345
A.12 Translational Oscillator with Rotating Actuator . . . . . . . . . . . . . 347

B Mathematical Review 349

x CONTENTS

C Composite Lyapunov Functions 355

C.1 Cascade Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355
C.2 Interconnected Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 357
C.3 Singularly Perturbed Systems . . . . . . . . . . . . . . . . . . . . . . . 359

D Proofs 363

Bibliography 369

Symbols 380

Index 382
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 ex-
ercises. The Companion Website for this book (http://www.pearsonhighered.com/
khalil) includes, among other things, an errata sheet, a link to report errors and
typos, pdf slides of the course, and Simulink models of selected examples.
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. The cover of the book depicts one of the first feedback control devices on
record, the ancient water clock of Ktesibios in Alexandria, Egypt, around the third
century B.C.
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

xi
This page intentionally left blank
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

1
2 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 3

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.
4 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 5

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.
6 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 7

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 ⎤


∂T ⎣
∂x1 ∂x2
⎦ −h (x1 ) −1/ε
= =
∂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].
8 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 9

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].
10 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 mathematical model that describes a wide variety of single-input–single-
ouput nonlinear systems is the nth-order differential equation
 
y (n) = g t, y, ẏ, . . . , y (n−1) , u
where u is the input and y the output. Find a state model.
1.4. EXERCISES 11

1.2 The nonlinear dynamic equations for a single-link manipulator with flexible
joints [135], damping ignored, is given by
I q̈1 + M gL sin q1 + k(q1 − q2 ) = 0, J q̈2 − k(q1 − q2 ) = u
where q1 and q2 are angular positions, I and J are moments of inertia, k is a spring
constant, M is the total mass, L is a distance, and u is a torque input.
(a) Using q1 , q̇1 , q2 , and q̇2 as state variables, find the state equation.
(b) Show that the right-hand side function is globally Lipschitz when u is constant.
(c) Find the equilibrium points when u = 0.
1.3 A synchronous generator connected to an infinite bus is represented by [103]
M δ̈ = P − Dδ̇ − η1 Eq sin δ, τ Ėq = −η2 Eq + η3 cos δ + EF
where δ is an angle in radians, Eq is voltage, P is mechanical input power, EF is
field voltage (input), D is damping coefficient, M is inertial coefficient, τ is time
constant, and η1 , η2 , and η3 are positive parameters.
(a) Using δ, δ̇, and Eq as state variables, find the state equation.
(b) Show that the right-hand side function is locally Lipschitz when P and EF are
constant. Is it globally Lipschitz?
(c) Show that when P and EF are constant and 0 < P < η1 EF /η2 , there is a
unique equilibrium point in the region 0 ≤ δ ≤ π/2.
1.4 The circuit shown in Figure 1.1 contains a nonlinear inductor and is driven
by a time-dependent current source. Suppose the nonlinear inductor is a Josephson
junction [25] described by iL = I0 sin kφL , where φL is the magnetic flux of the
inductor and I0 and k are constants.
(a) Using φL and vC as state variables, find the state equation.
(b) Show that the right-hand side function is locally Lipschitz when is is constant.
Is it globally Lipschitz?
(c) Find the equilibrium points when is = Is (constant) with 0 < Is < I0 .

1.5 Repeat the previous exercise when the nonlinear inductor is described by iL =
k1 φL + k2 φ3L , where k1 and k2 are positive constants. In Part (b), Is > 0.
1.6 Figure 1.2 shows a vehicle moving on a road with grade angle θ, where v the
vehicle’s velocity, M is its mass, and F is the tractive force generated by the engine.
Assume that the friction is due to Coulomb friction, linear viscous friction, and a
drag force proportional to v 2 . Viewing F as the control input and θ as a disturbance
input, find a state model of the system.
12 CHAPTER 1. INTRODUCTION

iL v

+ + F
is(t) R vC C vL
− −
Mg
θ

Figure 1.1: Exercises 1.4 and 1.5. Figure 1.2: Exercise 1.6.

θi + e u y
sin(.) G(s)
−

m
θo
y

Figure 1.3: Exercise 1.7. Figure 1.4: Exercise 1.8.

1.7 A phase-locked loop [45] can be represented by the block diagram of Figure 1.3.
Let {A, B, C} be a minimal realization of the scalar, strictly proper transfer function
G(s). Assume that all eigenvalues of A have negative real parts, G(0) = 0, and
θi = constant. Let z be the state of the realization {A, B, C}.
(a) Show that the closed-loop system can be represented by the state equations

ż = Az + B sin e, ė = −Cz

(b) Find the equilibrium points of the system.

1.8 Consider the mass–spring system shown in Figure 1.4. Assuming a linear
spring and nonlinear viscous damping described by c1 ẏ+c2 ẏ|ẏ|, find a state equation
that describes the motion of the system.

The next three exercises give examples of hydraulic systems [27].

1.9 Figure 1.5 shows a hydraulic system where liquid is stored in an open tank.
The cross-sectional area of the tank, A(h), is a function of h, the height of the liquid
h
level above the bottom of the tank. The liquid volume v is given by v = 0 A(λ) dλ.
For a liquid of density ρ, the absolute pressure p is given by p = ρgh + pa , where
pa is the atmospheric pressure (assumed constant) and g is the acceleration due to
1.4. EXERCISES 13

wi
pa

pa

wi p k wo
h
p k wo pa +Δp− pa
pa −Δp+
+Δp−

Figure 1.6: Exercise 1.10.

Figure 1.5: Exercise 1.9.

gravity. The tank receives liquid at a flow rate wi and loses liquid through a valve
√
that obeys the flow-pressure relationship wo = k p − pa . The rate of change of v
satisfies v̇ = wi − wo . Take wi to be the control input and h to be the output.
(a) Using h as the state variable, determine the state model.
(b) Using p − pa as the state variable, determine the state model.
(c) Find a constant input that maintains a constant output at h = r.

1.10 The hydraulic system shown in Figure 1.6 consists of a constant speed cen-
trifugal pump feeding a tank from which liquid flows through a pipe and a valve
√
that obeys the relationship w o = k p − pa . The pressure-flow characteristic of the
pump is given by p − pa = β 1 − wi /α for some positive constants α and β. The
cross-sectional area of the tank is uniform; therefore, v = Ah and p = pa + ρgv/A,
where the variables are defined in the previous exercise.
(a) Using (p − pa ) as the state variable, find the state model.
(b) Find the equilibrium points of the system.

1.11 The valves in the hydraulic system of Figure 1.7 obey the flow relationships
√ √
w1 = k1 p1 − p2 and w2 = k2 p2 − pa . The pump characteristic is p1 − pa =
β 1 − wp /α. All variables are defined in the previous two exercises.
(a) Using (p1 − pa ) and (p2 − pa ) as the state variables, find the state equation.
(b) Find the equilibrium points of the system.

1.12 For each of the following systems, investigate local and global Lipschitz prop-
erties. Assume that input variables are continuous functions of time.
(a) The pendulum equation (A.2).
(b) The mass-spring system (A.6).
14 CHAPTER 1. INTRODUCTION

pa pa

wp p1 k1 w1 p2 k2 w2
pa
pa

Figure 1.7: The hydraulic system of Exercise 1.11.

(c) The tunnel diode circuit (A.7).

(d) The van der Pol oscillator (A.13).

(e) The boost converter (A.16).

(f ) The biochemical reactor (A.19) with ν defined by (A.20).

(g) The DC motor (A.25) when fe and f are linear functions.

(h) The magnetic levitation system (A.30)–(A.32).

(i) The electrostatic actuator (A.33).

(j) The two-link robot manipulator (A.35)–(A.37).

(k) The inverted pendulum on a cart (A.41)–(A.44).

(l) The TORA system (A.49)–(A.52).

1.13 Find a global diffeomorphism z = T (x) that transforms the system

ẋ1 = x2 + g1 (x1 ), ẋ2 = x3 + g2 (x1 , x2 ), ẋ3 = g3 (x) + g4 (x)u, y = x1

into
ż1 = z2 , ż2 = z3 , ż3 = a(z) + b(z)u, y = z1
Assume g1 to g4 are smooth.

1.14 Find a diffeomorphism z = T (x) that transforms the system

ẋ1 = sin x2 , ẋ2 = −x21 + u, y = x1

into
ż1 = z2 , ż2 = a(z) + b(z)u, y = z1
and give the definitions of a and b.
Chapter 2

Two-Dimensional Systems

Two-dimensional time-invariant systems occupy an important place in the study of

nonlinear systems because solutions can be represented by curves in the plane. This
allows for easy visualization of the qualitative behavior of the system. The purpose
of this chapter is to use two-dimensional systems to introduce, in an elementary
context, some of the basic ideas of nonlinear systems. In particular, we will look at
the behavior of a nonlinear system near equilibrium points and the phenomenon of
nonlinear oscillation.1
A two-dimensional time-invariant system is represented by two simultaneous
differential equations

ẋ1 = f1 (x1 , x2 ), ẋ2 = f2 (x1 , x2 ) (2.1)

Assume that f1 and f2 are locally Lipschitz over the domain of interest and let
def
x(t) = (x1 (t), x2 (t)) be the solution of (2.1) that starts at x(0) = (x10 , x20 ) = x0 .
The locus in the x1 –x2 plane of the solution x(t) for all t ≥ 0 is a curve that passes
through the point x0 . This curve is called a trajectory or orbit of (2.1) from x0 .
The x1 –x2 plane is called the state plane or phase plane. Using vector notation we
rewrite (2.1) as
ẋ = f (x)
where f (x) is the vector (f1 (x), f2 (x)); it is tangent to the trajectory at x. We
consider f (x) as a vector field on the state plane, which means that to each point
x in the plane, we assign a vector f (x). For easy visualization, we represent f (x)
as a vector based at x; that is, we assign to x the directed line segment from x to
x + f (x). For example, if f (x) = (2x21 , x2 ), then at x = (1, 1), we draw an arrow
pointing from (1, 1) to (1, 1) + (2, 1) = (3, 2). (See Figure 2.1.)
Drawing the vector field at every point in a grid covering the plane, we obtain a
vector field diagram, such as the one shown in Figure 2.2 for the pendulum equation
1 The chapter follows closely the presentation of [25].

15
16 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

x2
1

x + f(x) = (3, 2)
0

x2
f (x)

−1
x = (1, 1)

−2
x1 −2 −1 0 1 2
x1

Figure 2.1: Vector field repre-

sentation. Figure 2.2: Vector field diagram
of the pendulum equation without
friction.

without friction:
ẋ1 = x2 , ẋ2 = − sin x1
In the figure, the length  of the arrow at a given point x is proportional to the
length of f (x), that is, f12 (x) + f22 (x). Sometimes, for convenience, we draw
arrows of equal length at all points. Since the vector field at a point is tangent to
the trajectory through that point, we can, in essence, construct trajectories from
the vector field diagram. Starting at a given initial point x0 , we can construct the
trajectory from x0 by moving along the vector field at x0 . This motion takes us to
a new point xa , where we continue the trajectory along the vector field at xa . If the
process is repeated carefully and the consecutive points are chosen close enough to
each other, we can obtain a reasonable approximation of the trajectory through x0 .
In the case of Figure 2.2, a careful implementation of the foregoing process would
show that the trajectory through (1.5, 0) is a closed curve.
The family of all trajectories is called the phase portrait of (2.1). An (approxi-
mate) picture of the phase portrait can be constructed by plotting trajectories from
a large number of initial states spread all over the x1 –x2 plane. Since numerical
programs for solving nonlinear differential equations are widely available, we can
easily construct the phase portrait by using computer simulation. (Some hints are
given in Section 2.5.) Note that since the time t is suppressed in a trajectory, it is
not possible to recover the solution (x1 (t), x2 (t)) associated with a given trajectory.
Hence, a trajectory gives only the qualitative, but not quantitative, behavior of the
associated solution. For example, a closed trajectory shows that there is a periodic
solution; that is, the system has a sustained oscillation, whereas a shrinking spi-
ral shows a decaying oscillation. In the rest of this chapter, we will qualitatively
analyze the behavior of two-dimensional systems by using their phase portraits.
2.1. QUALITATIVE BEHAVIOR OF LINEAR SYSTEMS 17

2.1 Qualitative Behavior of Linear Systems

Consider the linear time-invariant system

ẋ = Ax (2.2)

where A is a 2 × 2 real matrix. The solution of (2.2) for a given initial state x0 is
given by
x(t) = M exp(Jr t)M −1 x0
where Jr is the real Jordan form of A and M is a real nonsingular matrix such
that M −1 AM = Jr . We restrict our attention to the case when A has distinct
eigenvalues, different from zero.2 The real Jordan form takes one of two forms

λ1 0 α −β
or
0 λ2 β α

The first form occurs when the eigenvalues are real and the second when they are
complex.

Case 1. Both eigenvalues are real

In this case, M = [v1 , v2 ], where v1 and v2 are the real eigenvectors associated
with λ1 and λ2 . The change of coordinates z = M −1 x transforms the system into
two decoupled scalar (one-dimensional) differential equations,

ż1 = λ1 z1 , ż2 = λ2 z2

whose solutions, for a given initial state (z10 , z20 ), are given by

z1 (t) = z10 eλ1 t , z2 (t) = z20 eλ2 t

Eliminating t between the two equations, we obtain

λ /λ1
z2 = cz1 2 (2.3)

where c = z20 /(z10 )(λ2 /λ1 ) . The phase portrait of the system is given by the family
of curves generated from (2.3) by allowing the real number c to take arbitrary values.
The shape of the phase portrait depends on the signs of λ1 and λ2 .
Consider first the case when both eigenvalues are negative. Without loss of
generality, let λ2 < λ1 < 0. Here, both exponential terms eλ1 t and eλ2 t tend to
zero as t → ∞. Moreover, since λ2 < λ1 < 0, the term eλ2 t tends to zero faster
than eλ1 t . Hence, we call λ2 the fast eigenvalue and λ1 the slow eigenvalue. For
later reference, we call v2 the fast eigenvector and v1 the slow eigenvector. The
2 See [74, Section 2.1] for the case when A has zero or multiple eigenvalues.
18 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

z2

z1

Figure 2.3: Phase portrait of stable node in modal coordinates.

x2 x2
v2 v2

v1 v1

x1 x1

(a) (b)

Figure 2.4: Phase portraits of (a) stable node; (b) unstable node.

trajectory tends to the origin of the z1 –z2 plane along the curve of (2.3), which now
has (λ2 /λ1 ) > 1. The slope of the curve is given by

dz2 λ2 [(λ /λ )−1]

= c z1 2 1
dz1 λ1
Since [(λ2 /λ1 ) − 1] is positive, the slope approaches zero as |z1 | → 0 and approaches
∞ as |z1 | → ∞. Therefore, as the trajectory approaches the origin, it becomes
tangent to the z1 -axis; as it approaches ∞, it becomes parallel to the z2 -axis. These
observations allow us to sketch the typical family of trajectories shown in Figure 2.3.
When transformed back into the x-coordinates, the family of trajectories will have
the typical portrait shown in Figure 2.4(a). Note that in the x1 –x2 plane, the
trajectories become tangent to the slow eigenvector v1 as they approach the origin
and parallel to the fast eigenvector v2 far from the origin. In this situation, the
equilibrium point x = 0 is called stable node (or nodal sink).
2.1. QUALITATIVE BEHAVIOR OF LINEAR SYSTEMS 19

z2 x2

v2
v1

z1 x1

(a) (b)

Figure 2.5: Phase portrait of a saddle point (a) in modal coordinates; (b) in original
coordinates.

When λ1 and λ2 are positive, the phase portrait will retain the character of Fig-
ure 2.4(a), but with the trajectory directions reversed, since the exponential terms
eλ1 t and eλ2 t grow exponentially as t increases. Figure 2.4(b) shows the phase por-
trait for the case λ2 > λ1 > 0. The equilibrium point x = 0 is referred to in this
instance as unstable node (or nodal source).

Suppose now that the eigenvalues have opposite signs. In particular, let λ2 <
0 < λ1 . In this case, eλ1 t → ∞, while eλ2 t → 0 as t → ∞. Hence, we call λ2 the
stable eigenvalue and λ1 the unstable eigenvalue. Correspondingly, v2 and v1 are
called the stable and unstable eigenvectors, respectively. Equation (2.3) will have a
negative exponent (λ2 /λ1 ). Thus, the family of trajectories in the z1 –z2 plane will
take the typical form shown in Figure 2.5(a). Trajectories have hyperbolic shapes.
They become tangent to the z1 -axis as |z1 | → ∞ and tangent to the z2 -axis as
|z1 | → 0. The only exception to these hyperbolic shapes are the four trajectories
along the axes. The two trajectories along the z2 -axis are called the stable trajec-
tories since they approach the origin as t → ∞, while the two trajectories along the
z1 -axis are called the unstable trajectories since they approach infinity as t → ∞.
The phase portrait in the x1 –x2 plane is shown in Figure 2.5(b). Here the sta-
ble trajectories are along the stable eigenvector v2 and the unstable trajectories are
along the unstable eigenvector v1 . In this case, the equilibrium point is called saddle.

Case 2. Complex eigenvalues: λ1,2 = α ± jβ.

The change of coordinates z = M −1 x transforms the system (2.2) into the form

ż1 = αz1 − βz2 , ż2 = βz1 + αz2

The solution of these equations is oscillatory and can be expressed more conveniently
in the polar coordinates z1 = r cos θ and z2 = r sin θ, where we have two uncoupled
20 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

z2 z2 z2

z1 z1 z1

(a) (b) (c)

Figure 2.6: Typical trajectories in the case of complex eigenvalues.

(a) α < 0; (b) α > 0; (c) α = 0.

x2 x2 x2

x1 x1 x1

(a) (b) (c)

Figure 2.7: Phase portraits of (a) stable focus; (b) unstable focus; (c) center.

scalar differential equations:

ṙ = αr and θ̇ = β
The solution for a given initial state (r0 , θ0 ) is given by
r(t) = r0 eαt and θ(t) = θ0 + βt
which define a logarithmic spiral in the z1 –z2 plane. Depending on the value of α,
the trajectory will take one of the three forms shown in Figure 2.6. When α < 0,
the spiral converges to the origin; when α > 0, it diverges away from the origin.
When α = 0, the trajectory is a circle of radius r0 . Figure 2.7 shows the trajectories
in the x1 –x2 plane. The equilibrium point x = 0 is referred to as stable focus (or
spiral sink) if α < 0, unstable focus (or spiral source) if α > 0, and center if α = 0.
The local behavior of a nonlinear system near an equilibrium point may be
deduced by linearizing the system about that point and studying the behavior of
the resultant linear system. How conclusive this linearization is, depends to a great
extent on how the various qualitative phase portraits of a linear system persist
under perturbations. We can gain insight into the behavior of a linear system
under perturbations by examining the special case of linear perturbations. Suppose
A is perturbed to A + ΔA, where ΔA is a 2 × 2 real matrix whose elements have
arbitrarily small magnitudes. From the perturbation theory of matrices,3 we know
3 See [51, Chapter 7].
2.2. QUALITATIVE BEHAVIOR NEAR EQUILIBRIUM POINTS 21

that the eigenvalues of a matrix depend continuously on its parameters. This means
that, given any positive number ε, there is a corresponding positive number δ such
that if the magnitude of the perturbation in each element of A is less than δ, the
eigenvalues of the perturbed matrix A+ΔA will lie in open discs of radius ε centered
at the eigenvalues of A. Consequently, any eigenvalue of A that lies in the open
right-half plane (positive real part) or in the open left-half plane (negative real part)
will remain in its respective half of the plane after arbitrarily small perturbations.
On the other hand, eigenvalues on the imaginary axis, when perturbed, might go
into either the right-half or the left-half of the plane, since a disc centered on the
imaginary axis will extend in both halves no matter how small ε is. Consequently,
we can conclude that if the equilibrium point x = 0 of ẋ = Ax is a node, focus,
or saddle, then the equilibrium point x = 0 of ẋ = (A + ΔA)x will be of the same
type, for sufficiently small perturbations. The situation is quite different if the
equilibrium point is a center. Consider the perturbation of the real Jordan form in
the case of a center:

μ 1
−1 μ

where μ is a perturbation parameter. The equilibrium point of the perturbed system

is unstable focus when μ is positive, and stable focus when it is negative. This is
true no matter how small μ is, as long as it is different from zero. Because the phase
portraits of the stable and unstable foci are qualitatively different from the phase
portrait of the center, we see that a center equilibrium point may not persist under
perturbations. The node, focus, and saddle equilibrium points are said to be struc-
turally stable because they maintain their qualitative behavior under infinitesimally
small perturbations,4 while the center equilibrium point is not structurally stable.
The distinction between the two cases is due to the location of the eigenvalues of
A, with the eigenvalues on the imaginary axis being vulnerable to perturbations.
This brings in the definition of hyperbolic equilibrium points. The equilibrium point
is hyperbolic if A has no eigenvalue with zero real part.5

2.2 Qualitative Behavior Near Equilibrium Points

In this section we will see that, except for some special cases, the qualitative behavior
of a nonlinear system near an equilibrium point can be determined via linearization
with respect to that point.
Let p = (p1 , p2 ) be an equilibrium point of the nonlinear system (2.1) and
suppose that the functions f1 and f2 are continuously differentiable. Expanding f1

4 See [62, Chapter 16] for a rigorous and more general definition of structural stability.
5 This definition of hyperbolic equilibrium points extends to higher-dimensional systems. It also
carries over to equilibria of nonlinear systems by applying it to the eigenvalues of the linearized
system.
22 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

and f2 into their Taylor series about (p1 , p2 ), we obtain

ẋ1 = f1 (p1 , p2 ) + a11 (x1 − p1 ) + a12 (x2 − p2 ) + H.O.T.

ẋ2 = f2 (p1 , p2 ) + a21 (x1 − p1 ) + a22 (x2 − p2 ) + H.O.T.

where
 
∂f1 (x1 , x2 )  ∂f1 (x1 , x2 ) 
a11 =  , a12 = 
∂x1 x1 =p1 ,x2 =p2 ∂x2 x =p ,x =p
  1 1 2 2
∂f2 (x1 , x2 )  ∂f2 (x1 , x2 ) 
a21 =  , a22 = 
∂x1 x1 =p1 ,x2 =p2 ∂x2 x1 =p1 ,x2 =p2

and H.O.T. denotes higher-order terms of the expansion, that is, terms of the form
(x1 − p1 )2 , (x2 − p2 )2 , (x1 − p1 )(x2 − p2 ), and so on. Since (p1 , p2 ) is an equilibrium
point, we have
f1 (p1 , p2 ) = f2 (p1 , p2 ) = 0
Moreover, since we are interested in the trajectories near (p1 , p2 ), we define y1 =
x1 − p1 , y2 = x2 − p2 , and rewrite the state equations as

ẏ1 = ẋ1 = a11 y1 + a12 y2 + H.O.T.

ẏ2 = ẋ2 = a21 y1 + a22 y2 + H.O.T.

If we restrict attention to a sufficiently small neighborhood of the equilibrium point

such that the higher-order terms are negligible, then we may drop these terms and
approximate the nonlinear state equations by the linear state equations

ẏ1 = a11 y1 + a12 y2 , ẏ2 = a21 y1 + a22 y2

Rewriting the equations in a vector form, we obtain

⎡ ⎤ ⎡ ∂f1 ⎤
a11 a12
∂f1  
∂x1 ∂x2  ∂f 
ẏ = Ay, where A=⎣ ⎦= ⎣ ⎦ =
 ∂x x=p
a21 a22 ∂f2 ∂f2 
∂x 1 ∂x2 x=p

The matrix [∂f /∂x] is the Jacobian of f (x), and A is its evaluation x = p.
It is reasonable to expect the trajectories of the nonlinear system in a small
neighborhood of an equilibrium point to be “close” to the trajectories of its lin-
earization about that point. Indeed, it is true that6 if the origin of the linearized
state equation is a stable (respectively, unstable) node with distinct eigenvalues, a
stable (respectively, unstable) focus, or a saddle, then, in a small neighborhood of
6 The proof of this linearization property can be found in [58]. It is valid under the assumption

that f1 (x1 , x2 ) and f2 (x1 , x2 ) have continuous first partial derivatives in a neighborhood of the
equilibrium point (p1 , p2 ). A related, but different, linearization result will be stated in Chapter 3
for higher-dimensional systems. (See Theorem 3.2.)
2.2. QUALITATIVE BEHAVIOR NEAR EQUILIBRIUM POINTS 23

the equilibrium point, the trajectories of the nonlinear state equation will behave
like a stable (respectively, unstable) node, a stable (respectively, unstable) focus, or
a saddle. Consequently, we call an equilibrium point of the nonlinear state equation
(2.1) stable (respectively, unstable) node, stable (respectively, unstable) focus, or
saddle if the linearized state equation about the equilibrium point has the same
behavior.
The foregoing linearization property dealt only with cases where the linearized
state equation has no eigenvalues on the imaginary axis, that is, when the origin is
a hyperbolic equilibrium point of the linear system. We extend this definition to
nonlinear systems and say that an equilibrium point is hyperbolic if the Jacobian
matrix, evaluated at that point, has no eigenvalues on the imaginary axis. If the
Jacobian matrix has eigenvalues on the imaginary axis, then the qualitative behavior
of the nonlinear state equation near the equilibrium point could be quite distinct
from that of the linearized state equation. This should come as no surprise in
view of our earlier discussion on the effect of linear perturbations on the qualitative
behavior of linear systems. The example that follows considers a case when the
origin of the linearized state equation is a center.

Example 2.1 The system

ẋ1 = −x2 − μx1 (x21 + x22 ), ẋ2 = x1 − μx2 (x21 + x22 )

has an equilibrium point at the origin. The linearized state equation at the origin
has eigenvalues ±j. Thus, the origin is a center equilibrium point for the linearized
system. We can determine the qualitative behavior of the nonlinear system in the
polar coordinates x1 = r cos θ, x2 = r sin θ. The equations

ṙ = −μr3 and θ̇ = 1

show that the trajectories of the nonlinear system will resemble a stable focus when
μ > 0 and unstable focus when μ < 0. 

The preceding example shows that the qualitative behavior describing a center
in the linearized state equation is not preserved in the nonlinear state equation. De-
termining that a nonlinear system has a center must be done by nonlinear analysis.
For example, by constructing the phase portrait of the pendulum equation without
friction, as in Figure 2.2, it can be seen that the equilibrium point at the origin
(0, 0) is a center.
Determining the type of equilibrium points via linearization provides useful in-
formation that should be used when we construct the phase portrait of a two-
dimensional system. In fact, the first step in constructing the phase portrait should
be the calculation of all equilibrium points and determining the type of the isolated
ones via linearization, which will give us a clear idea about the expected portrait in
the neighborhood of these equilibrium points.
24 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

2.3 Multiple Equilibria

The linear system ẋ = Ax has an isolated equilibrium point at x = 0 if A has
no zero eigenvalues, that is, if det A = 0. When det A = 0, the system has a
continuum of equilibrium points. These are the only possible equilibria patterns that
a linear system may have. A nonlinear system can have multiple isolated equilibrium
points. In the following two examples, we explore the qualitative behavior of the
tunnel-diode circuit of Section A.3 and the pendulum equation of Section A.1. Both
systems exhibit multiple isolated equilibria.

Example 2.2 The state model of a tunnel-diode circuit is given by

1 1
ẋ1 = [−h(x1 ) + x2 ], ẋ2 = [−x1 − Rx2 + u]
C L

Assume that the circuit parameters are7 u = 1.2 V , R = 1.5 kΩ = 1.5 × 103 Ω,
C = 2 pF = 2 × 10−12 F , and L = 5 μH = 5 × 10−6 H. Measuring time in
nanoseconds and the currents x2 and h(x1 ) in mA, the state model is given by

def
ẋ1 = 0.5[−h(x1 ) + x2 ] = f1 (x)
def
ẋ2 = 0.2(−x1 − 1.5x2 + 1.2) = f2 (x)

Suppose h(·) is given by

h(x1 ) = 17.76x1 − 103.79x21 + 229.62x31 − 226.31x41 + 83.72x51

The equilibrium points are determined by the intersection of the curve x2 = h(x1 )
with the load line 1.5x2 = 1.2 − x1 . For the given numerical values, these two
curves intersect at three points: Q1 = (0.063, 0.758), Q2 = (0.285, 0.61), and Q3 =
(0.884, 0.21), as shown by the solid curves in Figure A.5.
The Jacobian matrix of f (x) is given by
⎡ ⎤
−0.5h (x1 ) 0.5
∂f ⎣ ⎦
=
∂x
−0.2 −0.3

where
dh
h (x1 ) = = 17.76 − 207.58x1 + 688.86x21 − 905.24x31 + 418.6x41
dx1

Evaluating the Jacobian matrix at the equilibrium points Q1 , Q2 , and Q3 , respec-

7 The numerical data are taken from [25].
2.3. MULTIPLE EQUILIBRIA 25

x2
1.6

1.4

1.2

0.8
Q1 Q2
0.6

0.4

0.2 Q3
0
x1
−0.2

−0.4
−0.4 −0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Figure 2.8: Phase portrait of the tunnel-diode circuit of Example 2.2.

tively, yields

−3.598 0.5
A1 = ; Eigenvalues : − 3.57, −0.33
−0.2 −0.3

1.82 0.5
A2 = ; Eigenvalues : 1.77, −0.25
−0.2 −0.3

−1.427 0.5
A3 = ; Eigenvalues : − 1.33, −0.4
−0.2 −0.3
Thus, Q1 is a stable node, Q2 a saddle, and Q3 a stable node. Examination of
the phase portrait in Figure 2.8 shows that, except for the stable trajectories of
the saddle Q2 , all other trajectories eventually approach either Q1 or Q3 . The two
stable trajectories of the saddle form a curve that divides the plane into two halves.
All trajectories originating from the left side of the curve approach Q1 , while those
originating from the right side approach Q3 . This special curve is called a separatrix
because it partitions the plane into two regions of different qualitative behavior.8
In an experimental setup, we shall observe one of the two steady-state operating
points Q1 or Q3 , depending on the initial capacitor voltage and inductor current.
The equilibrium point at Q2 is never observed in practice because the ever-present
physical noise would cause the trajectory to diverge from Q2 even if it were possible
to set up the exact initial conditions corresponding to Q2 .
8 In general, the state plane decomposes into a number of regions, within each of which the

trajectories may show a different type of behavior. The curves separating these regions are called
separatrices.
26 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

The phase portrait in Figure 2.8 shows the global qualitative behavior of the
tunnel-diode circuit. The bounding box is chosen so that all essential qualitative
features are displayed. The portrait outside the box does not contain new qualitative
features. 

Example 2.3 The pendulum equation with friction

def
ẋ1 = x2 = f1 (x)
def
ẋ2 = − sin x1 − 0.3x2 = f2 (x)

has equilibrium points at (nπ, 0) for n = 0, ±1, ±2, . . .. Evaluating the Jacobian
matrix

∂f 0 1
=
∂x − cos x1 −0.3

at the equilibrium points (0, 0) and (π, 0) yields, respectively, the two matrices

0 1
A1 = ; Eigenvalues : − 0.15 ± j0.9887
−1 −0.3

0 1
A2 = ; Eigenvalues : − 1.1612, 0.8612
1 −0.3

Thus, (0, 0) is a stable focus and (π, 0) is a saddle. The phase portrait, shown in
Figure 2.9, is periodic in x1 with period 2π. Consequently, all distinct features
of the system’s qualitative behavior can be captured by drawing the portrait in
the vertical strip −π ≤ x1 ≤ π. The equilibrium points (0, 0), (2π, 0), (−2π, 0),
etc., correspond to the downward equilibrium position (0, 0); they are stable foci.
On the other hand, the equilibrium points at (π, 0), (−π, 0), etc., correspond to
the upward equilibrium position (π, 0); they are saddles. The stable trajectories of
the saddles at (π, 0) and (−π, 0) form separatrices which contain a region with the
property that all trajectories in its interior approach the equilibrium point (0, 0).
This picture is repeated periodically. The fact that trajectories could approach
different equilibrium points correspond to the number of full swings a trajectory
would take before it settles at the downward equilibrium position. For example, the
trajectories starting at points A and B have the same initial position but different
speeds. The trajectory starting at A oscillates with decaying amplitude until it
settles down at equilibrium. The trajectory starting at B, on the other hand, has
more initial kinetic energy. It makes a full swing before it starts to oscillate with
decaying amplitude. Once again, notice that the “unstable” equilibrium position
(π, 0) cannot be maintained in practice, because noise would cause trajectories to
diverge away from it. 
2.4. LIMIT CYCLES 27

4
x2

3 B

1
A

0 x1

−1

−2

−3

−4
−8 −6 −4 −2 0 2 4 6 8

Figure 2.9: Phase portrait of the pendulum equation of Example 2.3.

2.4 Limit Cycles

Oscillation is one of the most important phenomena that occur in dynamical sys-
tems. A system oscillates when it has a nontrivial periodic solution
x(t + T ) = x(t), ∀t≥0
for some T > 0. The word “nontrivial” is used to exclude constant solutions corre-
sponding to equilibrium points. A constant solution satisfies the preceding equation,
but it is not what we have in mind when we talk of oscillation or periodic solutions.
From this point on whenever we refer to a periodic solution, we will mean a nontriv-
ial one. The image of a periodic solution in the phase portrait is a closed trajectory,
which is usually called a periodic orbit or a closed orbit.
We have already seen an example of oscillation in Section 2.1: the two-dimensional
linear system with eigenvalues ±jβ. The origin of that system is a center and the
trajectories are closed orbits. When the system is transformed into its real Jordan
form, the solution is given by
z1 (t) = r0 cos(βt + θ0 ), z2 (t) = r0 sin(βt + θ0 )
where   
z2 (0)
r0 = z12 (0) + z22 (0), θ0 = tan−1
z1 (0)
Therefore, the system has a sustained oscillation of amplitude r0 . It is usually
referred to as the harmonic oscillator. If we think of the harmonic oscillator as a
28 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

C L

Figure 2.10: A linear LC circuit for the harmonic oscillator.

model for the linear LC circuit of Figure 2.10, then we can see that the physical
mechanism leading to these oscillations is a periodic exchange (without dissipation)
of the energy stored in the capacitor’s electric field with the energy stored in the
inductor’s magnetic field. There are, however, two fundamental problems with
this linear oscillator. The first problem is one of robustness. We have seen that
infinitesimally small right-hand side (linear or nonlinear) perturbations will destroy
the oscillation. That is, the linear oscillator is not structurally stable. In fact, it
is impossible to build an LC circuit that realizes the harmonic oscillator, for the
resistance in the electric wires alone will eventually consume whatever energy was
initially stored in the capacitor and inductor. Even if we succeeded in building the
linear oscillator, we would face the second problem: the amplitude of oscillation is
dependent on the initial conditions.
The two fundamental problems of the linear oscillator can be eliminated in
nonlinear oscillators. It is possible to build physical nonlinear oscillators such that

• the oscillator is structurally stable, and

• the amplitude of oscillation (at steady state) is independent of initial condi-

tions.

The negative-resistance oscillator of Section A.4 is an example of such nonlinear

oscillators. It is modeled by

ẋ1 = x2 , ẋ2 = −x1 − εh (x1 )x2

where h satisfies the conditions of Section A.4. The system has only one equilibrium
point at x1 = x2 = 0, at which the Jacobian matrix is
⎡ ⎤
 0 1
∂f 
A= =⎣ ⎦
∂x x=0 
−1 −εh (0)

Since h (0) < 0, the origin is either unstable node or unstable focus, depending on
the value of εh (0). In either case, all trajectories starting near the origin diverge
away from it. This repelling feature is due to the negative resistance near the origin,
which means that the resistive element is “active” and supplies energy. This point
2.4. LIMIT CYCLES 29

−a x1
b

Figure 2.11: A sketch of h(x1 ) (solid) and −x1 h(x1 ) (dashed), which shows that Ė is
positive for −a ≤ x1 ≤ b.

can be seen also by examining the rate of change of energy. The total energy stored
in the capacitor and inductor at any time t is given by
2
E = 12 CvC + 12 Li2L

shown in Section A.4 that vC = x1 and iL = −h(x1 ) − x2 /ε. Recalling that

It is
ε = L/C, we can rewrite the energy as

E = 12 C{x21 + [εh(x1 ) + x2 ]2 }

The rate of change of energy is given by

Ė = C{x1 ẋ1 + [εh(x1 ) + x2 ][εh (x1 )ẋ1 + ẋ2 ]}

= C{x1 x2 + [εh(x1 ) + x2 ][εh (x1 )x2 − x1 − εh (x1 )x2 ]}
= −εCx1 h(x1 )

The preceding expression confirms that, near the origin, the trajectory gains energy
since for small |x1 | the term x1 h(x1 ) is negative. It also shows that there is a strip
−a ≤ x1 ≤ b such that the trajectory gains energy within the strip and loses energy
outside it. The strip boundaries −a and b are solutions of h(x1 ) = 0, as shown in
Figure 2.11. As a trajectory moves in and out of the strip, there is an exchange of
energy with the trajectory gaining energy inside the strip and losing it outside. A
stationary oscillation will occur if, along a trajectory, the net exchange of energy
over one cycle is zero. Such trajectory will be a closed orbit. It turns out that the
negative-resistance oscillator has an isolated closed orbit, which is illustrated in the
next example for the Van der Pol oscillator.

Example 2.4 Figures 2.12(a), 2.12(b), and 2.13(a) show the phase portraits of the
Van der Pol equation

ẋ1 = x2 , ẋ2 = −x1 + ε(1 − x21 )x2

30 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

4 x2 x2
3 3

2 2

1 1
0
0 x1 x1
−1 −1

−2 −2
−3
−2 0 2 4 −2 0 2 4
(a) (b)

Figure 2.12: Phase portraits of the Van der Pol oscillator: (a) ε = 0.2; (b) ε = 1.0.

3 z2
x2
10
2

5 1

0
z1
0 x1
−1

−5 −2

−3
−5 0 5 10 −2 0 2
(a) (b)

Figure 2.13: Phase portrait of the Van der Pol oscillator with ε = 5.0: (a) in x1 –x2
plane; (b) in z1 –z2 plane.

for three different values of the parameter ε: a small value of 0.2, a medium value of
1.0, and a large value of 5.0. In all three cases, the phase portraits show a unique
closed orbit that attracts all trajectories except the zero solution at the origin. For
ε = 0.2, the closed orbit is close to a circle of radius 2. This is typical for small ε
(say, ε < 0.3). For the medium value of ε = 1.0, the circular shape of the closed
orbit is distorted as shown in Figure 2.12(b). For the large value of ε = 5.0, the
closed orbit is severely distorted as shown in Figure 2.13(a). A more revealing phase
portrait in this case can be obtained when the state variables are chosen as z1 = iL
and z2 = vC , resulting in the state equations

1
ż1 = z2 , ż2 = −ε(z1 − z2 + 13 z23 )
ε
The phase portrait in the z1 –z2 plane for ε = 5.0 is shown in Figure 2.13(b). The
closed orbit is very close to the curve z1 = z2 − 13 z23 , except at the corners, where it
2.5. NUMERICAL CONSTRUCTION OF PHASE PORTRAITS 31

x2 x2

x1 x1

(a) (b)

Figure 2.14: (a) A stable limit cycle; (b) an unstable limit cycle.

becomes nearly vertical. The vertical portion of the closed orbit can be viewed as
if the closed orbit jumps from one branch of the curve to the other as it reaches the
corner. Oscillations where the jump phenomenon takes place are usually referred to
as relaxation oscillations. This phase portrait is typical for large values of ε (say,
ε > 3.0). 

The closed orbit in Example 2.4 is different from the harmonic oscillator’s closed
orbit. In the harmonic oscillator there is a continuum of closed orbits, while in the
Van der Pol oscillator there is only one isolated closed orbit. An isolated closed
orbit is called a limit cycle. The limit cycle of the Van der Pol oscillator has the
property that all trajectories in its vicinity approach it as time tends to infinity. A
limit cycle with this property is called stable limit cycle. We shall also encounter
unstable limit cycles, where all trajectories starting arbitrarily close to the limit
cycle move away as time progresses. An example of unstable limit cycle, shown in
Figure 2.14, is given by the Van der Pol equation in reverse time,

ẋ1 = −x2 , ẋ2 = x1 − ε(1 − x21 )x2

whose phase portrait is identical to that of the Van der Pol oscillator, except that
the arrowheads are reversed. Consequently, the limit cycle is unstable.

2.5 Numerical Construction of Phase Portraits

Computer programs for numerical solution of ordinary differential equations are
widely available. They can be effectively used to construct phase portraits for two-
dimensional systems.9 In this section, we give some hints for beginners.10
9 An effective MATLAB-based program for plotting the phase portrait is pplane.
10 These hints are taken from [104, Chapter 10], which contains more instructions on how to
generate informative phase portraits.
32 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

The first step in constructing the phase portrait is to find all equilibrium points
and determine the type of isolated ones via linearization.
Drawing trajectories involves three tasks:11
• Selection of a bounding box in the state plane where trajectories are to be
drawn. The box takes the form

x1min ≤ x1 ≤ x1max , x2min ≤ x2 ≤ x2max

• Selection of initial points (conditions) inside the bounding box.

• Calculation of trajectories.
Let us talk first about calculating trajectories. To find the trajectory passing
through a point x0 , solve the equation

ẋ = f (x), x(0) = x0

in forward time (with positive t) and in reverse time (with negative t). Solution in
reverse time is equivalent to solution in forward time of the equation

ẋ = −f (x), x(0) = x0

since the change of time variable τ = −t reverses the sign of the right-hand side.
The arrowhead on the forward trajectory is placed heading away from x0 , while the
one on the reverse trajectory is placed heading into x0 . Note that solution in reverse
time is the only way we can get a good portrait in the neighborhood of unstable
focus, unstable node, or unstable limit cycle. Trajectories are continued until they
get out of the bounding box. If processing time is a concern, you may want to add
a stopping criterion when trajectories converge to an equilibrium point.
The bounding box should be selected so that all essential qualitative features
are displayed. Since some of these features will not be known a priori, we may have
to adjust the bounding box interactively. However, our initial choice should make
use of all prior information. For example, the box should include all equilibrium
points. Care should be exercised when a trajectory travels out of bounds, for such
a trajectory is either unbounded or is attracted to a stable limit cycle.
The simplest approach to select initial points is to place them uniformly on a grid
throughout the bounding box. However, an evenly spaced set of initial conditions
rarely yields an evenly spaced set of trajectories. A better approach is to select the
initial points interactively after plotting the already calculated trajectories. Since
most computer programs have sophisticated plotting tools, this approach should be
quite feasible.
For a saddle point, we can use linearization to generate the stable and unstable
trajectories. This is useful because, as we saw in Examples 2.2 and 2.3, the stable
11 Afourth task that we left out is placing arrowheads on the trajectory. For the purpose of this
textbook, it can be conveniently done manually.
2.6. EXERCISES 33

trajectories of a saddle define a separatrix. Let the eigenvalues of the linearization

be λ1 > 0 > λ2 and the corresponding eigenvectors be v1 and v2 . The stable and
unstable trajectories of the nonlinear saddle will be tangent to the stable eigenvector
v2 and the unstable eigenvector v1 , respectively, as they approach the equilibrium
point p. Therefore, the two unstable trajectories can be generated from the initial
points x0 = p ± αv1 , where α is a small positive number. Similarly, the two stable
trajectories can be generated from the initial points x0 = p ± αv2 . The major parts
of the unstable trajectories will be generated by solution in forward time, while the
major parts of the stable ones will be generated by solution in reverse time.

2.6 Exercises
2.1 For each of the following systems,

(a) Find all equilibrium points and determine their types.

(b) Construct and discuss the phase portrait.

(1) ẋ1 = −x31 + x2 , ẋ2 = x1 − x32

(2) ẋ1 = x2 , ẋ2 = x1 − x2 − 0.5(x1 + x2 )2

(3) ẋ1 = x2 , ẋ2 = 2x1 + 3x21 + x31 − x2

(4) ẋ1 = −x2 + 14 x32 , ẋ2 = 4x1 − 2x2

(5) ẋ1 = x2 , ẋ2 = −x1 + x2 (1 − x21 + 0.1x41 )

(6) ẋ1 = −x1 + 2x2 + x1 x2 + x22 , ẋ2 = −x1 − x21 − x1 x2

(7) ẋ1 = x2 , ẋ2 = −(x1 + x2 )/(x1 + 2) (defined over the set {x1 > −2})

(8) ẋ1 = −x2 , ẋ2 = 2x1 + 3x2 + 2 sat(−3x2 )

(9) ẋ1 = x2 , ẋ2 = −x1 + 1 3

16 x1 + x2 − 13 x32

2.2 Consider the tunnel-diode circuit of Example 2.2. Keep all parameters the
same except u. Construct and discuss the phase portrait when

(a) u = 0.4;

(b) u = 2.

2.3 Consider the biochemical reactor (A.19) with ν defined by (A.20). Let α = 23,
β = 0.39, and γ = 0.57. Construct and discuss the phase portrait in the region
{x1 ≥ 0, x2 ≥ 0} when

(a) u = 0.5;
34 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

R1 C1

+ v1 −
+
+
v2 C2 R2 g(v2)
− −

Figure 2.15: Exercise 2.5.

(b) u = 1;

(c) u = 1.5.

2.4 Construct and discuss the phase portrait of the negative resistance oscilla-
tor (A.10) when

(a) h(v) = −v + v 3 − v 5 /5 + v 7 /98 and ε = 1.

(b) h(v) = −v + v 3 − v 5 /5 + v 7 /84 and ε = 1.

2.5 An equivalent circuit of the Wien–Bridge oscillator is shown in Figure 2.15

[26], where g(v2 ) is a nonlinear voltage-controlled voltage source.

(a) With x1 = (C1 v1 + C2 v2 )/(C1 + C2 ) and x2 = v2 as state variables, show that

the state model is given by

1 C1 + C2 1 1
ẋ1 = − x2 , ẋ2 = (x1 − x2 ) + g(x2 ) − x2
R2 (C1 + C2 ) C1 C2 R C2 R1 C2 R2

(b) Let C1 = C2 = C, R1 = R2 = R and g(v) = 3.234v − 2.195v 3 + 0.666v 5 .

Construct and discuss the phase portrait in the τ = t/(CR) time scale.

2.6 A reduced-order model of the electrostatic actuator (A.33) can be obtained

if the time constant T is sufficiently small. Writing the ẋ3 -equation as T ẋ3 = (·),
setting T = 0, and using the resultant algebraic equation to eliminate x3 , we obtain

4u2
ẋ1 = x2 , ẋ1 = −x1 − 2ζx2 +
27(1 − x1 )2

where 0 ≤ x1 ≤ 1 − δ. Let u be a constant input, u(t) ≡ U < 1.

(a) Show that there are two equilibrium points and determine their types.
1
(b) Construct and discuss the phase portrait when ζ = 2 and U = 34 .
2.6. EXERCISES 35

2.7 Consider the inverted pendulum equation (A.47) with a = 1.

(a) With u = 0, find all equilibrium points and determine their types.
(b) The energy E = 1 + cos x1 + 12 x22 is defined such that E = 2 at the upward
equilibrium position (0, 0) and E = 0 at the downward one (π, 0). To swing
up the pendulum we need to pump up the energy. Show that the control
u = k(2 − E)x2 cos(x1 ), with k > 0, will make Ė > 0 whenever E < 2 and
x2 cos x1 = 0.
(c) Consider the closed-loop system under the foregoing control. Construct and
discuss the phase portrait when k = 0.1.

2.8 Consider the system

ẋ1 = x2 , ẋ2 = −0.5x1 + 1.5x2 + 0.5u

Construct and discuss the phase portrait for u = 0, the feedback control u =
0.9x1 − 3.2x2 , and the constrained feedback control u = sat(0.9x1 − 3.2x2 ).

2.9 Construct and discuss the phase portrait of the system

 
ẋ1 = x2 , ẋ2 = x1 − x2 + μ sat −2x1 + 12 x2

for each of the values μ = 0.25, 1, and 3.

2.10 The elementary processing units in the central nervous system are the neu-
rons. The FitzHugh-Nagumo model [49] is a dimensionless model that attempts to
capture the dynamics of a single neuron. It is given by

u̇ = u − 13 u3 − w + I, ẇ = ε(b0 + b1 u − w)

where u, w, and I ≥ 0 are the membrane voltage, recovery variable, and applied
current, respectively. The constants ε, b0 and b1 are positive.
(a) Find all equilibrium points and determine their types when b1 > 1.
(b) Repeat part (a) when b1 < 1.
(c) Let ε = 0.1, b0 = 2 and b1 = 1.5. For each of the values I = 0 and I = 2,
construct the phase portrait and discuss the qualitative behavior of the system.
(d) Repeat (c) with b1 = 0.5.

2.11 Consider
 the magnetic
 levitation system (A.29) under the feedback control
u = −1 + sat 12 − x1 − x2 .
(a) Find all equilibrium points and determine their types when c = 1 and when
c = 1 but |(1 − c)/c| < 1.
36 CHAPTER 2. TWO-DIMENSIONAL SYSTEMS

(b) Construct and discuss the phase portrait when b = 0.01 and c = 23 , taking into
consideration the constraint 0 ≤ x1 ≤ 5.

2.12 [77] Consider the nonlinear system

ẋ1 = x2 , ẋ2 = 0.1x32 + u

under the feedback control u = −γ 2 x1 √ − γx2 , with γ > 0. The linearization at the
origin has the eigenvalues (γ/2)(−1 ± j 3), so the response of the linearization can
be made faster by increasing γ. To study the effect of increasing γ on the behavior
√
of the nonlinear system, it is convenient to apply the change of variables z1 = γx1 ,
√
z2 = x2 / γ, and τ = γt.

(a) Show that the transformed state equation in the τ time scale is given by

ż1 = z2 , ż2 = −z1 − z2 + 0.1z23

(b) Construct and discuss the phase portrait in the z1 –z2 plane.

(c) Discuss the effect of increasing γ on the phase portrait in the x1 –x2 plane.
Chapter 3

Stability of Equilibrium Points

Stability theory plays a central role in systems theory and engineering. There are
different kinds of stability problems that arise in the study of dynamical systems.
This chapter is concerned with stability of equilibrium points for time-invariant sys-
tems. In later chapters, we shall deal with time-varying systems and see other kinds
of stability, such as input-to-state stability and input–output stability. An equilib-
rium point is stable if all solutions starting at nearby points stay nearby; otherwise,
it is unstable. It is asymptotically stable if all solutions starting at nearby points
not only stay nearby, but also tend to the equilibrium point as time approaches
infinity. These notions are made precise in Section 3.1. In the same section we
see how to study stability of equilibrium points for linear systems and scalar (or
one-dimensional) nonlinear systems. In Section 3.2 we use linearization to study the
stability of equilibrium points of nonlinear systems. Lyapunov’s method is intro-
duced in Section 3.3 and an extension of the basic theory, known as the invariance
principle, is covered in Section 3.4. Sections 3.5 and 3.6 elaborate on two concepts
that are introduced in Section 3.1, namely, exponential stability and the region of
attraction. The chapter ends in Section 3.7 with two converse Lyapunov theorems,
one for asymptotic stability and the other for its special case of exponential stability.

3.1 Basic Concepts

Consider the n-dimensional time-invariant system

ẋ = f (x) (3.1)

where f is locally Lipschitz over a domain D ⊂ Rn . Suppose x̄ ∈ D is an equilibrium

point of (3.1); that is, f (x̄) = 0. Our goal is to characterize and study the stability
of x̄. For convenience, we state all definitions and theorems for the case when the
equilibrium point is at the origin of Rn ; that is, x̄ = 0. There is no loss of generality
in doing so because any equilibrium point can be shifted to the origin via a change

37
38 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

of variables. Suppose x̄ = 0 and consider the change of variables y = x − x̄. The

derivative of y is given by
def
ẏ = ẋ = f (x) = f (y + x̄) = g(y), where g(0) = 0

In the new variable y, the system has an equilibrium point at the origin. Therefore,
without loss of generality, we assume that f (x) satisfies f (0) = 0 and study the
stability of the origin x = 0.

Definition 3.1 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn , which contains the origin, and f (0) = 0. The equilibrium point x = 0 of
ẋ = f (x) is

• stable if for each ε > 0 there is δ > 0 (dependent on ε) such that

x(0)
< δ ⇒
x(t)
< ε, ∀ t ≥ 0

• unstable if it is not stable.

• asymptotically stable if it is stable and δ can be chosen such that

x(0)
< δ ⇒ lim x(t) = 0
t→∞

The ε–δ requirement for stability takes a challenge-answer form. To demonstrate

that the origin is stable, then, for any value of ε that a challenger may care to desig-
nate, we must produce a value of δ, possibly dependent on ε, such that a trajectory
starting in a δ neighborhood of the origin will never leave the ε neighborhood. The
three types of stability properties can be illustrated by the pendulum example of
Section A.1. The pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 − bx2

has equilibrium points at (0, 0) and (π, 0). Neglecting friction, by setting b = 0,
we have seen in Chapter 2 (Figure 2.2) that trajectories in the neighborhood of
the origin are closed orbits. Therefore, by starting sufficiently close to the origin,
trajectories can be guaranteed to stay within any specified ball centered at the
origin. Hence, the ε–δ requirement for stability is satisfied. The origin, however, is
not asymptotically stable since trajectories do not tend to it eventually. Instead,
they remain in their closed orbits. When friction is taken into consideration (b > 0),
the origin becomes a stable focus. Inspection of the phase portrait of a stable focus
shows that the ε–δ requirement for stability is satisfied. In addition, trajectories
starting close to the equilibrium point tend to it as time tends to infinity. The
equilibrium point at x1 = π is a saddle. Clearly the ε–δ requirement cannot be
satisfied since, for any ε > 0, there is always a trajectory that will leave the ball
{
x − x̄
≤ ε} even when x(0) is arbitrarily close to the equilibrium point x̄.
3.1. BASIC CONCEPTS 39

f(x) f(x) f(x)

x x x

Figure 3.1: Examples of f (x) for which the origin of ẋ = f (x) is unstable.

Having defined stability and asymptotic stability of equilibrium points, our task
now is to find ways to determine stability. The approach we used in the pendulum
example relied on our knowledge of the phase portrait. Trying to generalize this
approach amounts to actually finding all solutions of the state equation, which
may be difficult or even impossible. Lyapunov’s method provides us with a tool
to investigate stability of equilibrium points without the need to solve the state
equation. There are, however, two special cases where we can determine stability
without Lyapunov’ method, namely, scalar and linear systems.
In the scalar (one-dimensional) case, the behavior of x(t) in the neighborhood
of the origin can be determined by examining the sign of f (x). In particular, the
ε–δ requirement for stability is violated if xf (x) > 0 on either side of the origin
x = 0 because in this case a trajectory starting arbitrarily close to the origin will
have to move away from it and cannot be maintained in an ε neighborhood by
choosing δ small enough. Examples of such situation are shown in Figure 3.1.
Consequently, a necessary condition for the origin to be stable is to have xf (x) ≤ 0
in some neighborhood of the origin. It is not hard to see that this condition is also
sufficient. For asymptotic stability, we need to show that there is a neighborhood of
the origin such that trajectories starting in this neighborhood will converge to zero
as time tends to infinity. Clearly, this will not be the case if f (x) is identically zero
on either side of the origin, as in the examples shown in Figure 3.2. Again, it is not
hard to see that the origin will be asymptotically stable if and only if xf (x) < 0
in some neighborhood of the origin; examples of functions satisfying this condition
are shown in Figure 3.3. Let us note an important difference between the two
examples of Figure 3.3. For the system whose right-hand side function is shown in
Figure 3.3(a), limt→∞ x(t) = 0 holds only for initial states in the set {−a < x < b},
while for that of Figure 3.3(b) the limit holds for all initial states. In the first case,
the set {−a < x < b} is said to be the region of attraction, while in the second case
the origin is said to be globally asymptotically stable. These notions are defined
next for n-dimensional systems.

Definition 3.2 Let the origin be an asymptotically stable equilibrium point of the
system ẋ = f (x), where f is a locally Lipschitz function defined over a domain
D ⊂ Rn that contains the origin. Then,
• the region of attraction of the origin (also called region of asymptotic stability,
40 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

f(x) f(x) f(x)

x x x

Figure 3.2: Examples of f (x) for which the origin of ẋ = f (x) is stable but not
asymptotically stable.

f(x) f(x)

−a b x x

(a) (b)

Figure 3.3: Examples of f (x) for which the origin of ẋ = f (x) is asymptotically stable.

domain of attraction, or basin) is the set of all points x0 in D such that the
solution of
ẋ = f (x), x(0) = x0
is defined for all t ≥ 0 and converges to the origin as t tends to infinity;
• the origin is globally asymptotically stable if its region of attraction is the whole
space Rn .

If the origin is a globally asymptotically stable equilibrium point of a system,

then it must be the unique equilibrium point, for if there were another equilibrium
point x̄, the trajectory starting at x̄ would remain at x̄ for all t ≥ 0; hence, it would
not approach the origin, which contradicts the claim that the origin is globally
asymptotically stable. Therefore, global asymptotic stability is not studied for
multiple equilibria systems like the pendulum equation.
For the linear time-invariant system

ẋ = Ax (3.2)

stability properties of the origin can be characterized by the locations of the eigen-
values of the matrix A. Recall from linear system theory1 that the solution of (3.2)
1 See, for example, [2], [22], [59], [62], [70], or [114].
3.1. BASIC CONCEPTS 41

for a given initial state x(0) is given by

x(t) = exp(At)x(0) (3.3)

and that for any matrix A there is a nonsingular matrix P (possibly complex) that
transforms A into its Jordan form; that is,

P −1 AP = J = block diag[J1 , J2 , . . . , Jr ]

where Ji is a Jordan block associated with the eigenvalue λi of A. A Jordan block

of order one takes the form Ji = λi , while a Jordan block of order m > 1 takes the
form ⎡ ⎤
λi 1 0 ... ... 0
⎢ 0 λi 1 0 ... 0 ⎥
⎢ ⎥
⎢ .. .. .. ⎥
⎢ . . . ⎥
⎢ ⎥
Ji = ⎢ . . ⎥
⎢ .. .. 0 ⎥
⎢ ⎥
⎢ . .. ⎥
⎣ .. . 1 ⎦
0 . . . . . . . . . 0 λi m×m
Therefore,

r 
mi
exp(At) = P exp(Jt)P −1 = tk−1 exp(λi t)Rik (3.4)
i=1 k=1

where mi is the order of the Jordan block Ji . If an n × n matrix A has a repeated

eigenvalue λi of algebraic multiplicity qi ,2 then the Jordan blocks associated with λi
have order one if and only if rank(A−λi I) = n−qi . The next theorem characterizes
the stability properties of the origin.
Theorem 3.1 The equilibrium point x = 0 of ẋ = Ax is stable if and only if all
eigenvalues of A satisfy Re[λi ] ≤ 0 and for every eigenvalue with Re[λi ] = 0 and
algebraic multiplicity qi ≥ 2, rank(A − λi I) = n − qi , where n is the dimension of
x. The equilibrium point x = 0 is globally asymptotically stable if and only if all
eigenvalues of A satisfy Re[λi ] < 0. 3
Proof: From (3.3) we can see that the origin is stable if and only if exp(At) is a
bounded function of t for all t ≥ 0. If one of the eigenvalues of A has Re[λi ] > 0, the
corresponding exponential term exp(λi t) in (3.4) will grow unbounded as t → ∞.
Therefore, we must restrict the eigenvalues to have Re[λi ] ≤ 0. However, those
eigenvalues with Re[λi ] = 0 (if any) could give rise to unbounded terms if the order
of an associated Jordan block is higher than one, due to the term tk−1 in (3.4).
Therefore, we must restrict eigenvalues with zero real parts to have Jordan blocks
of order one, which is equivalent to the rank condition rank(A − λi I) = n − qi .
Thus, we conclude that the condition for stability is a necessary one. It is clear that
2 Equivalently, qi is the multiplicity of λi as a zero of det(λI − A).
42 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

(a) (b)

Figure 3.4: (a) Series connection; (b) parallel connection.

the condition is also sufficient to ensure that exp(At) is bounded. For asymptotic
stability of the origin, exp(At) must approach 0 as t → ∞. From (3.4), this is the
case if and only if Re[λi ] < 0, ∀ i. Since x(t) depends linearly on the initial state
x(0), asymptotic stability of the origin is global. 2

The proof shows, mathematically, why repeated eigenvalues on the imaginary

axis must satisfy the rank condition rank(A − λi I) = n − qi . The next example
sheds some light on the physical meaning of this requirement.

Example 3.1 Figure 3.4 shows a series connection and a parallel connection of two
identical systems. Each system is represented by the state model

0 1 0  
ẋ = x+ u, y= 1 0 x
−1 0 1

where u and y are the input and output, respectively. Let As and Ap be the matrices
of the series and parallel connections, when modeled in the form (3.2) (no driving
inputs). Then
⎡ ⎤ ⎡ ⎤
0 1 0 0 0 1 0 0
⎢ −1 0 0 0 ⎥ ⎢ 0 0 ⎥
Ap = ⎢ ⎥ and As = ⎢ −1 0 ⎥
⎣ 0 0 0 1 ⎦ ⎣ 0 0 0 1 ⎦
0 0 −1 0 1 0 −1 0

The matrices Ap and As have the same eigenvalues √ on the imaginary axis, ±j,
with algebraic multiplicity qi = 2, where j = −1. It can be easily checked that
rank(Ap −jI) = 2 = n−qi , while rank(As −jI) = 3 = n−qi . Thus, by Theorem 3.1,
the origin of the parallel connection is stable, while the origin of the series connection
is unstable. To physically see the difference between the two cases, notice that in
the parallel connection, nonzero initial conditions produce sinusoidal oscillations
of frequency 1 rad/sec, which are bounded functions of time. The sum of these
sinusoidal signals remains bounded. On the other hand, nonzero initial conditions
in the first component of the series connection produce a sinusoidal oscillation of
frequency 1 rad/sec, which acts as a driving input to the second component. Since
3.2. LINEARIZATION 43

the undamped natural frequency of the second component is 1 rad/sec, the driving
input causes “resonance” and the response grows unbounded. 
When all eigenvalues of A satisfy Re[λi ] < 0, A is called a Hurwitz matrix or a
stability matrix. The origin of ẋ = Ax is asymptotically stable if and only if A is
Hurwitz. In this case, its solution satisfies the inequality

x(t) ≤ kx(0)e−λt , ∀t≥0 (3.5)

for some positive constants k and λ, as can be seen from (3.3) and (3.4). When the
solution of a nonlinear system satisfies a similar inequality, the equilibrium point is
said to be exponentially stable, as stated in the following definition.
Definition 3.3 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn , which contains the origin, and f (0) = 0. The equilibrium point x = 0 of
ẋ = f (x) is exponentially stable if there exist positive constants c, k, and λ such
that inequality (3.5) is satisfied for all x(0) < c. It is globally exponentially stable
if the inequality is satisfied for every initial state x(0).
Exponential stability is a stronger form of asymptotic stability in the sense that
if the equilibrium point is exponentially stable, then it is asymptotically stable as
well. The opposite is not always true, as demonstrated by the following example.
Example 3.2 The origin is an asymptotically stable equilibrium point of the scalar
system ẋ = −x3 because xf (x) < 0 for x = 0. However, the solution
x(0)
x(t) = 
1 + 2tx2 (0)

does not satisfy inequality of the form (3.5) because if it did, it would follow that

e2λt
≤ k2
1 + 2tx2 (0)

for all t ≥ 0, which is impossible because limt→∞ e2λt /[1 + 2tx2 (0)] = ∞. 

3.2 Linearization
Consider the n-dimensional system

ẋ = f (x) (3.6)

where f is a continuously differentiable function over D = {x < r} for some

r > 0, and f (0) = 0. Let J(x) be the Jacobian matrix of f (x); that is,
∂f
J(x) = (x)
∂x
44 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

By (B.6) and f (0) = 0, we have

 1
f (x) = J(σx) dσ x
0

Let A = J(0). By adding and subtracting Ax to the right-hand side of the foregoing
equation, we obtain
 1
f (x) = [A + G(x)]x, where G(x) = [J(σx) − J(0)] dσ
0

By continuity of J(x), we see that G(x) → 0 as x → 0. This suggests that in a small

neighborhood of the origin we can approximate the nonlinear system ẋ = f (x) by its
linearization about the origin ẋ = Ax. The next theorem spells out conditions under
which we can draw conclusions about the stability of the origin as an equilibrium
point for the nonlinear system by investigating its stability as an equilibrium point
for the linear system. The theorem is known as Lyapunov’s indirect method.3
Theorem 3.2 Let x = 0 be an equilibrium point for the nonlinear system ẋ = f (x),
where f is continuously differentiable in a neighborhood of the origin. Let

∂f 
A= (x)
∂x x=0

and denote its eigenvalues by λ1 to λn . Then,

1. The origin is exponentially stable if and only if Re[λi ] < 0 for all eigenvalues.
2. The origin is unstable if Re[λi ] > 0 for one or more of the eigenvalues.
3
Theorem 3.2 provides us with a simple procedure for determining the stability of
an equilibrium point by testing the eigenvalues of the Jacobian at that point. The
theorem, however, does not say anything about the case when Re[λi ] ≤ 0 for all i,
with Re[λi ] = 0 for some i. In this case, linearization fails to determine the stability
of the equilibrium point, although we can say for sure that if the equilibrium point
turns out to be asymptotically stable, it will not be exponentially stable because
the condition for exponential stability is both necessary and sufficient.

Example 3.3 Consider the scalar system ẋ = ax3 . Linearizing the system about
the origin yields 
∂f  
A=  = 3ax2 x=0 = 0
∂x x=0
3 See [74, Theorems 4.7 and 4.15] for the proof of Theorem 3.2. For the first item of the theorem,

a proof of the sufficiency part is given in Section 3.5 while a proof of the necessity part is given in
Section 3.7.
3.3. LYAPUNOV’S METHOD 45

There is one eigenvalue that lies on the imaginary axis. Hence, linearization fails
to determine the stability of the origin. This failure is genuine in the sense that the
origin could be asymptotically stable, stable, or unstable, depending on the value
of the parameter a. If a < 0, the origin is asymptotically stable as can be seen from
the condition xf (x) = ax4 < 0 for x = 0. If a = 0, the system is linear and the
origin is stable according to Theorem 3.1. If a > 0, the origin is unstable as can be
seen from the condition xf (x) = ax4 > 0 for x = 0. Notice that when a < 0, the
origin is not exponentially stable, as we saw in Example 3.2.4 
Example 3.4 The pendulum equation
ẋ1 = x2 , ẋ2 = − sin x1 − bx2
has equilibrium points at (0, 0) and (π, 0). The Jacobian matrix is
⎡ ∂f1 ∂f1 ⎤ ⎡ ⎤
∂x1 ∂x2 0 1
∂f
=⎣ ⎦=⎣ ⎦
∂x ∂f2
∂x
∂f2
∂x
− cos x1 −b
1 2

To determine the stability of the origin, we evaluate the Jacobian at x = 0:



∂f  0 1
A= =
∂x x=0 −1 −b
√
The eigenvalues of A are λ1,2 = − 12 b ± 12 b2 − 4. For all b > 0, the eigenvalues
satisfy Re[λi ] < 0. Consequently, the equilibrium point at the origin is exponentially
stable. In the absence of friction (b = 0), both eigenvalues are on the imaginary
axis. Thus, we cannot determine the stability of the origin through linearization.
We know from the phase portrait (Figure 2.2) that, in this case, the origin is a
center; hence, it is a stable equilibrium point. To determine the stability of the
equilibrium point at (π, 0), we evaluate the Jacobian at that point. This is equivalent
to performing a change of variables z1 = x1 − π, z2 = x2 to shift the equilibrium
point to the origin and then evaluating the Jacobian [∂f /∂z] at z = 0:


∂f  0 1
à = =
∂x x1 =π,x2 =0 1 −b
√
The eigenvalues of à are λ1,2 = − 12 b ± 12 b2 + 4. For all b ≥ 0, there is one
eigenvalue with Re[λi ] > 0. Hence, the equilibrium point at (π, 0) is unstable. 

3.3 Lyapunov’s Method

Reconsider the pendulum equation
ẋ1 = x2 , ẋ2 = − sin x1 − bx2
4 While the example dealt with the case a = −1, its conclusion extends to any a < 0 by changing
the time variable from t to τ = −at, transforming ẋ = ax3 to dx/dτ = −x3 .
46 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

where b ≥ 0. Using phase portraits and/or linearization, we have argued in the

previous two sections that the origin is a stable equilibrium point when b = 0 and
asymptotically stable when b > 0. These conclusions can also be reached using
energy concepts. Let us define the energy of the pendulum E(x) as the sum of
its potential and kinetic energies, with the reference of the potential energy chosen
such that E(0) = 0; that is,
 x1
E(x) = sin y dy + 12 x22 = (1 − cos x1 ) + 12 x22
0

When friction is neglected (b = 0), the system is conservative; that is, there is no
dissipation of energy. Hence, E = constant during the motion of the system or, in
other words, dE/dt = 0 along the trajectories of the system. Since E(x) = c forms
a closed contour around x = 0 for small c, we can again arrive at the conclusion
that x = 0 is a stable equilibrium point. When friction is accounted for (b > 0),
energy will dissipate during the motion of the system, that is, dE/dt ≤ 0 along the
trajectories of the system. Due to friction, E cannot remain constant indefinitely
while the system is in motion. Hence, it keeps decreasing until it eventually reaches
zero, showing that the trajectory tends to x = 0 as t tends to ∞. Thus, by examining
the derivative of E along the trajectories of the system, it is possible to determine
the stability of the equilibrium point. In 1892, Lyapunov showed that certain other
functions could be used instead of energy to determine stability of an equilibrium
point.5 Let V (x) be a continuously differentiable real-valued function defined in a
domain D ⊂ Rn that contains the origin. The derivative of V along the trajectories
of ẋ = f (x), denoted by V̇ (x), is given by

n
∂V n
∂V
V̇ (x) = ẋi = fi (x)
i=1
∂xi i=1
∂xi
⎡ ⎤
f1 (x)
 ∂V ∂V ∂V
⎢
⎢ f2 (x) ⎥
⎥ ∂V
= ∂x1 , ∂x2 , ... , ⎢ .. ⎥ = f (x)
∂xn
⎣ . ⎦ ∂x
fn (x)

The derivative of V along the trajectories of a system is dependent on the system’s

equation. Hence, V̇ (x) will be different for different systems. If φ(t; x) is the solution
of ẋ = f (x) that starts at initial state x at time t = 0, then

d 
V̇ (x) = V (φ(t; x))
dt t=0

Therefore, if V̇ (x) is negative, V will decrease along the solution of ẋ = f (x). This
observation leads to Lyapunov’s stability theorem, given next.
5 Classical references on Lyapunov stability include [54, 78, 111, 152, 156].
3.3. LYAPUNOV’S METHOD 47

D
Br

Bδ

Ωβ

Figure 3.5: Geometric representation of sets in the proof of Theorem 3.3.

Theorem 3.3 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn , which contains the origin, and f (0) = 0. Let V (x) be a continuously differen-
tiable function defined over D such that

V (0) = 0 and V (x) > 0 for all x ∈ D with x = 0 (3.7)

V̇ (x) ≤ 0 for all x ∈ D (3.8)

Then, the origin is a stable equilibrium point of ẋ = f (x). Moreover, if

V̇ (x) < 0 for all x ∈ D with x = 0 (3.9)

then origin is asymptotically stable. Furthermore, if D = Rn , (3.7) and (3.9) hold

for all x = 0, and
x → ∞ ⇒ V (x) → ∞ (3.10)
then the origin is globally asymptotically stable. 3

Proof: Given ε > 0, choose r ∈ (0, ε] such that Br = {x ≤ r} ⊂ D.6 Let
α = minx=r V (x). By (3.7), α > 0 . Take β ∈ (0, α) and let

Ωβ = {x ∈ Br | V (x) ≤ β}

Then, Ωβ is in the interior of Br .7 (See Figure 3.5.) The set Ωβ has the property
6 The definition of stability should hold for any ε > 0, but the analysis has to be limited to the

domain D. Because r ≤ ε, if we show that x(t) ≤ r we will have x(t) ≤ ε.

7 This fact can be shown by contradiction. Suppose Ω is not in the interior of B , then there
β r
is a point p ∈ Ωβ that lies on the boundary of Br . At this point, V (p) ≥ α > β, but for all
x ∈ Ωβ , V (x) ≤ β, which is a contradiction.
48 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

that any trajectory starting in Ωβ at t = 0 stays in Ωβ for all t ≥ 0. This follows

from (3.8) since

V̇ (x(t)) ≤ 0 ⇒ V (x(t)) ≤ V (x(0)) ≤ β, ∀ t ≥ 0

Because Ωβ is a compact set,8 we conclude from Lemma 1.3 that ẋ = f (x) has a
unique solution defined for all t ≥ 0 whenever x(0) ∈ Ωβ . As V (x) is continuous
and V (0) = 0, there is δ > 0 such that

x ≤ δ ⇒ V (x) < β

Then,
Bδ ⊂ Ωβ ⊂ Br
and
x(0) ∈ Bδ ⇒ x(0) ∈ Ωβ ⇒ x(t) ∈ Ωβ ⇒ x(t) ∈ Br
Therefore,
x(0) < δ ⇒ x(t) < r ≤ ε, ∀ t ≥ 0
which shows that the origin is stable. Now assume that (3.9) holds. To show
asymptotic stability we need to show that x(t) → 0 as t → ∞. that is, for every
a > 0, there is T > 0 such that x(t) < a, for all t > T . By repetition of previous
arguments, we know that for every a > 0, we can choose b > 0 such that Ωb ⊂ Ba .
Therefore, it is sufficient to show that V (x(t)) → 0 as t → ∞. Since V (x(t)) is
monotonically decreasing and bounded from below by zero,

V (x(t)) → c ≥ 0 as t → ∞

We show that c = 0 by a contradiction argument. Suppose c > 0. By continuity of

V (x), there is d > 0 such that Bd ⊂ Ωc . The limit V (x(t)) → c > 0 implies that the
trajectory x(t) lies outside the ball Bd for all t ≥ 0. Let −γ = maxd≤x≤r V̇ (x),
which exists because the continuous function V̇ (x) has a maximum over the compact
set {d ≤ x ≤ r}.9 By (3.9), −γ < 0. It follows that
 t
V (x(t)) = V (x(0)) + V̇ (x(τ )) dτ ≤ V (x(0)) − γt
0

Since the right-hand side will eventually become negative, the inequality contra-
dicts the assumption that c > 0. Hence, the origin is asymptotically stable. To
show global asymptotic stability, note that condition (3.10) implies that the set
Ωc = {V (x) ≤ c} is compact for every c > 0. This is so because for any c > 0,
there is r > 0 such that V (x) > c whenever x > r. Thus, Ωc ⊂ Br . From the
foregoing arguments, it is clear that all solutions starting Ωc will converge to the
8Ω
β is closed by definition and bounded since it is contained in Br .
9 See [3, Theorem 4-20].
3.3. LYAPUNOV’S METHOD 49

c3
c2

V (x) = c1

c1<c2<c3

Figure 3.6: Level surfaces of a Lyapunov function.

origin. Since for any point p ∈ Rn , we can choose c = V (p), we conclude that the
origin is globally asymptotically stable. 2

A continuously differentiable function V (x) satisfying (3.7) and (3.8) is called a

Lyapunov function. The surface V (x) = c, for some c > 0, is called a Lyapunov or
level surface. Using Lyapunov surfaces, Figure 3.6 makes the theorem intuitively
clear. It shows Lyapunov surfaces for increasing values of c. The condition V̇ ≤ 0
implies that when a trajectory crosses a Lyapunov surface V (x) = c, it moves inside
the set Ωc = {V (x) ≤ c} and can never come out again. When V̇ < 0, the trajectory
moves from one Lyapunov surface to an inner Lyapunov surface with a smaller c. As
c decreases, the Lyapunov surface V (x) = c shrinks to the origin, showing that the
trajectory approaches the origin as time progresses. If we only know that V̇ ≤ 0, we
cannot be sure that the trajectory will approach the origin,10 but we can conclude
that the origin is stable since the trajectory can be contained inside any ball Bε by
requiring the initial state x(0) to lie inside a Lyapunov surface contained in that
ball.
A function V (x) satisfying (3.7)— that is, V (0) = 0 and V (x) > 0 for x = 0—
is said to be positive definite. If it satisfies the weaker condition V (x) ≥ 0 for
x = 0, it is said to be positive semidefinite. A function V (x) is said to be negative
definite or negative semidefiniteif −V (x) is positive definite or positive semidefinite,
respectively. If V (x) does not have a definite sign as per one of these four cases,
it is said to be indefinite. A function V (x) satisfying (3.10) is said to be radially
unbounded. With this terminology, we can rephrase Lyapunov’s theorem to say that
the origin is stable if there is a continuously differentiable positive definite function
V (x) so that V̇ (x) is negative semidefinite, and it is asymptotically stable if V̇ (x) is
negative definite. It is globally asymptotically stable if the conditions for asymptotic
stability hold globally and V (x) is radially unbounded.
The radial unboundedness condition (3.10) is used in the proof of Theorem 3.3
to ensure that the set Ωc = {V (x) ≤ c} is bounded for every c > 0. Without such
10 See, however, LaSalle’s theorem in Section 3.4
50 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

x2

x1

Figure 3.7: Lyapunov surfaces for V (x) = x21 /(1 + x21 ) + x22 .

condition the set Ωc might not be bounded for large c. Consider, for example, the
function
x21
V (x) = + x22
1 + x21
which is not radially unbounded because on the x1 -axis, limx→∞ V (x) = 1. Fig-
ure 3.7 shows the surfaces V (x) = c for various positive values of c. For small c,
the surface is closed; hence, Ωc is bounded since it is contained in a closed ball Br
for some r > 0. This is a consequence of the continuity and positive definiteness of
V (x). As c increases, the surface V (x) = c remains closed for c < 1, but for c > 1,
it is open and Ωc is unbounded.
For the quadratic form

n 
n
V (x) = xT P x = pij xi xj
i=1 j=1

where P is a real symmetric matrix, V (x) is positive definite (positive semidefinite)

if and only if all the eigenvalues of P are positive (nonnegative), which is true if and
only if all the leading principal minors of P are positive (all principal minors of P
are nonnegative).11 If V (x) = xT P x is positive definite (positive semidefinite), we
say that the matrix P is positive definite (positive semidefinite) and write P > 0
(P ≥ 0). Since xT P x ≥ λmin (P )x2 , where λmin (P ) is the minimum eigenvalue of
P , we see that if xT P x positive definite, it will also be radially unbounded.

Example 3.5 Consider

V (x) = ax21 + 2x1 x3 + ax22 + 4x2 x3 + ax23

⎡ ⎤⎡ ⎤
a 0 1 x1
= [x1 x2 x3 ] ⎣ 0 a 2 ⎦ ⎣ x2 ⎦ = xT P x
1 2 a x3
11 This is a well-known fact in matrix theory. Its proof can be found in [15] or [44].
3.3. LYAPUNOV’S METHOD 51

The leading principal

√ minors of P are a, a2 , and a(a2 −5). Therefore, V (x) is positive
definite if a > 5. For negative definiteness, the leading principal minors of −P
should be positive; that is, the leading principal minors of P should have alternating
signs, with the odd-numbered minors being negative and the even-numbered
√ minors
being positive. Consequently, V (x) is negative definite if a < − 5. By calculating
√
all principal minors, it can be √seen that V (x) √
is positive
√ semidefinite if a ≥ 5 and
negative semidefinite if a ≤ − 5. For a ∈ (− 5, 5), V (x) is indefinite. 

Lyapunov’s theorem can be applied without solving the differential equation

ẋ = f (x). On the other hand, there is no systematic method for finding Lyapunov
functions. In some cases, there are natural Lyapunov function candidates like energy
functions in electrical or mechanical systems. In other cases, it is basically a matter
of trial and error. The situation, however, is not as bad as it might seem. As we
go over various examples and applications throughout the book, some ideas and
approaches for searching for Lyapunov functions will be delineated.

Example 3.6 Consider the pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 − bx2

and study the stability of the equilibrium point at the origin. Let us start with the
no-friction case when b = 0. A natural Lyapunov function candidate is the energy

V (x) = (1 − cos x1 ) + 12 x22

Clearly, V (0) = 0 and V (x) is positive definite over the domain −2π < x1 < 2π.
The derivative of V (x) along the trajectories of the system is given by

V̇ (x) = ẋ1 sin x1 + x2 ẋ2 = x2 sin x1 − x2 sin x1 = 0

Thus, conditions (3.7) and (3.8) of Theorem 3.3 are satisfied, and we conclude that
the origin is stable. Since V̇ (x) ≡ 0, we can also conclude that the origin is not
asymptotically stable; for trajectories starting on a Lyapunov surface V (x) = c
remain on the same surface for all future time. Next, we consider the pendulum
with friction; that is, when b > 0. With V (x) = (1 − cos x1 ) + 12 x22 as a Lyapunov
function candidate, we obtain

V̇ (x) = ẋ1 sin x1 + x2 ẋ2 = − bx22

The derivative V̇ (x) is negative semidefinite. It is not negative definite because

V̇ (x) = 0 for x2 = 0 irrespective of the value of x1 ; that is, V̇ (x) = 0 along the
x1 -axis. Therefore, we can only conclude that the origin is stable. However, using
the phase portrait of the pendulum equation, we have seen that when b > 0, the
origin is asymptotically stable. The energy Lyapunov function fails to show this
fact. We will see later in Section 3.4 that LaSalle’s theorem will enable us to arrive
at a different conclusion. For now, let us look for a Lyapunov function V (x) that
52 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

would have a negative definite V̇ (x). Starting from the energy Lyapunov function,
let us replace the term 12 x22 by the more general quadratic form 12 xT P x for some
2 × 2 positive definite matrix P :

p11 p12 x1
V (x) = 12 xT P x + (1 − cos x1 ) = 12 [x1 x2 ] + (1 − cos x1 )
p12 p22 x2

The quadratic form xT P x is positive definite when

p11 > 0, p11 p22 − p212 > 0

The derivative V̇ (x) is given by

V̇ (x) = (p11 x1 + p12 x2 + sin x1 ) x2 + (p12 x1 + p22 x2 ) (− sin x1 − bx2 )

= (1 − p22 )x2 sin x1 − p12 x1 sin x1 + (p11 − p12 b) x1 x2 + (p12 − p22 b) x22

Now we want to choose p11 , p12 , and p22 such that V̇ (x) is negative definite. Since
the cross product terms x2 sin x1 and x1 x2 are sign indefinite, we will cancel them
by taking p22 = 1 and p11 = bp12 . With these choices, p12 must satisfy 0 < p12 < b
for V (x) to be positive definite. Let us take p12 = 12 b. Then,

V̇ (x) = − 12 bx1 sin x1 − 12 bx22

The term x1 sin x1 is positive for all 0 < |x1 | < π. Taking D = {|x1 | < π}, we
see that V (x) is positive definite and V̇ (x) is negative definite over D. Thus, by
Theorem 3.3, we conclude that the origin is asymptotically stable. 

This example emphasizes an important feature of Lyapunov’s stability theorem;

namely, the theorem’s conditions are only sufficient. Failure of a Lyapunov function
candidate to satisfy the conditions for stability or asymptotic stability does not
mean that the equilibrium point is not stable or asymptotically stable. It only
means that such stability property cannot be established by using this Lyapunov
function candidate. Whether the equilibrium point is stable (asymptotically stable)
or not can be determined only by further investigation.
In searching for a Lyapunov function in Example 3.6, we approached the problem
in a backward manner. We investigated an expression for the derivative V̇ (x) and
went back to choose the parameters of V (x) so as to make V̇ (x) negative definite.
This is a useful idea in searching for a Lyapunov function. A procedure that exploits
this idea is known as the variable gradient method. To describe the procedure, let
V (x) be a scalar function of x and g(x) = ∇V = (∂V /∂x)T . The derivative V̇ (x)
along the trajectories of ẋ = f (x) is given by
∂V
V̇ (x) = f (x) = g T (x)f (x)
∂x
The idea now is to try to choose g(x) as the gradient of a positive definite function
V (x) that would make V̇ (x) negative definite. It is not difficult to verify that g(x)
3.3. LYAPUNOV’S METHOD 53

is the gradient of a scalar function if and only if the Jacobian matrix [∂g/∂x] is
symmetric; that is,
∂gi ∂gj
= , ∀ i, j = 1, . . . , n
∂xj ∂xi
Under this constraint, we start by choosing g(x) such that g T (x)f (x) is negative
definite. The function V (x) is then computed from the integral
 x  x
n
T
V (x) = g (y) dy = gi (y) dyi
0 0 i=1

The integration is taken over any path joining the origin to x.12 Usually, this is
done along the axes; that is,
 x1  x2
V (x) = g1 (y1 , 0, . . . , 0) dy1 + g2 (x1 , y2 , 0, . . . , 0) dy2
0 0
 xn
+ ··· + gn (x1 , x2 , . . . , xn−1 , yn ) dyn
0

By leaving some parameters of g(x) undetermined, one would try to choose them to
ensure that V (x) is positive definite. The variable gradient method can be used to
arrive at the Lyapunov function of Example 3.6. Instead of repeating the example,
we illustrate the method on a slightly more general system.

Example 3.7 Consider the two-dimensional system

ẋ1 = x2 , ẋ2 = −h(x1 ) − ax2

where a > 0, h(·) is locally Lipschitz, h(0) = 0, and yh(y) > 0 for all y = 0, y ∈
(−b, c) for some positive constants b and c. The pendulum equation is a special
case of this system. To apply the variable gradient method, we want to choose a
two-dimensional vector g(x) that satisfies

∂g1 ∂g2
=
∂x2 ∂x1

V̇ (x) = g1 (x)x2 − g2 (x)[h(x1 ) + ax2 ] < 0, for x = 0

and  x
V (x) = g T (y) dy > 0, for x = 0
0
Let us try

φ1 (x1 ) + ψ1 (x2 )
g(x) =
φ2 (x1 ) + ψ2 (x2 )
12 The line integral of a gradient vector is independent of the path. (See [3, Theorem 10-37].)
54 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

where the scalar functions φ1 , φ2 , ψ1 , and ψ2 are to be determined. To satisfy the

symmetry requirement, we must have
∂ψ1 ∂φ2
=
∂x2 ∂x1
which can be achieved by taking ψ1 (x2 ) = γx2 and φ2 (x1 ) = γx1 for some constant
γ. The derivative V̇ (x) is given by

V̇ (x) = x2 φ1 (x1 ) + γx22 − γx1 h(x1 ) − aγx1 x2 − ψ2 (x2 )h(x1 ) − ax2 ψ2 (x2 )

To cancel the cross-product terms, we need

x2 φ1 (x1 ) − aγx1 x2 − ψ2 (x2 )h(x1 ) = 0

which can be achieved by taking ψ2 (x2 ) = δx2 and φ1 (x1 ) = aγx1 + δh(x1 ). Then,

aγx1 + δh(x1 ) + γx2

V̇ (x) = −γx1 h(x1 ) − (aδ − γ)x2 and g(x) =
2
γx1 + δx2

By integration, we obtain
 x1  x2
V (x) = [aγy1 + δh(y1 )] dy1 + (γx1 + δy2 ) dy2
0 0
 x1  x1
1 2 1 2 1 T
= 2 aγx1 + δ h(y) dy + γx1 x2 + 2 δx2 = 2 x P x + δ h(y) dy
0 0

where

aγ γ
P =
γ δ
Choosing δ > 0 and 0 < γ < aδ ensures that V (x) is positive definite and V̇ (x) is
negative definite. For example, taking γ = akδ for 0 < k < 1 yields the Lyapunov
function
 x1
δ ka2 ka
V (x) = xT x+δ h(y) dy
2 ka 1 0

which satisfies conditions (3.7) and (3.9) of Theorem 3.3 over the domain D =
{−b < x1 < c}. Therefore, the origin is asymptotically stable. Suppose now that
the condition yh(y) > 0 holds for all y = 0. Then, (3.7) and (3.9) hold globally
and V (x) is radially unbounded because V (x) ≥ 12 xT P x and xT P x is radially
unbounded. Therefore, the origin is globally asymptotically stable. 

3.4 The Invariance Principle

In our study of the pendulum equation with friction (Example 3.6), we saw that
the energy Lyapunov function fails to satisfy the asymptotic stability condition of
3.4. THE INVARIANCE PRINCIPLE 55

Theorem 3.3 because V̇ (x) = −bx22 is only negative semidefinite. Notice, however,
that V̇ (x) is negative everywhere, except on the line x2 = 0, where V̇ (x) = 0. For
the system to maintain the V̇ (x) = 0 condition, the trajectory of the system must
be confined to the line x2 = 0. Unless x1 = 0, this is impossible because from the
pendulum equation

x2 (t) ≡ 0 ⇒ ẋ2 (t) ≡ 0 ⇒ sin x1 (t) ≡ 0

Hence, in the segment −π < x1 < π of the x2 = 0 line, the system can maintain
the V̇ (x) = 0 condition only at the origin x = 0. Therefore, V (x(t)) must decrease
toward 0 and, consequently, x(t) → 0 as t → ∞, which is consistent with the fact
that, due to friction, energy cannot remain constant while the system is in motion.
The foregoing argument shows, formally, that if in a domain about the origin we
can find a Lyapunov function whose derivative along the trajectories of the system is
negative semidefinite, and if we can establish that no trajectory can stay identically
at points where V̇ (x) = 0, except at the origin, then the origin is asymptotically
stable. This idea follows from LaSalle’s invariance principle, which is the subject
of this section.13 To state and prove LaSalle’s invariance theorem, we need to
introduce a few definitions. Let x(t) be a solution of ẋ = f (x). A point p is a
positive limit point of x(t) if there is a sequence {tn }, with tn → ∞ as n → ∞, such
that x(tn ) → p as n → ∞. The set of all positive limit points of x(t) is called the
positive limit set of x(t). A set M is an invariant set with respect to ẋ = f (x) if

x(0) ∈ M ⇒ x(t) ∈ M, ∀ t ∈ R

That is, if a solution belongs to M at some time instant, then it belongs to M for
all future and past time. A set M is positively invariant if

x(0) ∈ M ⇒ x(t) ∈ M, ∀ t ≥ 0

We also say that x(t) approaches a set M as t approaches infinity, if for each ε > 0
there is T > 0 such that

dist(x(t), M ) < ε, ∀ t > T

where dist(p, M ) denotes the distance from a point p to a set M , that is, the shortest
distance from p to any point in M . More precisely,

dist(p, M ) = inf p − x
x∈M

These few concepts can be illustrated by examining asymptotically stable equilib-

rium points and stable limit cycles in the plane. An asymptotically stable equi-
librium point is the positive limit set of every solution starting in its vicinity. A
stable limit cycle is the positive limit set of every solution starting in its vicinity.
13 The presentation follows the original work [83].
56 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

In the latter case, the solution approaches the limit cycle as t → ∞, but does not
approach any specific point on it. In other words, the statement x(t) approaches
M as t → ∞ does not imply that limt→∞ x(t) exists. Equilibrium points and limit
cycles are invariant sets, since any solution starting in the set remains in it for all
t ∈ R. The set Ωc = {V (x) ≤ c}, with V̇ (x) ≤ 0 in Ωc , is positively invariant since,
as we saw in the proof of Theorem 3.3, a solution starting in Ωc remains in Ωc for
all t ≥ 0.
A fundamental property of limit sets is stated in the next lemma.14
Lemma 3.1 Let f (x) be a locally Lipschitz function defined over a domain D ⊂ Rn .
If a solution x(t) of ẋ = f (x) is bounded and belongs to D for t ≥ 0, then its positive
limit set L+ is a nonempty, compact, invariant set. Moreover, x(t) approaches L+
as t → ∞. 3
We are now ready to state LaSalle’s theorem.
Theorem 3.4 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn and Ω ⊂ D be a compact set that is positively invariant with respect to ẋ = f (x).
Let V (x) be a continuously differentiable function defined over D such that V̇ (x) ≤ 0
in Ω. Let E be the set of all points in Ω where V̇ (x) = 0, and M be the largest
invariant set in E. Then every solution starting in Ω approaches M as t → ∞. 3
Proof: Let x(t) be a solution of ẋ = f (x) starting in Ω. Since V̇ (x) ≤ 0 in Ω,
V (x(t)) is a decreasing function of t. Since V (x) is continuous on the compact set
Ω, it is bounded from below on Ω. Therefore, V (x(t)) has a limit a as t → ∞.
Note also that the positive limit set L+ is in Ω because Ω is a closed set. For
any p ∈ L+ , there is a sequence tn with tn → ∞ and x(tn ) → p as n → ∞. By
continuity of V (x), V (p) = limn→∞ V (x(tn )) = a. Hence, V (x) = a on L+ . Since
(by Lemma 3.1) L+ is an invariant set, V̇ (x) = 0 on L+ . Thus,
L+ ⊂ M ⊂ E ⊂ Ω
Since x(t) is bounded, x(t) approaches L+ as t → ∞ (by Lemma 3.1). Hence, x(t)
approaches M as t → ∞. 2

Unlike Lyapunov’s theorem, Theorem 3.4 does not require the function V (x) to
be positive definite. Note also that the construction of the set Ω does not have to be
tied in with the construction of the function V (x). However, in many applications
the construction of V (x) will itself guarantee the existence of a set Ω. In particular,
if Ωc = {V (x) ≤ c} is bounded and V̇ (x) ≤ 0 in Ωc , then we can take Ω = Ωc .
When V (x) is positive definite, then for sufficiently small c > 0 the set Ωc has a
bounded component that contains the origin.15
14 See [74, Lemma 4.1] for a proof of this lemma.
15 The set {V (x) ≤ c} may have more than one component. For example, if V (x) =x2 /(1 +
  √
4 1
x ), the set {V (x) ≤ 4 } has two components: a bounded component |x| ≤ 2 − 3 and an
  √ 
unbounded component |x| ≥ 2 + 3 .
3.4. THE INVARIANCE PRINCIPLE 57

When our interest is in showing that x(t) → 0 as t → ∞, we need to establish

that the largest invariant set in E is the origin. This is done by showing that
no solution can stay identically in E, other than the trivial solution x(t) ≡ 0.
Specializing Theorem 3.4 to this case and taking V (x) to be positive definite, we
obtain the following theorem that extends Theorem 3.3.16

Theorem 3.5 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn , which contains the origin, and f (0) = 0. Let V (x) be a continuously differ-
entiable positive definite function defined over D such that V̇ (x) ≤ 0 in D. Let
S = {x ∈ D | V̇ (x) = 0} and suppose that no solution can stay identically in S,
other than the trivial solution x(t) ≡ 0. Then, the origin is an asymptotically sta-
ble equilibrium point of ẋ = f (x). Moreover, let Γ ⊂ D be a compact set that is
positively invariant with respect to ẋ = f (x), then Γ is a subset of the region of
attraction. Finally, if D = Rn and V (x) is radially unbounded, then the origin is
globally asymptotically stable. 3

Proof: The stability of the origin follows from Theorem 3.3. To show asymptotic
stability, apply Theorem 3.4 with Ω = {x ∈ Br | V (x) ≤ c} where r is chosen such
that Br ⊂ D and c < minx=r V (x). The proof of Theorem 3.3 shows that Ω is
compact and positively invariant. The largest invariant set in E = Ω ∩ S is the
origin; hence, x(t) → 0 as t → ∞. To show that Γ is a subset of the region of
attraction, apply Theorem 3.4 with Ω = Γ. Finally, if D = Rn and V (x) is radially
unbounded, then the set Ωc = {V (x) ≤ c} is compact and positively invariant for
any c > 0. By choosing c large enough we can include any given initial state x(0)
in Ωc and application of Theorem 3.4 with Ω = Ωc ensures that x(t) converges to
zero. Since this can be done for any initial state, we conclude that the origin is
globally asymptotically stable. 2

Example 3.8 Consider the system

ẋ1 = x2 , ẋ2 = −h1 (x1 ) − h2 (x2 )

where h1 and h2 are locally Lipschitz and satisfy

hi (0) = 0, yhi (y) > 0, for 0 < |y| < a

The system has an isolated equilibrium point at the origin. Depending upon the
functions h1 and h2 , it might have other equilibrium points. The system can be
viewed as a generalized pendulum with h2 (x2 ) as the friction term. Therefore, a
Lyapunov function candidate may be taken as the energy-like function
 x1
V (x) = h1 (y) dy + 12 x22
0
16 Theorem 3.5 is known as the theorem of Barbashin and Krasovskii, who proved it before the
introduction of LaSalle’s invariance principle.
58 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

Let D = {|x1 | < a, |x2 | < a}; V (x) is positive definite in D and

V̇ (x) = h1 (x1 )x2 + x2 [−h1 (x1 ) − h2 (x2 )] = −x2 h2 (x2 ) ≤ 0

is negative semidefinite. To find S = {x ∈ D | V̇ (x) = 0}, note that

V̇ (x) = 0 ⇒ x2 h2 (x2 ) = 0 ⇒ x2 = 0, since |x2 | < a

Hence, S = {x ∈ D | x2 = 0}. Let x(t) be a solution that belongs identically to S:

x2 (t) ≡ 0 ⇒ ẋ2 (t) ≡ 0 ⇒ h1 (x1 (t)) ≡ 0 ⇒ x1 (t) ≡ 0

Therefore, the only solution that can stay identically in S is the trivial solution
x(t) ≡ 0. Thus, the origin is asymptotically stable. 

Example 3.9 Consider again the system of Example 3.8, but this time let a = ∞
and assume that h1 satisfies the additional condition:
 y
h1 (z) dz → ∞ as |y| → ∞
0
x
The Lyapunov function V (x) = 0 1 h1 (y) dy + 12 x22 is radially unbounded. Similar
to the previous example, it can be shown that V̇ (x) ≤ 0 in R2 , and the set

S = {V̇ (x) = 0} = {x2 = 0}

contains no solutions other than the trivial solution. Hence, the origin is globally
asymptotically stable. 

3.5 Exponential Stability

We have seen in Theorem 3.2 that the origin of ẋ = f (x) is exponentially stable
if and only if [∂f /∂x](0) is Hurwitz. This result, however, is local. It does not
give information about the region of attraction and cannot show global exponential
stability. In this section, we use Lyapunov functions to go beyond the local result
of Theorem 3.2.

Theorem 3.6 Let f (x) be a locally Lipschitz function defined over a domain D ⊂
Rn , which contains the origin, and f (0) = 0. Let V (x) be a continuously differen-
tiable function defined over D such that

k1 xa ≤ V (x) ≤ k2 xa (3.11)

V̇ (x) ≤ −k3 xa (3.12)

for all x ∈ D, where k1 , k2 , k3 , and a are positive constants. Then, the origin is an
exponentially stable equilibrium point of ẋ = f (x). If the assumptions hold globally,
the origin will be globally exponentially stable. 3
3.5. EXPONENTIAL STABILITY 59

Proof: Choose c > 0 small enough that {k1 xa ≤ c} ⊂ D. Then,

Ωc = {V (x) ≤ c} ⊂ {k1 xa ≤ c} ⊂ D

because V (x) ≤ c ⇒ k1 xa ≤ c by (3.11). Therefore, Ωc is compact and, from

(3.12), positively invariant. All trajectories starting in Ωc satisfy

k3
V̇ ≤ − V
k2
By separation of variables and integration, we obtain

V (x(t)) ≤ V (x(0))e−(k3 /k2 )t

Hence,

1/a
1/a
V (x(t)) V (x(0))e−(k3 /k2 )t
x(t) ≤ ≤
k1 k1

1/a  1/a
k2 x(0)a e−(k3 /k2 )t k2
≤ = e−γt x(0)
k1 k1

where γ = k3 /(k2 a), which shows that the origin is exponentially stable. If all
the assumptions hold globally, c can be chosen arbitrarily large and the foregoing
inequality holds for all x(0) ∈ Rn . 2

Example 3.10 In Example 3.7 we considered the system

ẋ1 = x2 , ẋ2 = −h(x1 ) − x2

 0, and used the

where h is locally Lipschitz, h(0) = 0, and yh(y) > 0 for all y =
Lyapunov function 17
 x1

T 1 1 1
V (x) = x P x + 2 h(y) dy, with P =
0 2 1 2

whose derivative satisfies

V̇ (x) = −x1 h(x1 ) − x22
to show that the origin is globally asymptotically stable. Suppose now that h
satisfies the stronger inequalities

c1 y 2 ≤ yh(y) ≤ c2 y 2
17 We 1
take a = 1, δ = 2, and k = 2
in Example 3.7.
60 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

for all y, for some positive constants c1 and c2 . Then,

xT P x ≤ V (x) ≤ xT P x + c2 x21

and
V̇ (x) ≤ −c1 x21 − x22
Using the fact that

λmin (P )x2 ≤ xT P x ≤ λmax (P )x2

we see that inequalities (3.11) and (3.12) are satisfied globally with a = 2. Hence,
the origin is globally exponentially stable. 

For the linear system ẋ = Ax, we can apply Theorem 3.6 with a quadratic
Lyapunov function V (x) = xT P x, where P is a real symmetric positive definite
matrix. The derivative of V along the trajectories of ẋ = Ax is given by

V̇ (x) = xT P ẋ + ẋT P x = xT (P A + AT P )x = −xT Qx

where Q is a symmetric matrix defined by

P A + AT P = −Q (3.13)

If Q is positive definite, we can conclude by Theorem 3.3 or 3.6 that the origin is
globally exponentially stable; that is, A is Hurwitz. The application of Theorem 3.6
is straightforward, while that of 3.3 uses the fact that, for linear systems, asymptotic
and exponential stability are equivalent. Here we follow the usual procedure of
Lyapunov’s method, where we choose V (x) to be positive definite and then check
the negative definiteness of V̇ (x). In the case of linear systems, we can reverse
the order of these two steps. Suppose we start by choosing Q as a real symmetric
positive definite matrix, and then solve equation (3.13) for P . If the equation has
a positive definite solution, then we can again conclude that the origin is globally
exponentially stable. Equation (3.13) is called the Lyapunov equation. The next
theorem characterizes exponential stability of the origin in terms of the solution of
the Lyapunov equation.18

Theorem 3.7 A matrix A is Hurwitz if and only if for every positive definite sym-
metric matrix Q there exists a positive definite symmetric matrix P that satisfies
the Lyapunov equation P A + AT P = −Q. Moreover, if A is Hurwitz, then P is the
unique solution. 3

Theorem 3.7 is important for nonlinear systems when we use linearization. We

have seen in Section 3.2 that the nonlinear system ẋ = f (x), with a continuously
18 See [74, Theorem 4.6] for the proof of Theorem 3.7. The MATLAB command “lyap(AT , Q)”
solves equation (3.13).
3.6. REGION OF ATTRACTION 61

differentiable f (x) and f (0) = 0, can be represented in some neighborhood of the

origin by
ẋ = [A + G(x)]x
where 
∂f 
A= and G(x) → 0 as x → 0
∂x x=0
When A is Hurwitz, we can solve the Lyapunov equation P A + AT P = −Q for a
positive definite Q, and use V (x) = xT P x as a Lyapunov function candidate for
the nonlinear system. Then

V̇ (x) = −xT Qx + 2xT P G(x)x

Since G(x) → 0 as x → 0, given any positive constant k < 1, we can find r > 0
such that 2P G(x) < kλmin (Q) in the domain D = {x < r}. Thus,

V̇ (x) ≤ −(1 − k)λmin (Q)x2

in D and Theorem 3.6 shows that the origin is an exponentially stable equilibrium
point of the nonlinear system ẋ = f (x).19 The function V (x) = xT P x can be used
to estimate the region of attraction, as we shall see in the next section.

3.6 Region of Attraction

Quite often, it is not sufficient to determine that a given system has an asymp-
totically stable equilibrium point. Rather, it is important to find the region of
attraction of that point, or at least an estimate of it. In this section we shall see
some properties of the region of attraction and how to estimate it using Lyapunov’s
method.20 We start with a lemma that states some properties of the region of
attraction.21

Lemma 3.2 The region of attraction of an asymptotically stable equilibrium point

is an open, connected, invariant set, and its boundary is formed by trajectories. 3

Lemma 3.2 suggests that one way to determine the region of attraction is to
characterize the trajectories that lie on its boundary. This process can be quite
difficult for high-dimensional systems, but can be easily seen for two-dimensional
ones by examining phase portraits in the state plane. Examples 3.11 and 3.12 show
typical cases. In the first example, the boundary of the region of attraction is a limit
cycle, while in the second one it is formed of stable trajectories of saddle points.
19 This is a proof of the sufficiency part of the first item of Theorem 3.2.
20 There is a vast literature on estimating the region of attraction. Some methods are described
or surveyed in [18, 23, 48, 65, 94, 99, 140, 155].
21 See [74, Lemma 8.1] for the proof of this lemma.
62 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

4
x2

0 x1

−2

−4
−4 −2 0 2 4

Figure 3.8: Phase portrait for Example 3.11.

Example 3.11 The two-dimensional system

ẋ1 = −x2 , ẋ2 = x1 + (x21 − 1)x2

is a Van der Pol equation in reverse time, that is, with t replaced by −t. The phase
portrait of Figure 3.8 shows that origin is stable focus surrounded by an unstable
limit cycle and all trajectories in the interior of the limit cycle spiral towards the
origin. Hence the origin is asymptotically stable, its region of attraction is the
interior of the limit cycle, and the boundary of the region of attraction is the limit
cycle itself. 

Example 3.12 The two-dimensional system

ẋ1 = x2 , ẋ2 = −x1 + 13 x31 − x2

√ √
has three isolated equilibrium points at (0, 0), ( 3, 0), and (− 3, 0). The phase
portrait (Figure 3.9) shows that the origin is a stable focus, and the other two
equilibrium points are saddles. Thus, the origin is asymptotically stable. From
the phase portrait, we can also see that the stable trajectories of the saddles form
two separatrices that are the boundaries of the region of attraction. The region is
unbounded. 

Lyapunov’s method can be used to estimate the region of attraction. By an

estimate, we mean a set Γ such that every trajectory starting in Γ converges the
origin as t → ∞. For the rest of this section, we will discuss some aspects of
estimating the region of attraction. Let us start by showing that the domain D of
Theorem 3.3 or 3.5 is not an estimate of the region of attraction. We have seen
in Theorems 3.3 and 3.5 that if D is a domain that contains the origin, and if
we can find a Lyapunov function V (x) that is positive definite in D and V̇ (x) is
negative definite in D or negative semidefinite, but no solution can stay identically
3.6. REGION OF ATTRACTION 63

4 x2

0 x1

−2

−4
−4 −2 0 2 4

Figure 3.9: Phase portrait for Example 3.12.

in the set {x ∈ D | V̇ (x) = 0} except the zero solution x = 0, then the origin is
asymptotically stable. One may jump to the conclusion that D is an estimate of the
region of attraction. This conjecture is not true, as illustrated by the next example.
Example 3.13 Reconsider Example 3.12, which is a special case of Example 3.7
with h(x1 ) = x1 − 13 x31 and a = 1. A Lyapunov function is given by

 x1
1 1
1 T
V (x) = 2 x x+2 (y − 13 y 3 ) dy = 32 x21 − 16 x41 + x1 x2 + x22
1 2 0

and
V̇ (x) = −x21 (1 − 13 x21 ) − x22
√ √
Therefore, the conditions of Theorem 3.3 are satisfied in D = {− 3 < x1 < 3}.
The phase portrait in Figure 3.9 shows that D is not a subset of the region of
attraction. 
In view of this example, it is not difficult to see why D of Theorem 3.3 or 3.5
is not an estimate of the region of attraction. Even though a trajectory starting in
D will move from one Lyapunov surface V (x) = c1 to an inner Lyapunov surface
V (x) = c2 , with c2 < c1 , there is no guarantee that the trajectory will remain
forever in D. Once the trajectory leaves D, there is no guarantee that V̇ (x) will
be negative. Consequently, the whole argument about V (x) decreasing to zero
falls apart. This problem does not arise when the estimate is a compact positively
invariant subset of D, as stated in Theorem 3.5. The simplest such estimate is the
set Ωc = {V (x) ≤ c} when Ωc is bounded and contained in D.22 For a quadratic
22 We can also estimate the region of attraction by the bounded open set {V (x) < c} ⊂ D.

As we saw in Footnote 15, the set Ωc may have more than one component, but our discussions
always apply to the bounded component that contains the origin. It is worthwhile to note that
this component is connected because every trajectory in it converges to the origin.
64 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

Lyapunov function V (x) = xT P x and D = {x < r}, we can ensure that Ωc ⊂ D
by choosing
c < min xT P x = λmin (P )r2
x=r

For D = {|b x| < r}, where b ∈ Rn , equation B.3 shows that

T

r2
min xT P x =
|bT x|=r bT P −1 b
Therefore, Ωc will be a subset of D = {|bTi x| < ri , i = 1, . . . , p}, if we choose
ri2
c < min
1≤i≤p bi P −1 bi
T

The simplicity of estimating the region of attraction by Ωc = {xT P x ≤ c}

has increased significance in view of the linearization discussion at the end of the
previous section. There we saw that if the Jacobian matrix A = [∂f /∂x](0) is
Hurwitz, then we can always find a quadratic Lyapunov function V (x) = xT P x by
solving the Lyapunov equation P A + AT P = −Q for any positive definite matrix
Q. Putting the pieces together, we see that whenever A is Hurwitz, we can estimate
the region of attraction of the origin. This is illustrated by the next example.
Example 3.14 The system
ẋ1 = −x2 , ẋ2 = x1 + (x21 − 1)x2
treated in Example 3.11, has an asymptotically stable origin since


∂f  0 −1
A= =
∂x x=0 1 −1
is Hurwitz. A Lyapunov function for the system can be found by taking Q = I and
solving the Lyapunov equation P A + AT P = −I for P . The unique solution is the
positive definite matrix

1.5 −0.5
P =
−0.5 1
The quadratic function V (x) = xT P x is a Lyapunov function for the system in
a certain neighborhood of the origin. Because our interest here is in estimating
the region of attraction, we need to determine a domain D about the origin where
V̇ (x) is negative definite and a constant c > 0 such that Ωc = {V (x) ≤ c} is a
subset of D. We are interested in the largest set Ωc that we can determine, that
is, the largest value for the constant c. Notice that we do not have to worry about
checking positive definiteness of V (x) in D because V (x) is positive definite for all
x. Moreover, V (x) is radially unbounded; hence Ωc is bounded for any c > 0. The
derivative of V (x) along the trajectories of the system is given by
V̇ (x) = −(x21 + x22 ) − (x31 x2 − 2x21 x22 )
3.6. REGION OF ATTRACTION 65

2 3
x2 x2
2
1
1

0 x1 0 x1
−1
−1
−2

−2 −3
−2 −1 0 1 2 −3 −2 −1 0 1 2 3
(a) (b)

Figure 3.10: (a) Contours of V̇ (x) = 0 (dashed), V (x) = 0.618 (dash-dot), and
V (x) = 2.25 (solid) for Example 3.14; (b) comparison of the region of attraction with
its estimate.

The right-hand side of V̇ (x) is written as the sum of two terms. The first term,
−(x21 + x22 ), is the contribution of the linear part Ax, while the second term is the
contribution of the nonlinear term f (x) − Ax. We know that there is an open ball
D = {x < r} such that V̇ (x) is negative definite in D. Once we find such a ball,
we can find Ωc ⊂ D by choosing c < λmin (P )r2 . Thus, to enlarge the estimate of
the region of attraction, we need to find the largest ball on which V̇ (x) is negative
definite. We have
√
5
V̇ (x) ≤ −x + |x1 | |x1 x2 | |x1 − 2x2 | ≤ −x +
2 2
x4
2
√
where we used |x1 | ≤ x, |x1 x2 | ≤ 12 x2 , and |x1 − 2x2 | ≤ 5x. Thus, V̇ (x) is
√
negative definite on a √ ball D with r2 = 2/ 5. Using λmin (P ) = 0.691, we choose
c = 0.61 < 0.691 × 2/ 5 = 0.618. The set Ωc with c = 0.61 is an estimate of the
region of attraction. A less conservative (that is, larger) estimate can be obtained
by plotting contours of V̇ (x) = 0 and V (x) = c for increasing values of c until we
determine the largest c for which V (x) = c will be in {V̇ (x) < 0}. This is shown in
Figure 3.10(a) where c is determined to be c = 2.25. Figure 3.10(b) compares this
estimate with the region of attraction whose boundary is a limit cycle. 

Estimating the region of attraction by Ωc = {V (x) ≤ c} is simple, but usually

conservative. We conclude this section by a couple of remarks that show how we
might reduce the conservatism of the estimate.

Remark 3.1 If Ω1 , Ω2 , . . . , Ωm are positively invariant subsets of the region of

attraction, then their union ∪m i=1 Ωi is also a positively invariant subset of the region
of attraction. Therefore, if we have multiple Lyapunov functions for the same system
66 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

and each function is used to estimate the region of attraction, we can enlarge the
estimate by taking the union of all the estimates. This idea will be illustrated in
Example 7.14.

Remark 3.2 According to Theorem 3.5, we can work with any compact set Γ ⊂
D provided we can show that Γ is positively invariant. This typically requires
investigating the vector field at the boundary of Γ to ensure that trajectories starting
in Γ cannot leave it. The next example illustrates this idea.

Example 3.15 Consider the system23

ẋ1 = x2 , ẋ2 = −4(x1 + x2 ) − h(x1 + x2 )

where h is a locally Lipschitz function that satisfies

h(0) = 0; uh(u) ≥ 0, ∀ |u| ≤ 1

Consider the quadratic function

2 1
V (x) = xT x = 2x21 + 2x1 x2 + x22
1 1

as a Lyapunov function candidate.24 The derivative V̇ (x) is given by

V̇ (x) = (4x1 + 2x2 )ẋ1 + 2(x1 + x2 )ẋ2

= −2x21 − 6(x1 + x2 )2 − 2(x1 + x2 )h(x1 + x2 )
≤ −2x21 − 6(x1 + x2 )2 , ∀ |x1 + x2 | ≤ 1

8 6
= −xT x
6 6

Therefore, V̇ (x) is negative definite in the set G = {|x1 + x2 | ≤ 1} and we can con-
clude that the origin is asymptotically stable. To estimate the region of attraction,
let us start by an estimate of the form Ωc = {V (x) ≤ c}. The largest c > 0 for
which Ωc ⊂ G is given by
1
c= min xT P x = =1
|x1 +x2 |=1 bT P −1 b

where bT = [1 1]. The set Ωc with c = 1 is shown in Figure 3.11. We can obtain
a better estimate by not restricting ourselves to the form Ωc . A key point in the
development is to observe that trajectories inside G cannot leave through certain
segments of the boundary |x1 + x2 | = 1. This can be seen by examining the vector
23 The example is taken from [148].
24 This Lyapunov function candidate can be derived by using the variable gradient method or
by applying the circle criterion of Section 7.3 and the Kalman–Yakubovich–Popov lemma.
3.6. REGION OF ATTRACTION 67

5
x2
(−3, 4)

V (x) = 10
0 x1

V (x) = 1

(3, −4)
−5
−5 0 5

Figure 3.11: Estimates of the region of attraction for Example 3.15.

field at the boundary or by the following analysis: Let σ = x1 + x2 such that the
boundary of G is given by σ = 1 and σ = −1. The derivative of σ 2 along the
trajectories of the system is given by
d 2
σ = 2σ(ẋ1 + ẋ2 ) = 2σx2 − 8σ 2 − 2σh(σ) ≤ 2σx2 − 8σ 2 , ∀ |σ| ≤ 1
dt
On the boundary σ = 1,
d 2
σ ≤ 2x2 − 8 ≤ 0, ∀ x2 ≤ 4
dt
This implies that when the trajectory is at any point on the segment of the boundary
σ = 1 for which x2 ≤ 4, it cannot move outside the set G because at such point σ 2
is nonincreasing. Similarly, on the boundary σ = −1,
d 2
σ ≤ −2x2 − 8 ≤ 0, ∀ x2 ≥ −4
dt
Hence, the trajectory cannot leave the set G through the segment of the boundary
σ = −1 for which x2 ≥ −4. This information can be used to form a closed, bounded,
positively invariant set Γ that satisfies the conditions of Theorem 3.5. Using the
two segments of the boundary of G just identified to define the boundary of Γ, we
now need two other segments to close the set. These segments should have the
property that trajectories cannot leave the set through them. We can take them
as segments of a Lyapunov surface. Let c1 be such that the Lyapunov surface
V (x) = c1 intersects the boundary x1 + x2 = 1 at x2 = 4, that is, at the point
(−3, 4). (See Figure 3.11.) Let c2 be such that the Lyapunov surface V (x) = c2
intersects the boundary x1 + x2 = −1 at x2 = −4, that is, at the point (3, −4). The
required Lyapunov surface is defined by V (x) = min{c1 , c2 }. The constants c1 and
68 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

c2 are given by

c1 = V (x)|x1 =−3,x2 =4 = 10, c2 = V (x)|x1 =3,x2 =−4 = 10

Therefore, we take c = 10 and define the set Γ by

Γ = {V (x) ≤ 10 and |x1 + x2 | ≤ 1}

This set is closed, bounded, and positively invariant. Moreover, V̇ (x) is negative
definite in Γ, since Γ ⊂ G. Thus, we conclude by Theorem 3.5 that Γ is a subset of
the region of attraction. 

3.7 Converse Lyapunov Theorems

Theorems 3.3 and 3.6 establish asymptotic stability or exponential stability of the
origin by requiring the existence of a Lyapunov function V (x) that satisfies certain
conditions. Requiring the existence of an auxiliary function V (x) that satisfies cer-
tain conditions is typical in many theorems of Lyapunov’s method. The conditions
of these theorems cannot be checked directly on the data of the problem. Instead,
one has to search for the auxiliary function. Faced with this searching problem,
two questions come to mind. First, is there a function that satisfies the conditions
of the theorem? Second, how can we search for such a function? In many cases,
Lyapunov theory provides an affirmative answer to the first question. The answer
takes the form of a converse Lyapunov theorem, which is the inverse of one of Lya-
punov’s theorems. For example, a converse theorem for asymptotic stability would
confirm that if the origin is asymptotically stable, then there is a Lyapunov function
that satisfies the conditions of Theorem 3.3. Most of these converse theorems are
proven by actually constructing auxiliary functions that satisfy the conditions of
the respective theorems. Unfortunately, this construction almost always assumes
knowledge of the solution of the differential equation. Therefore, the theorems do
not help in the practical search for an auxiliary function. However, they are use-
ful in using Lyapunov theory to draw conceptual conclusions about the behavior
of dynamical systems, as in Example 3.16. In this section, we give two converse
Lyapunov theorems without proofs.25 The first one is for the case when the origin
is exponentially stable, and the second when it is asymptotically stable.

Theorem 3.8 Let x = 0 be an exponentially stable equilibrium point for the n-

dimensional system ẋ = f (x), where f is continuously differentiable on D = {x <
r}. Let k, λ, and r0 be positive constants with r0 < r/k such that

x(t) ≤ kx(0)e−λt , ∀ x(0) ∈ D0 , ∀ t ≥ 0

25 The proofs are given in [74]; see Theorems 4.14 and 4.17, respectively.
3.7. CONVERSE LYAPUNOV THEOREMS 69

where D0 = {x < r0 }. Then, there is a continuously differentiable function V (x)

that satisfies the inequalities

∂V ∂V
c1 x2 ≤ V (x) ≤ c2 x2 , f (x) ≤ −c3 x2 ,
∂x ≤ c4 x
∂x
for all x ∈ D0 , with positive constants c1 , c2 , c3 , and c4 . Moreover, if f is continu-
ously differentiable for all x, globally Lipschitz, and the origin is globally exponen-
tially stable, then V (x) is defined and satisfies the aforementioned inequalities for
all x ∈ Rn . 3

Theorem 3.9 Let x = 0 be an asymptotically stable equilibrium point for the n-

dimensional system ẋ = f (x), where f is locally Lipschitz on a domain D ⊂ Rn
that contains the origin. Let RA ⊂ D be the region of attraction of x = 0. Then,
there is a smooth, positive definite function V (x) and a continuous, positive definite
function W (x), both defined for all x ∈ RA , such that

V (x) → ∞ as x → ∂RA
∂V
f (x) ≤ −W (x), ∀ x ∈ RA
∂x
and for any c > 0, {V (x) ≤ c} is a compact subset of RA . When RA = Rn , V (x)
is radially unbounded. 3

As an example of the use of converse Lyapunov functions, we show how Theo-

rem 3.8 can be used to prove the necessity part of the first item of Theorem 3.2.

Example 3.16 Consider the system ẋ = f (x) where f is continuously differentiable

in the neighborhood of the origin and f (0) = 0. Show that the origin is exponentially
stable only if A = [∂f /∂x](0) is Hurwitz. To proceed, let us recall that

f (x) = Ax + G(x)x

where G(x) → 0 as x → 0. Hence, given any L > 0, there is r1 > 0 such that
G(x) ≤ L for all x < r1 . Rewrite the linear system ẋ = Ax as

ẋ = Ax = f (x) − G(x)x

Because the origin of ẋ = f (x) is exponentially stable, let V (x) be the function
provided by Theorem 3.8 over the domain {x < r0 }. Using V (x) as a Lyapunov
function candidate for ẋ = Ax, we obtain
∂V ∂V ∂V
Ax = f (x) − G(x)x ≤ −c3 x2 + c4 Lx2 < (c3 − c4 L)x2
∂x ∂x ∂x
Taking L < c3 /c4 ), we see that the foregoing inequality holds for all x < min{r0 , r1 }.
Hence, by Theorem 3.6, the origin of ẋ = Ax is exponentially stable.
70 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

3.8 Exercises
3.1 Show that the scalar system ẋ = (k−x2 )/x, with k > 0, has two asymptotically
stable equilibrium points and find their regions of attraction.

3.2 Consider the scalar system ẋ = −g(x), where g(x) is locally Lipschitz and

g(0) = 0; xg(x) > 0, ∀ x

= 0 and x ∈ (−a, a)

It is shown in Section 3.1 that the origin is asymptotically stable. In this exercise
we arrive at the same conclusion using Lyapunov functions.
(a) Show that the asymptotic stability conditions of Theorem 3.3 are satisfied with
x
V (x) = 12 x2 or V (x) = 0 g(y) dy.
(b) If xg(x) > 0 for all x
= 0, show that the global asymptotic stability
 x conditions
of Theorem 3.3 are satisfied with V (x) = 12 x2 or V (x) = 12 x2 + 0 g(y) dy.
(c) Under what x conditions on g can you show global asymptotic stability using
V (x) = 0 g(y) dy? Give an example of g where the origin of ẋ = −g(x)
x
is globally asymptotic stability but V (x) = 0 g(y) dy does not satisfy the
global asymptotic stability conditions of Theorem 3.3.

3.3 For each of the following scalar systems, determine if the origin is unstable,
stable but not asymptotically stable, asymptotically stable but not globally asymp-
totically stable, or globally asymptotically stable.
(1) f (x) = −x for x ≥ 0 and f (x) = 0 for x < 0
(2) f (x) = x for x ≥ 0 and f (x) = 0 for x < 0
(3) f (x) = sin x
(4) f (x) = − sin x
(5) f (x) = −x − sin x

3.4 Euler equations for a rotating rigid spacecraft are given by [105]

J1 ω̇1 = (J2 −J3 )ω2 ω3 +u1 , J2 ω̇2 = (J3 −J1 )ω3 ω1 +u2 , J3 ω̇3 = (J1 −J2 )ω1 ω2 +u3

where ω = col(ω1 , ω2 , ω3 ) is the angular velocity vector along the principal axes,
u = col(u1 , u2 , u3 ) is the vector of torque inputs applied about the principal axes,
and J1 to J3 are the principal moments of inertia.
(a) Show that with u = 0 the origin ω = 0 is stable. Is it asymptotically stable?
(b) Let ui = −ki ωi , where k1 to k3 are positive constants. Show that the origin of
the closed-loop system is globally asymptotically stable.
3.8. EXERCISES 71

3.5 For each of the following systems, determine whether the origin is stable,
asymptotically stable or unstable.
(1) ẋ1 = x2 , ẋ2 = x3 , ẋ3 = −2 sin x1 − 2x2 − 2x3
(2) ẋ1 = x2 , ẋ2 = x3 , ẋ3 = 2 sin x1 − 2x2 − 2x3
(3) ẋ1 = x2 + x3 , ẋ2 = − sin x1 − x3 , ẋ3 = − sin x1 + x2

3.6 Consider the system

 
x21
ẋ1 = x2 , ẋ2 = −h(x1 ) − 2x2 , where h(x1 ) = x1 2+
1 + x21
x
Verify that V (x) = 0 1 h(σ) dσ + 12 (x1 + x2 )2 is positive definite and radially
unbounded; then use it to show that the origin is globally exponentially stable.

3.7 Consider the system

ẋ1 = −x1 + x3 , ẋ2 = x3 , ẋ3 = −x2 − x32 − x3

(a) Is the origin exponentially stable?

(b) Using V (x) = bx21 + 3x22 + 2x2 x3 + 2x23 + x42 , where b > 0 is to be chosen, show
that the origin is globally asymptotically stable.

3.8 ([132]) Consider the system

ẋ1 = x2 , ẋ2 = − tanh(x1 + x2 )

(a) Show that

 x1  x1 +x2
V (x) = tanh(σ) dσ + tanh(σ) dσ + x22
0 0

is positive definite for all x and radially unbounded.

(b) Show that the origin is globally asymptotically stable.

3.9 (Krasovskii’s Method) Consider the system ẋ = f (x) with f (0) = 0. As-
sume that f (x) is continuously differentiable and its Jacobian [∂f /∂x] satisfies


T
∂f ∂f
P (x) + (x) P ≤ −I, ∀ x ∈ Rn , where P = P T > 0
∂x ∂x
1 ∂f
(a) Using the representation f (x) = 0 ∂x
(σx)x dσ, show that

xT P f (x) + f T (x)P x ≤ −xT x, ∀ x ∈ Rn

72 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

(b) Using V (x) = f T (x)P f (x), show that the origin is globally asymptotically
stable.

3.10 ([121]) A closed-loop system under optimal stabilizing control is given by

 T
−1 ∂V
ẋ = f (x) − kG(x)R T
(x)G (x)
∂x

where V (x) is a continuously differentiable, positive definite function that satisfies

the Hamilton–Jacobi–Bellman equation
 T
∂V ∂V ∂V
f (x) + q(x) − 14 G(x)R−1 (x)GT (x) =0
∂x ∂x ∂x

q(x) is a positive semidefinite function, R(x) is a positive definite matrix for all x,
and k is a positive constant. Show that the origin is asymptotically stable when
(1) q(x) is positive definite and k ≥ 1/4.
(2) q(x) is positive semidefinite, k > 1/4, and the only solution of ẋ = f (x) that
stays identically in the set {q(x) = 0} is the trivial solution x(t) ≡ 0.

3.11 The system

ẋ1 = −x1 + x3 , ẋ2 = x3 , ẋ3 = −g1 (x1 ) − g2 (x2 ) − x3

where g1 and g2 are locally Lipschitz and satisfy gi (0) = 0 and ygi (y) > 0 for all
y
= 0, has a unique equilibrium point at the origin.
x x
(a) Verify that V (x) = 0 1 g1 (y) dy + 0 2 g2 (y) dy + 12 x23 is positive definite for all
x and use it to show asymptotic stability of the origin.
(b) Under what additional conditions on g1 and g2 can you show that the origin is
globally asymptotically stable.
(c) Under what additional conditions on g1 and g2 can you show that the origin is
exponentially stable?

3.12 An unforced mass-spring system (Section A.2) with nonlinear viscous friction
and nonlinear spring is modeled by

mÿ + b(1 + c|ẏ|)ẏ + g(y) = 0

where g(y) = k(1−a2 y 2 )y, with |ay| < 1, for a softening spring, g(y) = k(1+a2 y 2 )y
for a hardening spring, and all constants are positive. Take the state variables as
x1 = y, x2 = ẏ. Using an energy-type Lyapunov function, study the stability of the
origin for each spring type.
3.8. EXERCISES 73

3.13 Show that the origin of the system

ẋ1 = x2 , ẋ2 = −x1 + μx2 − x32

is globally asymptotically stable for μ ≤ 0 and unstable for μ > 0. Is it exponentially

stable when μ = 0 ?

3.14 Using V (x) = 12 xT x, show that the origin of

ẋ1 = −x1 + x2 cos x1 , ẋ2 = x1 cos x1 − x2 (cos x1 )2 + x3 , ẋ3 = −x2

is globally asymptotically stable. Is it exponentially stable?

3.15 Consider the pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + u

under the state feedback control u = −k1 sat(x1 ) − k2 sat(x2 ), where b, k1 and k2
are positive constants and k1 > 1.
(a) Show that the origin is the unique equilibrium point of the closed-loop system.
x
(b) Using V (x) = 0 1 [sin σ + k1 sat(σ)] dσ + 12 x22 , show that the origin of the
closed-loop system is globally asymptotically stable.

3.16 For each of the following systems,

(a) Find all equilibrium points and study their stability using linearization.
(b) Using quadratic Lyapunov functions, estimate the regions of the attraction of
each asymptotically stable equilibrium point. Try to make the estimates as
large as you can.
(c) Draw the phase portrait of the system to find the exact regions of attraction
and compare them with your estimates.

(1) ẋ1 = −(x1 + x31 ) + 2x2 , ẋ2 = 2x1 − (x2 + x32 )

(2) ẋ1 = x1 − x31 + x2 , ẋ2 = x1 − 3x2
(3) ẋ1 = x2 , ẋ2 = −x1 − x2 + 2x32 − 12 x52
(4) ẋ1 = −x2 , ẋ2 = 2x1 + 3x2 + 2 sat(−3x2 )

3.17 Consider the system

ẋ1 = x2 , ẋ2 = −x1 + x51 − x2

(a) Using the Lyapunov function candidate V (x) = 12 x21 − 16 x61 + 12 x22 , show that
the origin is asymptotically stable and estimate the region of attraction.
74 CHAPTER 3. STABILITY OF EQUILIBRIUM POINTS

(b) Draw the phase portrait of the system to find the exact region of attraction
and compare it with your estimate.

3.18 Consider the system

ẋ1 = x31 − x2 , ẋ2 = x1 − x2

Show that the origin is asymptotically stable? Is it exponentially stable? Is it

globally asymptotically stable? If not, estimate the region of attraction.

3.19 For each of the following systems, show that there is an equilibrium point
at the origin and investigate its stability. If the origin is asymptotically stable,
determine if it is globally asymptotically stable; if it is not so, estimate the region
of attraction.

(1) ẋ1 = x2 , ẋ2 = −x31 − x2

(2) ẋ1 = x2 , ẋ2 = −x31

(3) ẋ1 = x2 , ẋ2 = −x1 − 2x2 − (x21 + x22 ) tanh(x1 + x2 )

(4) ẋ1 = x2 , ẋ2 = −x1 + 13 x31 − 2x2

(5) ẋ1 = −x31 + x2 , ẋ2 = −x31 − x2

3.20 Show that the origin of the system

ẋ1 = −x1 + x2 , ẋ2 = −0.1x31 − x2 − 10 sin x3 , ẋ3 = x2 − x3

is asymptotically stable and estimate the region of attraction.

3.21 ([148]) Consider the system

ẋ1 = x2 , ẋ2 = −x1 − x2 − (x1 + 2x2 )(1 − x22 )

Show that the origin is asymptotically stable and find an estimate of the region
of attraction that includes the point x = (−1, 1). Hint: Use the variable gradient
method to find a quadratic Lyapunov function V (x) = xT P x such that V̇ (x) is
negative definite in the set {|x2 | ≤ 1}.
Chapter 4

Time-Varying and Perturbed

Systems

The chapter starts in Section 4.1 by extending Lyapunov stability theory to time-
varying systems. Towards that end, we introduce class K and class KL functions,
which will be used extensively in the rest of the chapter, and indeed the rest of the
book. We use them to define uniform stability and uniform asymptotic stability of
equilibrium points of time-varying systems, and give Lyapunov theorems for these
properties. The chapter turns then to the study of perturbed systems of the form

ẋ = f (x) + g(t, x)

when the origin of ẋ = f (x) is asymptotically stable. Two cases are studied sepa-
rately. Vanishing perturbations, when g(t, 0) = 0, are studied in Section 4.2 where
the goal is to find conditions on g under which we can confirm that the origin of
the perturbed system is uniformly asymptotically, or exponentially, stable. Nonva-
nishing perturbations, when g(t, 0)
= 0, are studied in Section 4.3 after introducing
the notion of ultimate boundedness and giving Lyapunov-like theorems for it.1 The
results on ultimate boundedness lead naturally to the notion of input-to-state sta-
bility, which is introduced in Section 4.4 together with its Lyapunov-like theorems.

4.1 Time-Varying Systems

Consider the n-dimensional time-varying system

ẋ = f (t, x) (4.1)
1 The nonvanishing perturbation is referred to in [54, 78] as “persistent disturbance.” The
results on nonvanishing perturbations are related to the concept of total stability [54, Section 56].

75
76 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

where f (t, x) is piecewise continuous in t and locally Lipschitz in x for all t ≥ 0

and x ∈ D, where D ⊂ Rn is a domain that contains the origin x = 0. The origin
is an equilibrium point of (4.1) at t = 0 if f (t, 0) = 0, ∀ t ≥ 0. The notions of
stability and asymptotic stability of equilibrium points of time-varying systems are
basically the same as those introduced in Definition 3.1 for time-invariant systems.
The new element here is that, while the solution of a time-invariant system depends
only on (t − t0 ), the solution of a time-varying system may depend on both t and
t0 . Therefore, the stability behavior of the equilibrium point will, in general, be
dependent on t0 . The origin x = 0 is a stable equilibrium point for ẋ = f (t, x) if,
for each ε > 0, and any t0 ≥ 0 there is δ > 0, dependent on both ε and t0 , such that

x(t0 ) < δ ⇒ x(t) < ε, ∀ t ≥ t0

The existence of δ for every t0 does not necessarily guarantee that there is one δ,
dependent only on ε, that works for all t0 . Similar dependence on t0 appears in
studying asymptotic stability. Therefore, we refine Definition 3.1 to define stability
and asymptotic stability as uniform properties with respect to the initial time. The
new definition will use class K and class KL functions, which are defined next.

Definition 4.1
• A scalar continuous function α(r), defined for r ∈ [0, a), belongs to class K if
it is strictly increasing and α(0) = 0. It belongs to class K∞ if it is defined
for all r ≥ 0 and α(r) → ∞ as r → ∞.
• A scalar continuous function β(r, s), defined for r ∈ [0, a) and s ∈ [0, ∞),
belongs to class KL if, for each fixed s, the mapping β(r, s) belongs to class K
with respect to r and, for each fixed r, the mapping β(r, s) is decreasing with
respect to s and β(r, s) → 0 as s → ∞.

Example 4.1
• α(r) = tan−1 (r) is strictly increasing since α (r) = 1/(1 + r2 ) > 0. It belongs
to class K, but not to class K∞ since limr→∞ α(r) = π/2 < ∞.
• α(r) = rc , with c > 0, is strictly increasing since α (r) = c rc−1 > 0. More-
over, limr→∞ α(r) = ∞; thus, it belongs to class K∞ .
• α(r) = min{r, r2 } is continuous, strictly increasing, and limr→∞ α(r) = ∞.
Hence, it belongs to class K∞ . It is not continuously differentiable at r = 1.
Continuous differentiability is not required for a class K function.
• β(r, s) = r/(ksr + 1), with positive constant k, belongs to class KL because it
is strictly increasing in r since ∂β/∂r = 1/(ksr + 1)2 > 0, strictly decreasing
in s since ∂β/∂s = −kr2 /(ksr + 1)2 < 0, and β(r, s) → 0 as s → ∞.
• β(r, s) = rc e−as , with positive constants a and c, belongs to class KL.
4.1. TIME-VARYING SYSTEMS 77

The next two lemmas state some properties of class K and class KL functions.2
Lemma 4.1 Let α1 and α2 be class K functions on [0, a1 ) and [0, a2 ), respectively,
with a1 ≥ limr→a2 α2 (r), and β be a class KL function defined on [0, limr→a2 α2 (r))×
[0, ∞) with a1 ≥ limr→a2 β(α2 (r), 0). Let α3 and α4 be class K∞ functions. Denote
the inverse of αi by αi−1 . Then,
• α1−1 is defined on [0, limr→a1 α1 (r)) and belongs to class K.
• α3−1 is defined on [0, ∞) and belongs to class K∞ .
• α1 ◦ α2 is defined on [0, a2 ) and belongs to class K.
• α3 ◦ α4 is defined on [0, ∞) and belongs to class K∞ .
• σ(r, s) = α1 (β(α2 (r), s)) is defined on [0, a2 ) × [0, ∞) and belongs to class KL.
3

Lemma 4.2 Let V : D → R be a continuous positive definite function defined on a

domain D ⊂ Rn that contains the origin. Let Br ⊂ D for some r > 0. Then, there
exist class K functions α1 and α2 , defined on [0, r], such that

α1 (x) ≤ V (x) ≤ α2 (x)

for all x ∈ Br . If D = Rn and V (x) is radially unbounded, then there exist class
K∞ functions α1 and α2 such the foregoing inequality holds for all x ∈ Rn . 3

Definition 4.2 The equilibrium point x = 0 of ẋ = f (t, x) is

• uniformly stable if there exist a class K function α and a positive constant c,
independent of t0 , such that

x(t) ≤ α(x(t0 )), ∀ t ≥ t0 ≥ 0, ∀ x(t0 ) < c

• uniformly asymptotically stable if there exist a class KL function β and a

positive constant c, independent of t0 , such that

x(t) ≤ β(x(t0 ), t − t0 ), ∀ t ≥ t0 ≥ 0, ∀ x(t0 ) < c

• globally uniformly asymptotically stable if the foregoing inequality holds ∀ x(t0 ).

• exponentially stable if there exist positive constants c, k, and λ such that

x(t) ≤ kx(t0 )e−λ(t−t0 ) , ∀ x(t0 ) < c

• globally exponentially stable if the foregoing inequality holds ∀ x(t0 ). 3

2 The proof of Lemma 4.1 is straightforward and that of 4.2 is given in [74, Lemma 4.3].
78 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

The following three theorems, stated without proofs,3 extend Theorems 3.3 and
3.6 to time-varying systems.

Theorem 4.1 Let the origin x = 0 be an equilibrium point of ẋ = f (t, x) and

D ⊂ Rn be a domain containing x = 0. Suppose f (t, x) is piecewise continuous in
t and locally Lipschitz in x for all t ≥ 0 and x ∈ D. Let V (t, x) be a continuously
differentiable function such that

W1 (x) ≤ V (t, x) ≤ W2 (x) (4.2)

∂V ∂V
+ f (t, x) ≤ 0 (4.3)
∂t ∂x
for all t ≥ 0 and x ∈ D, where W1 (x) and W2 (x) are continuous positive definite
functions on D. Then, the origin is uniformly stable. 3

Theorem 4.2 Suppose the assumptions of Theorem 4.1 are satisfied with inequality
(4.3) strengthened to
∂V ∂V
+ f (t, x) ≤ −W3 (x) (4.4)
∂t ∂x
for all t ≥ 0 and x ∈ D, where W3 (x) is a continuous positive definite function
on D. Then, the origin is uniformly asymptotically stable. Moreover, if r and c
are chosen such that Br = {x ≤ r} ⊂ D and c < minx=r W1 (x), then every
trajectory starting in {x ∈ Br | W2 (x) ≤ c} satisfies

x(t) ≤ β(x(t0 ), t − t0 ), ∀ t ≥ t0 ≥ 0

for some class KL function β. Finally, if D = Rn and W1 (x) is radially unbounded,

then the origin is globally uniformly asymptotically stable. 3

Theorem 4.3 Suppose the assumptions of Theorem 4.1 are satisfied with inequal-
ities (4.2) and (4.3) strengthened to

k1 xa ≤ V (t, x) ≤ k2 xa (4.5)

∂V ∂V
+ f (t, x) ≤ −k3 xa (4.6)
∂t ∂x
for all t ≥ 0 and x ∈ D, where k1 , k2 , k3 , and a are positive constants. Then, the
origin is exponentially stable. If the assumptions hold globally, the origin will be
globally exponentially stable. 3

A function V (t, x) is said to be positive semidefinite if V (t, x) ≥ 0. It is said

to be positive definite if V (t, x) ≥ W1 (x) for some positive definite function W1 (x),
radially unbounded if W1 (x) is so, and decrescent if V (t, x) ≤ W2 (x). A function
V (t, x) is said to be negative definite (semidefinite) if −V (t, x) is positive definite
3 The proofs can be found in [74, Section 4.5].
4.1. TIME-VARYING SYSTEMS 79

(semidefinite). Therefore, Theorems 4.1 and 4.2 say that the origin is uniformly
stable if there is a continuously differentiable, positive definite, decrescent function
V (t, x), whose derivative along the trajectories of the system is negative semidefinite.
It is uniformly asymptotically stable if the derivative is negative definite, and globally
uniformly asymptotically stable if the conditions for uniform asymptotic stability
hold globally with a radially unbounded V (t, x).
Example 4.2 Consider the scalar system
ẋ = −[1 + g(t)]x3
where g(t) is continuous and g(t) ≥ 0 for all t ≥ 0. Using V (x) = 12 x2 , we obtain

V̇ (t, x) = −[1 + g(t)]x4 ≤ −x4 , ∀ x ∈ R, ∀ t ≥ 0

The assumptions of Theorem 4.2 are satisfied globally with W1 (x) = W2 (x) = V (x)
and W3 (x) = x4 . Hence, the origin is globally uniformly asymptotically stable.
Example 4.3 Consider the system
ẋ1 = −x1 − g(t)x2 , ẋ2 = x1 − x2
where g(t) is continuously differentiable and satisfies
0 ≤ g(t) ≤ k and ġ(t) ≤ g(t), ∀ t ≥ 0
Taking V (t, x) = x21 + [1 + g(t)]x22 , it can be seen that
x21 + x22 ≤ V (t, x) ≤ x21 + (1 + k)x22 , ∀ x ∈ R2
Hence, V (t, x) is positive definite, decrescent, and radially unbounded. The deriva-
tive of V along the trajectories of the system is given by
V̇ (t, x) = −2x21 + 2x1 x2 − [2 + 2g(t) − ġ(t)]x22
Using the inequality
2 + 2g(t) − ġ(t) ≥ 2 + 2g(t) − g(t) ≥ 2
we obtain

T

x1 2 −1 x1 def
V̇ (t, x) ≤ −2x21 + 2x1 x2 − 2x22 = − = −xT Qx
x2 −1 2 x2

where Q is positive definite; therefore, V̇ (t, x) is negative definite. Thus, all the
assumptions of Theorem 4.2 are satisfied globally with positive definite quadratic
functions W1 , W2 , and W3 . Recalling that a positive definite xT P x satisfies
λmin (P )x2 ≤ xT P x ≤ λmax (P )x2
we see that the conditions of Theorem 4.3 are satisfied globally with a = 2. Hence,
the origin is globally exponentially stable.
80 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

4.2 Perturbed Systems

Consider the system
ẋ = f (x) + g(t, x) (4.7)
where f is locally Lipschitz and g is piecewise continuous in t and locally Lipschitz
in x, for all t ≥ 0 and x ∈ D, in which D ⊂ Rn is a domain that contains the origin
x = 0. Suppose f (0) = 0 and g(t, 0) = 0 so that the origin is an equilibrium point
of the system (4.7). We think of the system (4.7) as a perturbation of the nominal
system4
ẋ = f (x) (4.8)
The perturbation term g(t, x) could result from modeling errors, aging, or uncertain-
ties and disturbances, which exist in any realistic problem. In a typical situation,
we do not know g(t, x), but we know some information about it, like knowing an
upper bound on g(t, x). Suppose the nominal system (4.8) has an asymptotically
stable equilibrium point at the origin, what can we say about the stability of the
origin as an equilibrium point of the perturbed system (4.7)? A natural approach
to address this question is to use a Lyapunov function for the nominal system as a
Lyapunov function candidate for the perturbed system.
Let us start with the case when the origin is an exponentially stable equilibrium
point of the nominal system (4.8). Let V (x) be a Lyapunov function that satisfies

∂V ∂V
c1 x ≤ V (x) ≤ c2 x ,
2 2
f (x) ≤ −c3 x ,
2 ≤ c4 x (4.9)
∂x ∂x

for all x ∈ D for some positive constants c1 , c2 , c3 , and c4 .5 Suppose the perturba-
tion term g(t, x) satisfies the linear growth bound

g(t, x) ≤ γx, ∀ t ≥ 0, ∀ x ∈ D (4.10)

where γ is a nonnegative constant. Any function g(t, x) that vanishes at the origin
and is locally Lipschitz in x, uniformly in t for all t ≥ 0, in a bounded neighborhood
of the origin satisfies (4.10) over that neighborhood.6 We use V as a Lyapunov
function candidate to investigate the stability of the origin as an equilibrium point
for the perturbed system (4.7). The derivative of V along the trajectories of (4.7)
is given by
∂V ∂V
V̇ (t, x) = f (x) + g(t, x)
∂x ∂x
4 For convenience, the nominal system is assumed to be time invariant. The results of this

section extend in a straightforward way to the case when the nominal system is time varying; see
[74, Section 9.1].
5 The existence of a Lyapunov function satisfying (4.9) is guaranteed by (the converse Lyapunov)

Theorem 3.8 when f is continuously differentiable.

6 Note, however, that the linear growth bound (4.10) becomes restrictive when required to hold

globally because it would require g to be globally Lipschitz in x.

4.2. PERTURBED SYSTEMS 81

The first term on the right-hand side is the derivative of V (x) along the trajectories
of the nominal system, which is negative definite. The second term, [∂V /∂x]g, is
the effect of the perturbation. Since we do not have complete knowledge of g the
best we can do is worst case analysis where [∂V /∂x]g is bounded by a nonnegative
term. Using (4.9) and (4.10), we obtain

∂V

V̇ (t, x) ≤ −c3 x +
2 g(t, x) ≤ −c3 x2 + c4 γx2
∂x

If γ is small enough to satisfy the bound

c3
γ< (4.11)
c4

then
V̇ (t, x) ≤ −(c3 − γc4 )x2 , (c3 − γc4 ) > 0
In this case, we can conclude by Theorem 4.3 that the origin is an exponentially
stable equilibrium point of the perturbed system (4.7).

Example 4.4 Consider the system

ẋ = Ax + g(t, x)

where A is Hurwitz and g(t, x) ≤ γx for all t ≥ 0 and x ∈ Rn . Let Q = QT > 0
and solve the Lyapunov equation P A + AT P = −Q for P . From Theorem 3.7,
we know that there is a unique solution P = P T > 0. The quadratic Lyapunov
function V (x) = xT P x satisfies (4.9). In particular,

λmin (P )x2 ≤ V (x) ≤ λmax (P )x2

∂V
Ax = −xT Qx ≤ −λmin (Q)x2
∂x

∂V

∂x = 2x P  ≤ 2P x = 2λmax (P )x
T

The derivative of V (x) along the trajectories of the perturbed system satisfies

V̇ (t, x) ≤ −λmin (Q)x2 + 2λmax (P )γx2

Hence, the origin is globally exponentially stable if γ < λmin (Q)/(2λmax (P )). Since
this bound depends on the choice of Q, one may wonder how to choose Q to maxi-
mize the ratio λmin (Q)/λmax (P ). It turns out that this ratio is maximized with the
choice Q = I.7
7 See [74, Exercise 9.1].
82 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

Example 4.5 Consider the two-dimensional system

ẋ1 = x2 , ẋ2 = −4x1 − 2x2 + βx32

where the constant β ≥ 0 is unknown. We view the system as a perturbed system

of the form (4.7) with

0 1 x1 0
f (x) = Ax = and g(x) =
−4 −2 x2 βx32
√
The eigenvalues of A are −1 ± j 3. Hence, A is Hurwitz. The solution of the
Lyapunov equation P A + AT P = −I is
⎡ 3 1 ⎤
2 8
P =⎣ ⎦
1 5
8 16

As we saw in Example 4.4, the Lyapunov function V (x) = xT P x satisfies (4.9)

with c3 = 1 and c4 = 2λmax (P ) = 2 × 1.513 = 3.026. The perturbation term g(x)
does not satisfy the linear growth condition g(x) ≤ γx globally, but satisfies it
over compact sets. Consider the compact set Ωc = {xT P x ≤ c} for c > 0 and let
k2 = maxx∈Ωc |x2 |. Then, by (B.4),
  √   √
k2 = max | 0 1 x| = c 0 1 P −1/2  = 1.8194 c
xT P x≤c

where P −1/2 is the inverse of P 1/2 , the square root matrix of P . Using V (x) as a
Lyapunov function candidate for the perturbed system, we obtain

V̇ (x) ≤ −x2 + 3.026βk22 x2

Hence, V̇ (x) will be negative definite in Ωc if

1 1 0.1
β< = ≈
2
3.026k2 3.026 × (1.8194) c
2 c

We conclude that if β < 0.1/c, the origin will be exponentially stable and Ωc will
be an estimate of the region of attraction. Let us use this example to illustrate the
conservative nature of the bound (4.11). Using this bound, we came up with the
inequality β < 1/(3.026k22 ), which allows the perturbation term g(t, x) to be any
two-dimensional vector that satisfies g(t, x) ≤ βk22 x. This class of perturbations
is more general than the perturbation we have in this specific problem. We have
a structured perturbation in the sense that the first component of g is always zero,
while our analysis allowed for unstructured perturbation where the vector g could
change in all directions. Such disregard of the structure of the perturbation will,
in general, lead to conservative bounds. Suppose we repeat the analysis, this time
taking into consideration the structure of the perturbation. Instead of using the
4.2. PERTURBED SYSTEMS 83

general bound of (4.11), we calculate the derivative of V (x) along the trajectories
of the perturbed system to obtain
V̇ (x) = −x2 + 2xT P g(x) = −x2 + 18 βx32 (2x1 + 5x2 )
√
≤ −x2 + 8 βk2 x
29 2 2

where we wrote (2x1 + 5x2 ) as y T x and√used the inequality |y T x| ≤ xy. Hence,

V̇ (x) is negative definite for β < 8/( 29k22 ). Using, again, the fact that for all
x ∈ Ωc , |x2 |2 ≤ k22 = (1.8194)2 c, we arrive at the bound β < 0.448/c, which is
more than four times the bound we obtained by using (4.11). Finally, note that
the inequality β < 0.448/c shows a tradeoff between the estimate of the region
of attraction and the estimate of the upper bound on β. The smaller the upper
bound on β, the larger the estimate of the region of attraction. This tradeoff is not
artificial; it does exist in this example. The change of variables
 
3β 3β
z1 = x2 , z2 = (4x1 + 2x2 − βx32 ), τ = 2t
2 8
transforms the state equation into
dz1 dz2
= −z2 , = z1 + (z12 − 1)z2
dτ dτ
which was shown in Example 3.11 to have a bounded region of attraction surrounded
by an unstable limit cycle. When transformed into the x-coordinates, the region of
attraction will expand with decreasing β and shrink with increasing β.
When the origin of the nominal system ẋ = f (x) is asymptotically, but not
exponentially, stable, the stability analysis of the perturbed system is more involved.
Suppose the nominal system has a positive definite Lyapunov function V (x) that
satisfies
∂V
f (x) ≤ −W3 (x)
∂x
for all x ∈ D, where W3 (x) is positive definite and continuous. The derivative of V
along the trajectories of (4.7) is given by

∂V ∂V ∂V
V̇ (t, x) = f (x) +
g(t, x) ≤ −W3 (x) + g(t, x)
∂x ∂x ∂x

Our task now is to show that [∂V /∂x]g(t, x) < W3 (x) for all t ≥ 0 and x ∈ D,
a task that cannot be done by putting a linear growth bound on g(t, x), as we
have done in the exponential stability case. The growth bound on g(t, x) will
depend on the nature of the Lyapunov function of the nominal system. One class
of Lyapunov functions for which the analysis is almost as simple as in exponential
stability is the case when V (x) satisfies

∂V ∂V
f (x) ≤ −c3 φ2 (x),
∂x ∂x ≤ c4 φ(x) (4.12)
84 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

for all x ∈ D for some positive constants c3 and c4 , where the scalar function
φ(x) is positive definite and continuous. A function satisfying (4.12) is called a
quadratic-type Lyapunov function. It is clear that a Lyapunov function satisfying
(4.9) is quadratic type, but a quadratic-type Lyapunov function may exist even
when the origin is not exponentially stable. We will illustrate this point shortly
by an example. If the nominal system ẋ = f (x) has a quadratic-type Lyapunov
function V (x), then its derivative along the trajectories of (4.7) satisfies

V̇ (t, x) ≤ −c3 φ2 (x) + c4 φ(x)g(t, x)

Suppose now that the perturbation term satisfies the bound

c3
g(t, x) ≤ γφ(x), with γ <
c4

Then,
V̇ (t, x) ≤ −(c3 − c4 γ)φ2 (x)
which shows that V̇ (t, x) is negative definite.

Example 4.6 Consider the scalar system

ẋ = −x3 + g(t, x)

The nominal system ẋ = −x3 has a globally asymptotically stable equilibrium

point at the origin, but, as we saw in Example 3.2, the origin is not exponentially
stable. Thus, there is no Lyapunov function that satisfies (4.9). The Lyapunov
function V (x) = x4 satisfies (4.12), with φ(x) = |x|3 and c3 = c4 = 4. Suppose the
perturbation term g(t, x) satisfies the bound |g(t, x)| ≤ γ|x|3 for all x, with γ < 1.
Then, the derivative of V along the trajectories of the perturbed system satisfies

V̇ (t, x) ≤ −4(1 − γ)φ2 (x)

Hence, the origin is a globally uniformly asymptotically stable equilibrium point of

the perturbed system.

In contrast to the case of exponential stability, it is important to notice that

a nominal system with asymptotically, but not exponentially, stable origin is not
robust to smooth perturbations with arbitrarily small linear growth bounds of the
form of (4.10). This point is illustrated by the next example.

Example 4.7 Consider the system of the previous example with perturbation g =
γx where γ > 0; that is,
ẋ = −x3 + γx
It can be seen, via linearization, that for any γ > 0 the origin is unstable.
4.3. BOUNDEDNESS AND ULTIMATE BOUNDEDNESS 85

4.3 Boundedness and Ultimate Boundedness

The scalar equation
ẋ = −x + δ sin t, x(t0 ) = a, a>δ>0
has no equilibrium points and its solution is given by
 t
x(t) = e−(t−t0 ) a + δ e−(t−τ ) sin τ dτ
t0

For all t ≥ t0 , the solution satisfies the bound

 t  
|x(t)| ≤ e−(t−t0 ) a + δ e−(t−τ ) dτ = e−(t−t0 ) a + δ 1 − e−(t−t0 ) ≤ a
t0

Hence, the solution is bounded uniformly in t0 , that is, with a bound independent
of t0 . While this bound is valid for all t ≥ t0 , it becomes a conservative estimate
of the solution as time progresses because it does not take into consideration the
exponentially decaying term. If, on the other hand, we pick any number b such that
δ < b < a, it can be seen that
 
a−δ
|x(t)| ≤ b, ∀ t ≥ t0 + ln
b−δ
The bound b, which again is independent of t0 , gives a better estimate of the solution
after a transient period has passed. In this case, the solution is said to be uniformly
ultimately bounded and b is called an ultimate bound.
The following definition formalizes the notions of boundedness and ultimate
boundedness of the solutions of the n-dimensional system
ẋ = f (t, x), x(t0 ) = x0 (4.13)
in which t0 ≥ 0, f is piecewise continuous in t and locally Lipschitz in x for all t ≥ 0
and x ∈ D, where D ⊂ Rn is a domain that contains the origin.
Definition 4.3 The solutions of (4.13) are
• uniformly bounded if there exists c > 0, independent of t0 , and for every
a ∈ (0, c), there is β > 0, dependent on a but independent of t0 , such that
x(t0 ) ≤ a ⇒ x(t) ≤ β, ∀ t ≥ t0 (4.14)

• globally uniformly bounded if (4.14) holds for arbitrarily large a.

• uniformly ultimately bounded with ultimate bound b if there exists a positive
constant c, independent of t0 , and for every a ∈ (0, c), there is T ≥ 0, depen-
dent on a and b but independent of t0 , such that
x(t0 ) ≤ a ⇒ x(t) ≤ b, ∀ t ≥ t0 + T (4.15)
86 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

• globally uniformly ultimately bounded if (4.15) holds for arbitrarily large a.

In the case of time-invariant systems, we may drop the word “uniformly” since the
solution depends only on t − t0 .
Lyapunov analysis can be used to study boundedness and ultimate boundedness
of the solutions of ẋ = f (t, x). Consider a continuously differentiable, positive
definite function V (x) and suppose that the set {V (x) ≤ c} is compact, for some
c > 0. Let Λ = {ε ≤ V (x) ≤ c} for some positive constant ε < c. Suppose the
derivative of V along the trajectories of the system ẋ = f (t, x) satisfies

V̇ (t, x) ≤ −W3 (x), ∀ x ∈ Λ, ∀ t ≥ 0 (4.16)

where W3 (x) is a continuous positive definite function. Inequality (4.16) implies

that the sets Ωc = {V (x) ≤ c} and Ωε = {V (x) ≤ ε} are positively invariant since
on the boundaries ∂Ωc and ∂Ωε , the derivative V̇ is negative. A sketch of the sets
Λ, Ωc , and Ωε is shown in Figure 4.1. Since V̇ is negative in Λ, a trajectory starting
in Λ must move in a direction of decreasing V (x(t)). In fact, while in Λ, V satisfies
inequalities (4.2) and (4.4) of Theorem 4.2. Therefore, the trajectory behaves as if
the origin was uniformly asymptotically stable and satisfies

x(t) ≤ β(x(t0 ), t − t0 )

for some class KL function β. The function V (x(t)) will continue decreasing until
the trajectory enters the set Ωε in finite time and stays therein for all future time.
The fact that the trajectory enters Ωε in finite time can be shown as follows: Let
k = minx∈Λ W3 (x) > 0. The minimum exists because W3 (x) is continuous and Λ is
compact. It is positive since W3 (x) is positive definite. Hence,

V̇ (t, x) ≤ −k, ∀ x ∈ Λ, ∀ t ≥ t0

Therefore,
V (x(t)) ≤ V (x(t0 )) − k(t − t0 ) ≤ c − k(t − t0 )
which shows that V (x(t)) reduces to ε within the time interval [t0 , t0 + (c − ε)/k].
In many problems, the inequality V̇ ≤ −W3 is obtained by using norm inequal-
ities. In such cases, it is more likely to arrive at

V̇ (t, x) ≤ −W3 (x), ∀ x ∈ D with x ≥ μ, ∀ t ≥ 0 (4.17)

If μ is sufficiently small, we can choose c and ε such that the set Λ is nonempty and
contained in D ∩ {x ≥ μ}. In particular, let α1 and α2 be class K functions such
that8
α1 (x) ≤ V (x) ≤ α2 (x) (4.18)
for all x ∈ D. From the left inequality of (4.18), we have

V (x) ≤ c ⇒ α1 (x) ≤ c
8 By Lemma 4.2, it is always possible to find such class K functions.
4.3. BOUNDEDNESS AND ULTIMATE BOUNDEDNESS 87

Ω c Ω c

Ω ε Ω ε

Bb
Λ Bμ

Figure 4.1: Representation of the Figure 4.2: Representation of the sets Ωε ,

set Λ, Ωε and Ωc . Ωc (solid) and Bμ , Bb (dashed).

Therefore, we can choose c > 0 such that Ωc = {V (x) ≤ c} is compact and contained
in D. In particular, if Br ⊂ D, c can be chosen as α1 (r). From the right inequality
of (4.18), we have
x ≤ μ ⇒ V (x) ≤ α2 (μ)
Taking ε = α2 (μ) ensures that Bμ ⊂ Ωε . To obtain ε < c, we must have μ < α2−1 (c).
The foregoing argument shows that all trajectories starting in Ωc enter Ωε within a
finite time T .9 To calculate the ultimate bound on x(t), we use the left inequality
of (4.18) to write
V (x) ≤ ε ⇒ α1 (x) ≤ ε = α2 (μ) ⇒ x ≤ α1−1 (α2 (μ))
Therefore, the ultimate bound can be taken as b = α1−1 (α2 (μ)). A sketch of the
sets Ωc , Ωε , Bμ , and Bb is shown in Figure 4.2.
If D = Rn and V (x) is radially unbounded, α1 and α2 can be chosen as class
K∞ functions, and so is α2−1 ◦ α1 . By choosing c large enough we can satisfy
the inequality μ < α2−1 (c) for any μ > 0 and include any initial state in the set
{V (x) ≤ c}. Our conclusions are summarized in the following theorem.
Theorem 4.4 Let D ⊂ Rn be a domain containing Bμ and V (x) be a continuously
differentiable function such that
α1 (x) ≤ V (x) ≤ α2 (x) (4.19)
∂V
f (t, x) ≤ −W3 (x), ∀ x ∈ D with x ≥ μ, ∀ t ≥ 0 (4.20)
∂x
where α1 and α2 are class K functions and W3 (x) is a continuous positive definite
function. Choose c > 0 such that Ωc = {V (x) ≤ c} is compact and contained in D
9 If the trajectory starts in Ωε , T = 0.
88 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

and suppose that μ < α2−1 (c). Then, Ωc is positively invariant for the system (4.13)
and there exists a class KL function β such that for every initial state x(t0 ) ∈ Ωc ,
the solution of (4.13) satisfies
 
x(t) ≤ max β(x(t0 ), t − t0 ), α1−1 (α2 (μ)) , ∀ t ≥ t0 (4.21)

If D = Rn and V (x) is radially unbounded, then (4.21) holds for any initial state
x(t0 ), with no restriction on how large μ is. 3

Inequality (4.21) shows that x(t) is uniformly bounded for all t ≥ t0 and uni-
formly ultimately bounded with the ultimate bound α1−1 (α2 (μ)). The ultimate
bound is a class K function of μ; hence, the smaller the value of μ, the smaller the
ultimate bound. As μ → 0, the ultimate bound approaches zero.

Example 4.8 It is shown in Section A.2 that a mass–spring system with a hard-
ening spring, linear viscous damping, and periodic force can be represented by

mÿ + cẏ + ky + ka2 y 3 = A cos ωt

Taking x1 = y, x2 = ẏ and assuming certain numerical values for the various

constants, the system is represented by the state model

ẋ1 = x2 , ẋ2 = −(1 + x21 )x1 − x2 + M cos ωt

where M ≥ 0 is proportional to the amplitude of the periodic force. When M = 0,

the system has an equilibrium point at the origin. It is shown in Example 3.7 that
the origin is globally asymptotically stable and a Lyapunov function is10
⎡ 1 1 ⎤ ⎡ 1 1 ⎤
2 2
 x1 2 2
V (x) = xT ⎣ ⎦x + 2 (y + y 3 ) dy = xT ⎣ ⎦ x + x21 + 1 x41
2
1 0 1
2 1 2 1
⎡ 3 1 ⎤
2 2
= xT ⎣ ⎦ x + 1 x41 def
2 = xT P x + 12 x41
1
2 1

When M > 0, we apply Theorem 4.4 with V (x) as a candidate function. The
function V (x) is positive definite and radially unbounded. From the inequalities

λmin (P )x2 ≤ xT P x ≤ V (x) ≤ xT P x + 12 x4 ≤ λmax (P )x2 + 12 x4

we see that V (x) satisfies (4.19) globally with α1 (r) = λmin (P )r2 and α2 (r) =
λmax (P )r2 + 12 r4 . The derivative of V along the trajectories of the system is
√
V̇ = −x21 − x41 − x22 + (x1 + 2x2 )M cos ωt ≤ −x2 − x41 + M 5x
10 The 1
constants δ and k of Example 3.7 are taken as δ = 2 and k = 2
.
4.3. BOUNDEDNESS AND ULTIMATE BOUNDEDNESS 89

|y T x| ≤ xy. To
where we wrote (x1 + 2x2 ) as y T x and used the inequality √
satisfy (4.20), we want to use part of −x to dominate M 5x for large x.
2

Towards that end, we rewrite the foregoing inequality as

√
V̇ ≤ −(1 − θ)x2 − x41 − θx2 + M 5x

where 0 < θ < 1. Then,

√
M 5
V̇ ≤ −(1 − θ)x − 2
x41 , ∀ x ≥
θ
√
which shows that inequality (4.20) is satisfied globally with μ = M 5/θ. Hence,
the solutions are globally uniformly ultimately bounded, with the ultimate bound

α2 (μ) λmax (P )μ2 + μ4 /2

b = α1−1 (α2 (μ)) = =
λmin (P ) λmin (P )

The next theorem is a special case of Theorem 4.4, which arises often in appli-
cations.
Theorem 4.5 Suppose the assumptions of Theorem 4.4 are satisfied with

c1 x2 ≤ V (x) ≤ c2 x2 (4.22)

∂V
f (t, x) ≤ −c3 x2 , ∀ x ∈ D with x ≥ μ, ∀ t ≥ 0 (4.23)
∂x

for some positive constants c1 to c3 , and μ < c/c2 . Then, the set Ωc = {V (x) ≤ c}
is positively invariant for the system (4.13) and for every initial state x(t0 ) ∈ Ωc ,
V (x(t)) and x(t) satisfy the inequalities
! "
V (x(t)) ≤ max V (x(t0 ))e−(c3 /c2 )(t−t0 ) , c2 μ2 , ∀ t ≥ t0 (4.24)
 ! "
c2
x(t) ≤ max x(t0 )e−(c3 /c2 )(t−t0 )/2 , μ , ∀ t ≥ t0 (4.25)
c1
If the assumptions hold globally, then (4.24) and (4.25) hold for any initial state
x(t0 ), with no restriction on how large μ is. 3
Proof: Inequality (4.25) follows from Theorem 4.4, except for the explicit expres-
sion of the class KL function β. When c2 μ2 ≤ V ≤ c, V̇ satisfies the inequality
V̇ ≤ −(c3 /c2 )V . By integration we see that V (x(t)) ≤ V (x(t0 ))e−(c3 /c2 )(t−t0 ) .
This bound is valid until x(t) enters the positively invariant set {V (x) ≤ c2 μ2 },
after which V (x(t)) ≤ c2 μ2 . The maximum of the two expressions gives the bound
(4.24). Inequality (4.25) follows from (4.24) by using (4.22). 2
90 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

Example 4.9 Consider the system

ẋ1 = x2 , ẋ2 = −x1 + 13 x31 − x2 + u(t)

where |u(t)| ≤ d for all t ≥ 0. With u = 0, it is shown in Example 3.7 that the
origin is asymptotically stable and a Lyapunov function is

 x1
1 T k k
V (x) = 2 x x+ h(y) dy
k 1 0

where 0 < k < 1 and h(y) = y − 13 y 3 . To satisfy (4.22), we limit our analysis to the
set {|x1 | ≤ 1}, where it can be shown that
 x1
2 2
x
3 1 ≤ x1 h(x1 ) ≤ x2
1 and 5 2
x
12 1 ≤ h(y) dy ≤ 12 x21
0

Therefore, V (x) satisfies xT P1 x ≤ V (x) ≤ xT P2 x, where

5

1 k+ 6 k 1 k+1 k
P1 = 2 , P2 = 2
k 1 k 1

and (4.22) holds with c1 = λmin (P1 ) and c2 = λmax (P2 ). The derivative of V is

V̇ = −kx1 h(x1 ) − (1 − k)x22 + (kx1 + x2 )u(t) ≤ − 23 kx21 − (1 − k)x22 + |kx1 + x2 | d

Taking 23 k = 1 − k ⇒ k = 35 , we obtain c1 = 0.2894, c2 = 0.9854, and

  2 def
V̇ ≤ − 25 x2 + 1 + 35 x d ≤ − 0.2 5 x ,
2
∀ x ≥ 3.2394 d = μ

where we wrote − 25 x2 as −(1−θ) 25 x2 −θ 25 x2 , with θ = 0.9, and used −θ 25 x2
to dominate the linear-in-x term. Let c = min|x1 |=1 V (x) = 0.5367, then Ωc =

{V (x) ≤ c} ⊂ {|x1 | ≤ 1}. To meet the condition μ < c/c2 we need d < 0.2278.
Now all the conditions of Theorem 4.5 are satisfied and  inequalities (4.24) and (4.25)
hold for all x(0) ∈ Ωc . The ultimate bound is b = μ c2 /c1 = 5.9775 d.

Theorems 4.4 and 4.5 play an important role in studying the stability of per-
turbed systems of the form
ẋ = f (x) + g(t, x) (4.26)
where the origin is an asymptotically stable equilibrium point of the nominal system
ẋ = f (x). We considered such perturbed system in Section 4.1 when g(t, 0) = 0
and derived conditions under which the origin is an asymptotically stable equilib-
rium point of the perturbed system. When g(t, 0)
= 0 the origin is no longer an
equilibrium point of the perturbed system and we should not expect its solution to
approach zero as t → ∞. The best we can hope for is that x(t) will be ultimately
bounded by a small bound, if the perturbation term g(t, x) is small in some sense.
We start with the case when the origin of the nominal system is exponentially stable.
4.3. BOUNDEDNESS AND ULTIMATE BOUNDEDNESS 91

Lemma 4.3 Consider the perturbed system ẋ = f (x) + g(t, x), where f and g are
locally Lipschitz in x and g is piecewise continuous in t, for all t ≥ 0 and x ∈ Br .
Let the origin be an exponentially stable equilibrium point of the nominal system
ẋ = f (x) and V (x) be a Lyapunov function that satisfies the inequalities11

∂V ∂V
c1 x ≤ V (x) ≤ c2 x ,
2 2
f (x) ≤ −c3 x ,
2 ≤ c4 x (4.27)
∂x ∂x

for all x ∈ Br with some positive constants c1 to c4 . Suppose g(t, x) satisfies


c3 c1
g(t, x) ≤ δ < θr (4.28)
c4 c2

for all t ≥ 0 and x ∈ Br , with some positive constant θ < 1. Then, for all x(t0 ) ∈
{V (x) ≤ c1 r2 }, the solution x(t) of the perturbed system satisfies

x(t) ≤ max {k exp[−γ(t − t0 )]x(t0 ), b} , ∀ t ≥ t0 (4.29)

where  
c2 (1 − θ)c3 δc4 c2
k= , γ= , b=
c1 2c2 θc3 c1
3
Proof: We use V (x) as a Lyapunov function candidate for the perturbed system.
∂V ∂V
V̇ (t, x) = f (x) + g(t, x)
∂x ∂x
∂V
≤ −c3 x2 +
∂x g(t, x)
≤ −c3 x2 + c4 δx
= −(1 − θ)c3 x2 − θc3 x2 + c4 δx, 0 < θ < 1
≤ −(1 − θ)c3 x2 , ∀ x ≥ δc4 /(θc3 )

Application of Theorem 4.5 completes the proof. 2

The ultimate bound b in Lemma 4.3 is proportional to δ (the upper bound on

the perturbation). This shows a robustness property of nominal systems having
exponentially stable equilibrium points at the origin, because arbitrarily small (uni-
formly bounded) perturbations will not result in large steady-state deviations from
the origin.

Example 4.10 Consider the two-dimensional system

ẋ1 = x2 , ẋ2 = −4x1 − 2x2 + βx32 + d(t)

11 When f is continuously differentiable, the existence of V is guaranteed by Theorem 3.8.
92 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

where β ≥ 0 is unknown and d(t) is a uniformly bounded disturbance that satisfies

|d(t)| ≤ δ for all t ≥ 0. This is the same system we studied in Example 4.5, except for
the additional perturbation term d(t). The system can be viewed as a perturbation
of a nominal linear system that has a Lyapunov function
⎡ 3 1 ⎤
2 8
V (x) = xT P x = xT ⎣ ⎦x
1 5
8 16

We use V (x) as a Lyapunov function candidate for the perturbed system, but treat
the two perturbation terms βx32 and d(t) differently, since the first term vanishes at
the origin while the second one does not. Calculating the derivative of V (x) along
the trajectories of the perturbed system, over the compact set Ωc = {xT P x ≤ c},
we obtain
   
V̇ (t, x) = −x2 + 2βx22 18 x1 x2 + 16
5 2
x2 + 2d(t) 18 x1 + 16
5
x2
√ √
29 2 29δ
≤ −x +2
βk2 x +
2
x
8 8
√ √
where k2 = maxx∈Ωc |x2 | = 1.8194 c. Suppose β ≤ 8(1 − ζ)/( 29k22 ), where
0 < ζ < 1. Then,
√ √
29δ 29δ
V̇ (t, x) ≤ −ζx +
2
x ≤ −(1 − θ)ζx , ∀ x ≥ μ =
2
8 8ζθ
where 0 < θ < 1. Thus, if β ≤ 0.448(1 − ζ)/c and δ is small enough that
μ2 λmax (P ) < c, then Bμ ⊂ Ωc and all trajectories starting inside Ωc remain for
all future time in Ωc . Furthermore, the conditions of Theorem 4.5 are satisfied
in Ωc . Therefore, the solutions of the perturbed system are uniformly ultimately
bounded by
√
29δ λmax (P )
b=
8ζθ λmin (P )

In the more general case when the origin x = 0 is an asymptotically, but not
exponentially, stable equilibrium point of ẋ = f (x), the analysis of the perturbed
system proceeds in a similar manner.
Lemma 4.4 Consider the perturbed system ẋ = f (x) + g(t, x), where f and g are
locally Lipschitz in x and g is piecewise continuous in t, for all t ≥ 0 and x ∈ Br .
Let the origin be an asymptotically stable equilibrium point of the nominal system
ẋ = f (x) and V (x) be a Lyapunov function that satisfies the inequalities12

∂V ∂V
α1 (x) ≤ V (x) ≤ α2 (x), f (x) ≤ −α3 (x), (x) ≤k
(4.30)
∂x ∂x
12 The
existence of a Lyapunov function satisfying these inequalities is guaranteed by Theorem 3.9
and Lemma 4.2.
4.3. BOUNDEDNESS AND ULTIMATE BOUNDEDNESS 93

for all x ∈ Br , where αi , i = 1, 2, 3, are class K functions. Suppose the perturbation

term g(t, x) satisfies the bound
θα3 (α2−1 (α1 (r)))
g(t, x) ≤ δ < (4.31)
k
for all t ≥ 0 and x ∈ Br , with some positive constant θ < 1. Then, for all x(t0 ) ∈
{V (x) ≤ α1 (r)}, the solution x(t) of the perturbed system satisfies
x(t) ≤ max {β(x(t0 ), t − t0 ), ρ(δ)} , ∀ t ≥ t0
for some class KL function β, and a class K function ρ defined by
   
δk
ρ(δ) = α1−1 α2 α3−1
θ
3
Proof: The derivative of V (x) along the trajectories of the perturbed system
satisfies
∂V ∂V
V̇ (t, x) = f (x) + g(t, x)
∂x ∂x

∂V
≤ −α3 (x) +
∂x g(t, x)
≤ −α3 (x) + δk
≤ −(1 − θ)α3 (x) − θα3 (x) + δk, 0 < θ < 1
 
δk
≤ −(1 − θ)α3 (x), ∀ x ≥ α3−1
θ
Application of Theorem 4.4 completes the proof. 2

This lemma is similar to Lemma 4.3 but there is an important difference between
the two cases. In the case of exponential stability, δ is required to satisfy (4.28),
whose right-hand side approaches ∞ as r → ∞. Therefore, if the assumptions hold
globally, we can conclude that for all uniformly bounded perturbations, the solution
of the perturbed system will be uniformly bounded. This is the case because, for
any δ, we can choose r large enough to satisfy (4.28). In the case of asymptotic
stability, δ is required to satisfy (4.31). Without further information about the
class K functions and how k depends on r, we cannot conclude that uniformly
bounded perturbations will result in bounded solutions irrespective of the size of
the perturbation. The following example illustrates this point.
Example 4.11 Consider the scalar system ẋ = −x/(1+x2 ) whose origin is globally
asymptotically stable. The Lyapunov function V (x) = x4 satisfies inequalities (4.30)
over Br with
4|x|4
α1 (|x|) = α2 (|x|) = |x|4 ; α3 (|x|) = ; k = 4r3
1 + |x|2
94 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

The right-hand side of (4.31) is given by

θα3 (α2−1 (α1 (r))) θr 1

= < 2
k 1 + r2
Consider now the perturbed system ẋ = −x/(1 + x2 ) + δ where δ > 12 is constant.
The right-hand side of the perturbed system will be always positive and x(t) will
escape to infinity, for any initial state x(0).

4.4 Input-to-State Stability

Consider the system
ẋ = f (x, u) (4.32)
where f is locally Lipschitz in x and u. The input u(t) is a piecewise continuous,
bounded function of t for all t ≥ 0. Suppose the unforced system

ẋ = f (x, 0) (4.33)

has a globally asymptotically stable equilibrium point at the origin x = 0. What

can we say about the behavior of the system (4.32) in the presence of a bounded
input u(t)? For the linear system

ẋ = Ax + Bu

with a Hurwitz matrix A, we can write the solution as

 t
x(t) = e(t−t0 )A x(t0 ) + e(t−τ )A Bu(τ ) dτ
t0

and use the bound e(t−t0 )A  ≤ k e−λ(t−t0 ) to estimate the solution by

 t
x(t) ≤ k e−λ(t−t0 ) x(t0 ) + k e−λ(t−τ ) Bu(τ ) dτ
t0
kB
≤ k e−λ(t−t0 ) x(t0 ) + sup u(τ )
λ t0 ≤τ ≤t

This estimate shows that the zero-input response decays to zero exponentially fast
while the zero-state response is bounded for every bounded input. In fact, the
estimate shows more than a bounded-input–bounded-state property. It shows that
the bound on the zero-state response is proportional to the bound on the input. How
much of this behavior should we expect for the nonlinear system ẋ = f (x, u)? For a
general nonlinear system, it should not be surprising that these properties may not
hold even when the origin of the unforced system is globally asymptotically stable.
The scalar system
ẋ = −3x + (1 + 2x2 )u
4.4. INPUT-TO-STATE STABILITY 95

has a globally exponentially stable origin when u = 0. Yet, when x(0) = 2 and
u(t) ≡ 1, the solution x(t) = (3 − et )/(3 − 2et ) is unbounded; it even has a finite
escape time.
Let us view the system ẋ = f (x, u) as a perturbation of the unforced system
ẋ = f (x, 0). Suppose we have a Lyapunov function V (x) for the unforced system
and let us calculate the derivative of V in the presence of u. Due to the boundedness
of u, it is plausible that in some cases it should be possible to show that V̇ is negative
outside a ball of radius μ, where μ depends on sup u. This would be expected,
for example, when the function f (x, u) satisfies the Lipschitz condition

f (x, u) − f (x, 0) ≤ Lu (4.34)

Showing that V̇ is negative outside a ball of radius μ would enable us to apply

Theorem 4.4 to show that
 
x(t) ≤ max β(x(t0 ), t − t0 ), α1−1 (α2 (μ)) , ∀ t ≥ t0

where β is a class KL function and α1−1 ◦ α2 is a class K function. This motivates

the next definition of input-to-state stability.13
Definition 4.4 The system ẋ = f (x, u) is input-to-state stable if there exist a class
KL function β and a class K function γ such that for any t0 ≥ 0, any initial state
x(t0 ), and any bounded input u(t), the solution x(t) exists for all t ≥ t0 and satisfies
  
x(t) ≤ max β(x(t0 ), t − t0 ), γ sup u(τ ) , ∀ t ≥ t0 (4.35)
t0 ≤τ ≤t

Input-to-state stability of ẋ = f (x, u) implies the following properties:

• For any bounded input u(t), the state x(t) is bounded;
 
• x(t) is ultimately bounded by γ supt0 ≤τ ≤t u(τ ) ;
• if u(t) converges to zero as t → ∞, so does x(t);
• The origin of the unforced system ẋ = f (x, 0) is globally asymptotically stable.
The Lyapunov-like theorem that follows gives a sufficient condition for input-
to-state stability.14
Theorem 4.6 Suppose f (x, u) is locally Lipschitz in (x, u) for all x ∈ Rn and
u ∈ Rm . Let V (x) be a continuously differentiable function such that, for all x ∈ Rn
and u ∈ Rm ,
α1 (x) ≤ V (x) ≤ α2 (x) (4.36)
13 Input-to-state stability was introduced in [131]; more recent results are in [133]. It can,

equivalently, be defined by inequality of the form x(t) ≤ β(x(t0 ), t−t0 )+γ(supt0 ≤τ ≤t u(τ )).
14 It is shown in [134] that the existence of a function V that satisfies inequalities (4.36) and

(4.37) is also necessary. In the literature, it is common to abbreviate input-to-state stability as

ISS and to call the function V of Theorem 4.6 an ISS-Lyapunov function.
96 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

∂V
f (x, u) ≤ −W3 (x), whenever x ≥ ρ(u) (4.37)
∂x
where α1 , α2 are class K∞ functions, ρ is class K function, and W3 (x) is a continu-
ous positive definite function on Rn . Then, the system ẋ = f (x, u) is input-to-state
stable with γ = α1−1 ◦ α2 ◦ ρ. 3

Proof: By applying the global version of Theorem 4.4 with μ = ρ(supτ ≥t0 u(τ )),
we find that for any x(t0 ) and any bounded u(t), the solution x(t) exists and satisfies
  
x(t) ≤ max β(x(t0 ), t − t0 ), γ sup u(τ ) , ∀ t ≥ t0 (4.38)
τ ≥t0

Since x(t) depends only on u(τ ) for t0 ≤ τ ≤ t, the supremum on the right-hand
side of (4.38) can be taken over [t0 , t], which yields (4.35). 2

The next lemma is an immediate consequence of the converse Lyapunov theorem

for global exponential stability (Theorem 3.8).15

Lemma 4.5 Suppose f (x, u) is continuously differentiable and globally Lipschitz

in (x, u). If ẋ = f (x, 0) has a globally exponentially stable equilibrium point at the
origin, then the system ẋ = f (x, u) is input-to-state stable. 3

In the absence of global exponential stability or globally Lipschitz functions, we

may still be able to show input-to-state stability by applying Theorem 4.6. This
process is illustrated by the three examples that follow.

Example 4.12 When u = 0, the system ẋ = −x3 +u has a globally asymptotically

stable origin. With V = 12 x2 and 0 < θ < 1,
 1/3
|u|
V̇ = −x4 + xu = −(1 − θ)x4 − θx4 + xu ≤ −(1 − θ)x4 , ∀ |x| ≥
θ

Thus, the system is input-to-state stable with γ(r) = (r/θ)1/3 .

Example 4.13 The system

ẋ = f (x, u) = −x − 2x3 + (1 + x2 )u2

has a globally exponentially stable origin when u = 0, but Lemma 4.5 does not
apply since f is not globally Lipschitz. Taking V = 12 x2 , we obtain

V̇ = −x2 − 2x4 + x(1 + x2 )u2 ≤ −x4 , ∀ |x| ≥ u2

Thus, the system is input-to-state stable with γ(r) = r2 .

15 See [74, Lemma 4.6] for the proof of this lemma.
4.4. INPUT-TO-STATE STABILITY 97

In Examples 4.12 and 4.13, the function V (x) = 12 x2 satisfies (4.36) with
α1 (r) = α2 (r) = 12 r2 . Hence, α1−1 (α2 (r)) = r and γ(r) reduces to ρ(r). In higher-
dimensional systems, the calculation of γ is more involved.
Example 4.14 Consider the system
ẋ1 = −x1 + x2 , ẋ2 = −x31 − x2 + u
With the Lyapunov function V (x) = 14 x41 + 12 x22 we can show that the origin is
globally asymptotically stable when u = 0. Using V (x) as a candidate function for
Theorem 4.6, we obtain
V̇ = −x41 − x22 + x2 u ≤ −x41 − x22 + |x2 | |u|
To use −x41 − x22 to dominate |x2 | |u|, we rewrite the foregoing inequality as
V̇ ≤ −(1 − θ)[x41 + x22 ] − θx41 − θx22 + |x2 | |u|
where 0 < θ < 1. The term −θx22 + |x2 | |u| has a maximum value of u2 /(4θ) and is
less than or equal to zero for |x2 | ≥ |u|/θ. Hence,
|u| u2
x21 ≥ or x22 ≥ 2 ⇒ −θx41 − θx22 + |x2 | |u| ≤ 0
2θ θ
Consequently,
|u| u2
x2 ≥ ⇒ −θx41 − θx22 + |x2 | |u| ≤ 0
+ 2
2θ θ

Defining the class K function ρ by ρ(r) = r/(2θ) + r2 /θ2 , we see that inequality
(4.37) is satisfied as
V̇ ≤ −(1 − θ)[x41 + x22 ], ∀ x ≥ ρ(|u|)
Since V (x) is positive definite and radially unbounded, Lemma 4.2 shows that there
are class K∞ functions α1 and α2 that satisfy (4.36) globally. Hence, the system is
input-to-state stable.
An interesting property of input-to-state stability is stated in the following
lemma.16
Lemma 4.6 If the systems η̇ = f1 (η, ξ) and ξ˙ = f2 (ξ, u) are input-to-state stable,
then the cascade connection
η̇ = f1 (η, ξ), ξ˙ = f2 (ξ, u)
is input-to-state stable. Consequently, If η̇ = f1 (η, ξ) is input-to-state stable and the
origin of ξ˙ = f2 (ξ) is globally asymptotically stable, then the origin of the cascade
connection
η̇ = f1 (η, ξ), ξ˙ = f2 (ξ)
is globally asymptotically stable. 3
16 See [130] for the proof.
98 CHAPTER 4. TIME-VARYING AND PERTURBED SYSTEMS

Example 4.15 The system

ẋ1 = −x1 + x22 , ẋ2 = −x2 + u

is a cascade connection of ẋ1 = −x1 + x22 and ẋ2 = −x2 + u. The system ẋ1 =
−x1 + x22 is input-to-state stable, as seen from Theorem 4.6 with V (x1 ) = 12 x21 ,
whose derivative satisfies

V̇ = −x21 + x1 x22 ≤ −(1 − θ)x21 , for |x1 | ≥ x22 /θ

where 0 < θ < 1. The linear system ẋ2 = −x2 + u is input-to-state stable by
Lemma 4.5. Hence, the cascade connection is input-to-state stable.

The notion of input-to-state stability is defined for the global case where the
initial state and input can be arbitrarily large. Regional and local versions are
presented next.
Definition 4.5 Let X ⊂ Rn and U ⊂ Rm be bounded sets containing their respec-
tive origins as interior points. The system ẋ = f (x, u) is regionally input-to-state
stable with respect to X × U if there exist a class KL function β and a class K
function γ such that for any initial state x(t0 ) ∈ X and any input u with u(t) ∈ U
for all t ≥ t0 , the solution x(t) belongs to X for all t ≥ t0 and satisfies
  
x(t) ≤ max β(x(t0 ), t − t0 ), γ sup u(τ ) (4.39)
t0 ≤τ ≤t

The system ẋ = f (x, u) is locally input-to-state stable if it is regionally input-to-state

stable with respect to some neighborhood of the origin (x = 0, u = 0).

Theorem 4.7 Suppose f (x, u) is locally Lipschitz in (x, u) for all x ∈ Br and
u ∈ Bλ . Let V (x) be a continuously differentiable function that satisfies (4.36) and
(4.37) for all x ∈ Br and u ∈ Bλ , where α1 , α2 , and ρ are class K functions,
and W3 (x) is a continuous positive definite function. Suppose α1 (r) > α2 (ρ(λ))
and let Ω = {V (x) ≤ α1 (r)}. Then, the system ẋ = f (x, u) is regionally input-to-
state stable with respect to Ω × Bλ . The class K function γ in (4.39) is given by
γ = α1−1 ◦ α2 ◦ ρ. 3
 
Proof: Apply Theorem 4.4 with μ = ρ supt≥t0 u(t) . 2

Local input-to-state stability of ẋ = f (x, u) is equivalent to asymptotic stability

of the origin of ẋ = f (x, 0), as shown in the following lemma.

Lemma 4.7 Suppose f (x, u) is locally Lipschitz in (x, u) in some neighborhood of

(x = 0, u = 0). Then, the system ẋ = f (x, u) is locally input-to-state stable if and
only if the unforced system ẋ = f (x, 0) has an asymptotically stable equilibrium
point at the origin. 3
4.5. EXERCISES 99

Proof: With u(t) ≡ 0, (4.39) reduces to x(t) ≤ β(x(t0 ), t − t0 ); hence, the
origin of ẋ = f (x, 0) is asymptotically stable. On the other hand, Theorem 3.9
shows that if the origin of ẋ = f (x, 0) is asymptotically stable, then there is a
Lyapunov function V (x) that satisfies

∂V
f (x, 0) ≤ −U (x)
∂x
in some neighborhood of the origin, where U (x) is continuous and positive definite.
Lemma 4.2 shows that V (x) satisfies (4.36) and U (x) satisfies U (x) ≥ α3 (x)
for some class K functions α1 , α2 , and α3 . The derivative of V with respect to
ẋ = f (x, u) satisfies

∂V ∂V
V̇ = f (x, 0) + [f (x, u) − f (x, 0)] ≤ −α3 (x) + kLu
∂x ∂x
where k is an upper bound on [∂V /∂x] and L is a Lipschitz constant of f with
respect to u. Hence,

V̇ ≤ −(1 − θ)α3 (x) − θα3 (x) + kLu

where 0 < θ < 1. Then,

 
kLu
∀ x ≥ α3−1
def
V̇ ≤ −(1 − θ)α3 (x), = ρ(u)
θ

in some neighborhood of (x = 0, u = 0). Taking r and λ sufficiently small with

α1 (r) > α2 (ρ(λ)), we see that the conditions of Theorem 4.7 are satisfied. Hence,
the system ẋ = f (x, u) is locally input-to-state stable. 2

4.5 Exercises
4.1 Consider the linear time-varying system ẋ = [A + B(t)]x where A is Hurwtiz
and B(t) is piecewise continuous. Let P be the solution of the Lyapunov equa-
tion P A + AT P = −I. Show that the origin is globally exponentially stable if
2||P B(t) ≤ a < 1 for all t ≥ 0.

4.2 For each of the following systems determine whether the origin is uniformly
stable, uniformly asymptotically stable, exponentially stable, or none of the above.
In all cases, g(t) is piecewise continuous and bounded.

(1) ẋ1 = −x31 + g(t)x2 , ẋ2 = −g(t)x1 − x2

(2) ẋ = −g(t)h(x), where xh(x) ≥ ax2 ∀ x with a > 0 and g(t) ≥ k > 0 ∀ t ≥ 0

(3) ẋ1 = g(t)x2 , ẋ2 = −g(t)x1

100 CHAPTER 4. TIME-VARYING AND PERTURBED

(4) ẋ1 = −g(t)x1 + x2 , ẋ2 = x1 − x2 , g(t) ≥ 2 ∀ t ≥ 0

(5) ẋ1 = g(t)x1 + x2 , ẋ2 = −x1 − x2 , where |g(t)| ≤ 1
4 ∀t≥0

4.3 Consider the linear system

ẋ1 = x2 /g(t), ẋ2 = −x1 − x2 /g(t)

where g(t) = 2 − e−t/2 . Using V (t, x) = [1 + 2g(t)]x21 + 2x1 x2 + 2x22 show that the
origin is exponentially stable.

4.4 Consider the linear system

ẋ1 = x2 , ẋ2 = −g(t)x1 − 2x2

where g(t) is piecewise continuous and bounded. Using V (t, x) = [2 + g(t)]x21 +

2x1 x2 +x22 show that the origin is exponentially stable if g(t) ≥ 0 and −2g(t)+ ġ(t) ≤
−a < 0 for all t ≥ 0.

4.5 Let g(t) be piecewise continuous and g(t) ≥ g0 > 0 for all t ≥ 0. Show that if
the origin of the n-dimensional system ẋ = f (x) is asymptotically stable, then the
origin of ẋ = g(t)f (x) is uniformly asymptotically stable.

4.6 Consider the system

ẋ1 = −x1 + x2 , ẋ2 = ax1 − ax2 − g(t)x32

where a is a positive constant and g(t) is piecewise continuous and satisfies g(t) ≥
b > 0 for all t ≥ 0. Show that the origin is globally uniformly asymptotically stable.

4.7 Consider the system ẋ = Ax + Bu with the stabilizing control u = −F x,

where (A − BF ) is Hurwitz. Suppose that, due to physical limitations, we have
to use a limiter so that |ui (t)| ≤ L. The closed-loop system is given by ẋ =
Ax − BL sat(F x/L), where sat(v) is a vector whose ith component is the saturation
function. By adding and subtracting BF x, we can write the closed-loop system as
ẋ = (A − BF )x − Bh(F x), where h(v) = L sat(v/L) − v. Thus, the effect of the
limiter can be viewed as a perturbation of the nominal system ẋ = (A − BF )x.
(a) Show that |hi (v)| ≤ [δ/(1 + δ)]|vi |, ∀ |vi | ≤ L(1 + δ), where δ > 0.
(b) Show that the origin of the perturbed system is exponentially stable if δ/(1 +
δ) < 12 P B F , where P is the solution of P (A − BF ) + (A − BF )T P = −I,
and discuss how would you estimate the region of attraction.

4.8 For each of the following systems, find a bound on |a| such that the origin is
exponentially stable.
(1) ẋ1 = −x1 − 2x2 + a(x1 − x2 )x2 , ẋ2 = 2x1 − x2 − a(x1 + x2 )x2
4.5. EXERCISES 101

(2) ẋ1 = ax21 + x2 , ẋ2 = −x1 − x31 /(1 + x21 ) − x2

(3) ẋ1 = x2 , ẋ2 = −x1 − x2 + a(x1 − x2 )

4.9 For each of the following systems, find a bound on |a| such that the origin is
globally asymptotically stable.

(1) ẋ1 = ax31 /(1 + x21 ) + x2 , ẋ2 = −x1 − x31 /(1 + x21 ) − x2

(2) ẋ1 = −x31 + x2 , ẋ2 = −x31 − x2 + ax31

(3) ẋ1 = −x1 + h(x2 ), ẋ2 = −x1 − h(x2 ) − ax1 , where h is locally Lipschitz,
h(0) = 0, and yh(y) > 0 ∀ y
= 0.

4.10 Consider the system

x2 −x1
ẋ1 = −x1 + , ẋ2 = − x2 + a(x1 + x2 )
1 + x21 1 + x21

(a) With a = 0, show that the origin is globally exponentially stable.

(b) Find an upper bound on |a| such that the origin is globally exponentially stable.

(c) Show that the origin is exponentially stable for a < 1 and unstable for a > 1.

(d) Estimate the region of attraction when a = 0.9.

(e) Replace a(x1 + x2 ) by a. Apply Theorem 4.5 to show that x(t) is globally
ultimately bounded and estimate the ultimate bound in terms of a.

4.11 For each of the following systems, find a compact set Ω such that for all
x(0) ∈ Ω, the solution x(t) is ultimately bounded and estimate the ultimate bound
in terms of d. Hint: Apply Theorem 4.5 using V (x) from Example 3.7 in (1) and
(2) and from Example 3.14 in (3).

(1) ẋ1 = x2 , ẋ2 = − sin(x1 ) − x2 + d

(2) ẋ1 = x2 , ẋ2 = −x1 + 19 x31 − 2x2 + d

(3) ẋ1 = x2 , ẋ2 = x1 + (x21 − 1)x2 + d

4.12 For each of the following systems, show that the origin is globally asymptot-
ically stable when a = 0 and the solution is globally ultimately bounded by a class
K function of |a| when a
= 0.

(1) ẋ1 = −x31 + x2 , ẋ2 = −x2 + a

(2) ẋ1 = −x31 + x2 + a, ẋ2 = −x2

102 CHAPTER 4. TIME-VARYING AND PERTURBED

4.13 Consider the system

ẋ1 = x2 , ẋ2 = −x31 − x2 + M cos(ωt)

(a) With M = 0, show that global asymptotic stability of the origin can be es-
tablished using the Lyapunov function V (x) = 12 x21 + x1 x2 + x22 + 12 x41 or the
Lyapunov function V (x) = 14 x41 + 12 x22 .

(b) Calculate an ultimate bound on the solution as a class K function of M .

4.14 For each of the following systems, investigate input-to-state stability. The
function h is locally Lipschitz, h(0) = 0, and yh(y) ≥ ay 2 ∀ y, with a > 0.

(1) ẋ = −h(x) + u2

(2) ẋ = h(x) + u

(3) ẋ = −x/(1 + x2 ) + u

(4) ẋ = −x3 /(1 + x2 ) + u

(5) ẋ1 = −x1 + x2 , ẋ2 = −x1 (2 − x1 )2 − x2 + u

(6) ẋ1 = −x1 − x31 + x2 , ẋ2 = −x1 − x2 + u

(7) ẋ1 = −x1 − x31 + x2 , ẋ2 = −x1 − x2 + x3 , ẋ3 = −x33 + u

(8) ẋ1 = x2 , ẋ2 = −h(x1 ) − x2 + u

(9) ẋ1 = x2 , ẋ2 = −h(x1 ) − x2 + x3 , ẋ3 = −h(x3 ) + u

(10) ẋ1 = −x1 + x2 , ẋ2 = −h(x1 ) − x2 + u

4.15 Show that each of the following systems is not input-to-state stable, but
regionally input-to-state stable, and estimate the sets X and U of Definition 4.5.

(1) ẋ1 = −x2 , ẋ2 = x1 + (x21 − 1)x2 + u

(2) ẋ1 = x2 , ẋ2 = −x1 + 13 x31 − x2 + u

(3) ẋ1 = x2 , ẋ2 = −4(x1 + x2 ) − (x1 + x2 )[1 − (x1 + x2 )2 ] + u

Chapter 5

Passivity

Passivity provides us with a useful tool for the analysis of nonlinear systems, which
relates to Lyapunov stability of the previous two chapters and L2 stability of the
next one. We start in Section 5.1 by defining passivity of memoryless nonlinearities
and introduce the related notion of sector nonlinearities. We extend the definition
of passivity to dynamical systems, represented by state models, in Section 5.2. In
both cases, we use electrical networks to motivate the definitions. In Section 5.3,
we study positive real and strictly positive real transfer functions and show that
they represent passive and strictly passive systems, respectively. The connection
between passivity and Lyapunov stability is established in Section 5.4.1

5.1 Memoryless Functions

Our goal in this section is to define passivity of the memoryless function y = h(t, u),
where u and y are m-dimensional vectors and t ≥ 0. We use electrical networks
to motivate the definition. Figure 5.1(a) shows a one-port resistive element, which
we view as a system with the voltage u as input and the current y as output. The
resistive element is passive if the inflow of power is always nonnegative; that is, if
uy ≥ 0 for all points (u, y) on its u–y characteristic. Geometrically, this means that
the u–y curve must lie in the first and third quadrants, as shown in Figure 5.1(b).
The simplest such resistive element is a linear resistor that obeys Ohm’s law u = Ry
or y = Gu, where R is the resistance and G = 1/R is the conductance. For positive
resistance, the u–y characteristic is a straight line of slope G and the product
uy = Gu2 is always nonnegative. In fact, it is always positive except at the origin
point (0, 0). Nonlinear passive resistive elements have nonlinear u–y curves lying in
the first and third quadrants; examples are shown in Figures 5.2(a) and (b). Notice
that the tunnel-diode characteristic of Figure 5.2(b) is still passive even though the
1 Thecoverage of passivity in this chapter and Section 7.1 is based on [20, 60, 61, 81, 121, 139,
144]. An expanded treatment of positive real transfer functions is given in [1].

103
104 CHAPTER 5. PASSIVITY

+ y
y

u
u

(a) (b)

Figure 5.1: (a) A passive resistor; (b) u–y characteristic lies in the first–third quadrant.

y y y

u u u

(a) (b) (c)

Figure 5.2: (a) and (b) are examples of nonlinear passive resistor characteristics; (c) is
an example of a nonpassive resistor.

curve has negative slope in some region. As an example of an element that is not
passive, Figure 5.2(c) shows the u–y characteristic of a negative resistance that is
used in Section A.4 to construct the negative resistance oscillator. For a multiport
network where u and y are vectors, the power flow into the network is the inner
product uT y = Σm i=1 ui yi = Σi=1 ui hi (t, u). The network is passive if u y ≥ 0 for
m T

all u. This concept of passivity is now abstracted and assigned to any function
y = h(t, u) irrespective of its physical origin. We think of uT y as the power flow
into the system and say that the system is passive if uT y ≥ 0 for all u. For the
scalar case, the graph of the input–output relation must lie in the first and third
quadrants. We also say that the graph belongs to the sector [0, ∞], where zero and
infinity are the slopes of the boundaries of the first–third quadrant region. The
graphical representation is valid even when h is time varying. In this case, the u–y
curve will be changing with time, but will always belong to the sector [0, ∞]. For
a vector function, we can give a graphical representation in the special case when
h(t, u) is decoupled in the sense that hi (t, u) depends only on ui ; that is,
h(t, u) = col (h1 (t, u1 ), h2 (t, u2 ), . . . , hm (t, um )) (5.1)
In this case, the graph of each component belongs to the sector [0, ∞]. In the general
case, such graphical representation is not possible, but we will continue to use the
sector terminology by saying that h belongs to the sector [0, ∞] if uT h(t, u) ≥ 0 for
all (t, u).
5.1. MEMORYLESS FUNCTIONS 105

An extreme case of passivity happens when uT y = 0. In this case, we say that

the system is lossless. An example of a lossless system is the ideal transformer
shown in Figure 5.3. Here y = Su, where

v1 i1 0 −N
u= , y= , S=
i2 v2 N 0

The matrix S is skew-symmetric; that is, S + S T = 0. Hence, uT y = uT Su =

1 T T
2 u (S + S )u = 0.

N
+ +
i1 i2
v2 = Nv1
v1 v2 i1 = −Ni2

− −

Figure 5.3: Ideal transformer

Consider now a function h satisfying uT y = uT h(t, u) ≥ uT ϕ(u) > 0 for all

u = 0. The function h is called input strictly passive because passivity is strict in
the sense that uT y = 0 only if u = 0. Equivalently, in the scalar case, the u–y graph
does not touch the u-axis, except at the origin.
On the other hand, suppose uT y = uT h(t, u) ≥ y T ρ(y) > 0 for all y = 0. Then,
h is called output strictly passive because passivity is strict in the sense that uT y = 0
only if y = 0. Equivalently, in the scalar case, the u–y graph does not touch the
y-axis, except at the origin. For convenience, we summarize the various notions of
passivity in the next definition.

Definition 5.1 The system y = h(t, u) is

• passive if uT y ≥ 0.

• lossless if uT y = 0.

• input strictly passive if uT y ≥ uT ϕ(u) > 0, ∀ u = 0.

• output strictly passive if uT y ≥ y T ρ(y) > 0, ∀ y = 0.

In all cases, the inequality should hold for all (t, u).

Consider next a scalar function y = h(t, u), which satisfies the inequalities

αu2 ≤ uh(t, u) ≤ βu2 (5.2)

106 CHAPTER 5. PASSIVITY

y = βu y = βu
y y

y = αu

u u

y = αu

(a) α > 0 (b) α > 0

Figure 5.4: The sector [α, β] for β > 0 and (a) α > 0; (b) α < 0.

or, equivalently,
[h(t, u) − αu][h(t, u) − βu] ≤ 0 (5.3)
for all (t, u), where α and β are real numbers with β ≥ α. The graph of this function
belongs to a sector whose boundaries are the lines y = αu and y = βu. We say
that h belongs to the sector [α, β]. Figure 5.4 shows the sector [α, β] for β > 0 and
different signs of α. If strict inequality is satisfied on either side of (5.2), we say
that h belongs to a sector (α, β], [α, β), or (α, β), with obvious implications. To
extend the sector definition to the vector case, consider first a function h(t, u) that
is decoupled as in (5.1). Suppose each component hi satisfies the sector condition
(5.2) with constants αi and βi > αi . Taking

K1 = diag(α1 , α2 , . . . , αm ), K2 = diag(β1 , β2 , . . . , βm )

it can be easily seen that

[h(t, u) − K1 u]T [h(t, u) − K2 u] ≤ 0 (5.4)

for all (t, u). Note that K = K2 − K1 is a positive definite symmetric (diagonal)
matrix. Inequality (5.4) may hold for more general vector functions. For example,
suppose h(t, u) satisfies the inequality

h(t, u) − Lu ≤ γu

for all (t, u). Taking K1 = L − γI and K2 = L + γI, we can write

[h(t, u) − K1 u]T [h(t, u) − K2 u] = h(t, u) − Lu2 − γ 2 u2 ≤ 0

Once again, K = K2 − K1 is a positive definite symmetric (diagonal) matrix. We

use inequality (5.4) with a positive definite symmetric matrix K = K2 − K1 as a
definition of the sector [K1 , K2 ] in the vector case. The next definition summaries
the sector terminology.
5.2. STATE MODELS 107

ũ + û u y + ỹ
K−1 y = h(t, u)
+ −

K1

Figure 5.5: A function in the sector [K1 , K2 ] can be transformed into a function in the
sector [0, ∞] by input feedforward followed by output feedback.

Definition 5.2 A memoryless function h(t, u) belongs to the sector

• [0, ∞] if uT h(t, u) ≥ 0.
• [K1 , ∞] if uT [h(t, u) − K1 u] ≥ 0.
• [0, K2 ] with K2 = K2T > 0 if hT (t, u)[h(t, u) − K2 u] ≤ 0.
• [K1 , K2 ] with K = K2 − K1 = K T > 0 if
[h(t, u) − K1 u]T [h(t, u) − K2 u] ≤ 0 (5.5)

In all cases, the inequality should hold for all (t, u). If in any case the inequality is
strict, we write the sector as (0, ∞), (K1 , ∞), (0, K2 ), or (K1 , K2 ). In the scalar
case, we write (α, β], [α, β), or (α, β) to indicate that one or both sides of (5.2) is
satisfied as a strict inequality.
The sector [0, K2 ] with K2 = (1/δ)I > 0 corresponds to output strict passivity with
ρ(y) = δy. A function in the sector [K1 , K2 ] is transformed into a function in the
sector [0, ∞] as shown in Figure 5.5. Define ũ, û, and ỹ as in the figure. Then,
ỹ = h(t, u) − K1 u and û = Ku = ũ + ỹ. From [h(t, u) − K1 u]T [h(t, u) − K2 u] ≤ 0
we have
ỹ T (ỹ − Ku) ≤ 0 ⇒ ỹ T (ỹ − û) ≤ 0 ⇒ ỹ T (−ũ) ≤ 0 ⇔ ỹ T ũ ≥ 0

5.2 State Models

Let us now define passivity for a dynamical system represented by the state model
ẋ = f (x, u), y = h(x, u) (5.6)
where f is locally Lipschitz, h is continuous, f (0, 0) = 0, and h(0, 0) = 0. The
system has the same number of inputs and outputs. The following RLC circuit
motivates the definition.
108 CHAPTER 5. PASSIVITY

v2 = h2(i2)
L

y i2 iL
+ v2 −
+ + +
+
u v1 i1 = h1(v1) vc C v3 i3 = h3(v3)
−
− − −
i1 i3

Figure 5.6: RLC circuit of Example 5.1.

Example 5.1 The RLC circuit of Figure 5.6 features a voltage source connected
to an RLC network with linear inductor and capacitor and nonlinear resistors. The
nonlinear resistors 1 and 3 are represented by their v–i characteristics i1 = h1 (v1 )
and i3 = h3 (v3 ), while resistor 2 is represented by its i–v characteristic v2 = h2 (i2 ).
We take the voltage u as the input and the current y as the output. The product
uy is the power flow into the network. Taking the current x1 through the inductor
and the voltage x2 across the capacitor as the state variables, the state model is
Lẋ1 = u − h2 (x1 ) − x2 , C ẋ2 = x1 − h3 (x2 ), y = x1 + h1 (u)
The new feature of an RLC network over a resistive network is the presence of the
energy-storing elements L and C. The system is passive if the energy absorbed by
the network over any period of time [0, t] is greater than or equal to the change in
the energy stored in the network over the same period; that is,
 t
u(s)y(s) ds ≥ V (x(t)) − V (x(0)) (5.7)
0

where V (x) = 12 Lx21 + 12 Cx22 is the energy stored in the network. If (5.7) holds with
strict inequality, then the difference between the absorbed energy and the change in
the stored energy must be the energy dissipated in the resistors. Since (5.7) must
hold for every t ≥ 0, the instantaneous power inequality
u(t)y(t) ≥ V̇ (x(t), u(t)) (5.8)
must hold for all t; that is, the power flow into the network must be greater than or
equal to the rate of change of the energy stored in the network. We can investigate
inequality (5.8) by calculating the derivative of V along the trajectories of the
system. We have
V̇ = Lx1 ẋ1 + Cx2 ẋ2 = x1 [u − h2 (x1 ) − x2 ] + x2 [x1 − h3 (x2 )]
= x1 [u − h2 (x1 )] − x2 h3 (x2 )
= [x1 + h1 (u)]u − uh1 (u) − x1 h2 (x1 ) − x2 h3 (x2 )
= uy − uh1 (u) − x1 h2 (x1 ) − x2 h3 (x2 )
5.2. STATE MODELS 109

Thus,
uy = V̇ + uh1 (u) + x1 h2 (x1 ) + x2 h3 (x2 )

If h1 , h2 , and h3 are passive, uy ≥ V̇ and the system is passive. Other possibilities

are illustrated by four different special cases of the network.

Case 1: If h1 = h2 = h3 = 0, then uy = V̇ and there is no energy dissipation in

the network; that is, the system is lossless.

Case 2: If h2 and h3 are passive, then uy ≥ V̇ + uh1 (u). If uh1 (u) > 0 for all
u = 0, the energy absorbed over [0, t] will be equal to the change in the stored
energy only when u(t) ≡ 0. This is a case of input strict passivity.

Case 3: If h1 = 0 and h3 is passive, then uy ≥ V̇ + yh2 (y). When yh2 (y) > 0 for
all y = 0, we have output strict passivity because the energy absorbed over
[0, t] will be equal to the change in the stored energy only when y(t) ≡ 0.

Case 4: If h1 ∈ [0, ∞], h2 ∈ (0, ∞), and h3 ∈ (0, ∞), then

uy ≥ V̇ + x1 h2 (x1 ) + x2 h3 (x2 )

where x1 h2 (x1 ) + x2 h3 (x2 ) is a positive definite function of x. This is a case

of state strict passivity because the energy absorbed over [0, t] will be equal
to the change in the stored energy only when x(t) ≡ 0. A system having this
property is called state strictly passive or, simply, strictly passive.

Definition 5.3 The system (5.6) is passive if there exists a continuously differen-
tiable positive semidefinite function V (x) (called the storage function) such that

∂V
uT y ≥ V̇ = f (x, u), ∀ (x, u) (5.9)
∂x
Moreover, it is

• lossless if uT y = V̇ .

• input strictly passive if uT y ≥ V̇ + uT ϕ(u) and uT ϕ(u) > 0, ∀ u = 0.

• output strictly passive if uT y ≥ V̇ + y T ρ(y) and y T ρ(y) > 0, ∀ y = 0.

• strictly passive if uT y ≥ V̇ + ψ(x) for some positive definite function ψ.

In all cases, the inequality should hold for all (x, u).
110 CHAPTER 5. PASSIVITY

u + x=y
h(.)
u x=y −

u x y
+ h(.)

(a) (b) (c)

Figure 5.7: Example 5.2

Example 5.2 The integrator of Figure 5.7(a), represented by

ẋ = u, y=x
is a lossless system since, with V (x) = 12 x2 as the storage function, uy = V̇ . When
a memoryless function is connected in parallel with the integrator, as shown in
Figure 5.7(b), the system is represented by
ẋ = u, y = x + h(u)
With V (x) = 1 2
2x as the storage function, uy = V̇ +uh(u). If h ∈ [0, ∞], the system
is passive. If uh(u) > 0 for all u = 0, the system is input strictly passive. When the
loop is closed around the integrator with a memoryless function, as in Figure 5.7(c),
the system is represented by
ẋ = −h(x) + u, y=x
With V (x) = 12 x2 as the storage function, uy = V̇ + yh(y). If h ∈ [0, ∞], the system
is passive. If yh(y) > 0 for all y = 0, the system is output strictly passive. 

Example 5.3 The cascade connection of an integrator and a passive memoryless

function, shown in Figure 5.8(a), is represented by
ẋ = u, y = h(x)
x x
Passivity of h guarantees that 0 h(σ) dσ ≥ 0 for all x. With V (x) = 0 h(σ) dσ as
the storage function, V̇ = h(x)ẋ = yu. Hence, the system is lossless. Suppose now
the integrator is replaced by the transfer function 1/(as + 1) with a > 0, as shown
in Figure 5.8(b). The system can be represented by the state model
aẋ = −x + u, y = h(x)
x
With V (x) = a 0
h(σ) dσ as the storage function,

V̇ = h(x)(−x + u) = yu − xh(x) ≤ yu
Hence, the system is passive. If xh(x) > 0 ∀ x = 0, it is strictly passive. 
5.2. STATE MODELS 111

u x y u 1 x y
h(.) h(.)
as + 1

(a) (b)

Figure 5.8: Example 5.3

Example 5.4 Consider the system

ẋ1 = x2 , ẋ2 = −h(x1 ) − ax2 + u, y = bx2 + u

where h ∈ [α1 , ∞], and a, b, α1 are positive constants. Let

 x1  x1
V (x) = α h(σ) dσ + 12 αxT P x = α h(σ) dσ + 12 α(p11 x21 + 2p12 x1 x2 + p22 x22 )
0 0

where α, p11 , and p11 p22 − p212 are positive. Use V as a storage function candidate.

uy − V̇ = u(bx2 +u)−α[h(x1 )+p11 x1 +p12 x2 ]x2 −α(p12 x1 +p22 x2 )[−h(x1 )−ax2 +u]

Take p22 = 1, p11 = ap12 , and α = b to cancel the cross product terms x2 h(x1 ),
x1 x2 , and x2 u, respectively. Then,

uy − V̇ = u2 − bp12 x1 u + bp12 x1 h(x1 ) + b(a − p12 )x22

 2
= u − 12 bp12 x1 − 14 b2 p212 x21 + bp12 x1 h(x1 ) + b(a − p12 )x22
 
≥ bp12 α1 − 14 bp12 x21 + b(a − p12 )x22

 p12 1= ak, where 0 < k < min{1, 4α1 /(ab)}, ensures that p11 , p11 p22 − p12 ,
2
Taking
bp12 α1 − 4 bp12 , and b(a − p12 ) are positive. Thus, the preceding inequality shows
that the system is strictly passive. 

Example 5.5 Consider the pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + cu

where b ≥ 0 and c > 0. View y = x2 as the output and use the total energy

V (x) = α[(1 − cos x1 ) + 12 x22 ]

where α > 0, as a storage function candidate. Note that when viewed as a function
on the whole space R2 , V (x) is positive semidefinite but not positive definite because
it is zero at points other than the origin. We have

uy − V̇ = ux2 − α[x2 sin x1 − x2 sin x1 − bx22 + cx2 u]

112 CHAPTER 5. PASSIVITY

Taking α = 1/c to cancel the cross product term x2 u, we obtain

uy − V̇ = (b/c)x22 ≥ 0

Hence, the system is passive when b = 0 and output strictly passive when b > 0.


5.3 Positive Real Transfer Functions

Definition 5.4 An m × m proper rational transfer function matrix G(s) is positive
real if
• poles of all elements of G(s) are in Re[s] ≤ 0,
• for all real ω for which jω is not a pole of any element of G(s), the matrix
G(jω) + GT (−jω) is positive semidefinite, and
• any pure imaginary pole jω of any element of G(s) is a simple pole and the
residue matrix lims→jω (s − jω)G(s) is positive semidefinite Hermitian.
It is strictly positive real if G(s − ε) is positive real for some ε > 0.

When m = 1, G(jω)+GT (−jω) = 2Re[G(jω)], an even function of ω. Therefore,

the second condition of Definition 5.4 reduces to Re[G(jω)] ≥ 0, ∀ ω ∈ [0, ∞), which
holds when the Nyquist plot of G(jω) lies in the closed right-half complex plane.
This is a condition that can be satisfied only if the relative degree of the transfer
function is zero or one.2
The next lemma gives an equivalent characterization of strictly positive real
transfer functions.3
Lemma 5.1 Let G(s) be an m × m proper rational transfer function matrix, and
suppose det [G(s) + GT (−s)] is not identically zero.4 Then, G(s) is strictly positive
real if and only if
• G(s) is Hurwitz; that is, poles of all elements of G(s) have negative real parts,
• G(jω) + GT (−jω) is positive definite for all ω ∈ R, and
• either G(∞) + GT (∞) is positive definite or it is positive semidefinite and
limω→∞ ω 2(m−q) det[G(jω)+GT (−jω)] > 0, where q = rank[G(∞)+GT (∞)].
In the case m = 1, the frequency-domain condition of the lemma reduces to
Re[G(jω)] > 0 for all ω ∈ [0, ∞) and either G(∞) > 0 or G(∞) = 0 and
limω→∞ ω 2 Re[G(jω)] > 0.
2 The relative degree of a rational transfer function G(s) = n(s)/d(s) is deg d - deg n. For a

proper transfer function, the relative degree is a nonnegative integer.

3 The proof is given in [30].
4 Equivalently, G(s) + GT (−s) has a normal rank m over the field of rational functions of s.
5.3. POSITIVE REAL TRANSFER FUNCTIONS 113

Example 5.6 The transfer function G(s) = 1/s is positive real since it has no poles
in Re[s] > 0, has a simple pole at s = 0 whose residue is 1, and Re[G(jω)] = 0,
∀ ω = 0. It is not strictly positive real since 1/(s − ε) has a pole in Re[s] > 0
for any ε > 0. The transfer function G(s) = 1/(s + a) with a > 0 is positive real,
since it has no poles in Re[s] ≥ 0 and Re[G(jω)] = a/(ω 2 + a2 ) > 0, ∀ ω ∈ [0, ∞)
Since this is so for every a > 0, we see that for any ε ∈ (0, a) the transfer function
G(s − ε) = 1/(s + a − ε) will be positive real. Hence, G(s) = 1/(s + a) is strictly
positive real. The same conclusion can be drawn from Lemma 5.1 by noting that

ω2 a
lim ω 2 Re[G(jω)] = lim =a>0
ω→∞ ω→∞ ω 2 + a2
The transfer function G(s) = 1/(s2 + s + 1) is not positive real because its relative
degree is two. We can see it also by calculating

1 − ω2
Re[G(jω)] = < 0, ∀ ω > 1
(1 − ω 2 )2 + ω 2
Consider the 2 × 2 transfer function matrix

1 s+1 1
G(s) =
s+1 −1 2s + 1

Since det[G(s) + GT (−s)] is not identically zero, we can apply Lemma 5.1. Noting
that G(∞) + GT (∞) is positive definite and
2

2 ω +1 −jω
G(jω) + GT (−jω) = 2
ω +1 jω 2ω 2 + 1

is positive definite for all ω ∈ R, we can conclude that G(s) is strictly positive real.
Finally, the 2 × 2 transfer function matrix
⎡ s+2 ⎤
s+1
1
s+2

G(s) = ⎣ ⎦ has G(∞) + GT (∞) = 2 0

−1 2 0 0
s+2 s+1

It can be verified that

⎡ ⎤
2(2+ω 2 ) −2jω
⎢ 1+ω 2 4+ω 2 ⎥
G(jω) + GT (−jω) = ⎣ ⎦
2jω 4
4+ω 2 1+ω 2

is positive definite for all ω ∈ R and limω→∞ ω 2 det[G(jω) + GT (−jω)] = 4. Con-

sequently, by Lemma 5.1, we conclude that G(s) is strictly positive real. 
Passivity properties of positive real transfer functions can be shown by using the
next two lemmas, which are known, respectively, as the positive real lemma and the
114 CHAPTER 5. PASSIVITY

Kalman–Yakubovich–Popov lemma.5 The lemmas give algebraic characterization of

positive real and strictly positive real transfer functions.

Lemma 5.2 (Positive Real) Let G(s) = C(sI −A)−1 B +D be an m×m transfer
function matrix where (A, B) is controllable and (A, C) is observable. Then, G(s)
is positive real if and only if there exist matrices P = P T > 0, L, and W such that

P A + AT P = −LT L (5.10)
PB = C T − LT W (5.11)
WTW = D + DT (5.12)

Lemma 5.3 (Kalman–Yakubovich–Popov) Let G(s) = C(sI − A)−1 B + D be

an m × m transfer function matrix, where (A, B) is controllable and (A, C) is ob-
servable. Then, G(s) is strictly positive real if and only if there exist matrices
P = P T > 0, L, and W , and a positive constant ε such that

P A + AT P = −LT L − εP (5.13)
PB = C −L W
T T
(5.14)
T
W W = D + DT (5.15)

Proof of Lemma 5.3: Suppose there exist P = P T > 0, L, W , and ε > 0 that
satisfy (5.13) through (5.15). Set μ = 12 ε and recall that G(s − μ) = C(sI − μI −
A)−1 B + D. From (5.13), we have

P (A + μI) + (A + μI)T P = −LT L (5.16)

It follows from Lemma 5.2 that G(s − μ) is positive real. Hence, G(s) is strictly
positive real. On the other hand, suppose G(s) is strictly positive real. There exists
μ > 0 such that G(s − μ) is positive real. It follows from Lemma 5.2 that there
are matrices P = P T > 0, L, and W , which satisfy (5.14) through (5.16). Setting
ε = 2μ shows that P , L, W , and ε satisfy (5.13) through (5.15). 2

Lemma 5.4 The linear time-invariant minimal realization

ẋ = Ax + Bu, y = Cx + Du

with G(s) = C(sI − A)−1 B + D is

• passive if G(s) is positive real;
• strictly passive if G(s) is strictly positive real.
5 The proof of Lemma 5.2 is given in [74, Lemma 6.2].
5.4. CONNECTION WITH STABILITY 115

3
1 T
Proof: Apply Lemmas 5.2 and 5.3, respectively, and use V (x) = 2x Px as the
storage function.
∂V
uT y − (Ax + Bu) = uT (Cx + Du) − xT P (Ax + Bu)
∂x
= uT Cx + 12 uT (D + DT )u − 12 xT (P A + AT P )x − xT P Bu
= uT (B T P + W T L)x + 12 uT W T W u
+ 12 xT LT Lx + 12 εxT P x − xT P Bu
= 1
2 (Lx + W u)T (Lx + W u) + 12 εxT P x ≥ 1 T
2 εx P x

In the case of Lemma 5.2, ε = 0 and we conclude that the system is passive, while in
the case of Lemma 5.3, ε > 0 and we conclude that the system is strictly passive. 2

5.4 Connection with Stability

In this section we consider a passive system of the form
ẋ = f (x, u), y = h(x, u) (5.17)
where f is locally Lipschitz, h is continuous, f (0, 0) = 0, and h(0, 0) = 0, and study
the stability of the origin as an equilibrium point of ẋ = f (x, 0).
Lemma 5.5 If the system (5.17) is passive with a positive definite storage function
V (x), then the origin of ẋ = f (x, 0) is stable.
Proof: Take V as a Lyapunov function candidate for ẋ = f (x, 0). Then V̇ ≤ 0. 2

To show asymptotic stability of the origin of ẋ = f (x, 0), we need to either show
that V̇ is negative definite or apply the invariance principle. In the next lemma, we
apply the invariance principle by considering a case where V̇ = 0 when y = 0 and
then require the additional property that
y(t) ≡ 0 ⇒ x(t) ≡ 0 (5.18)
for all solutions of (5.17) when u = 0. Equivalently, no solutions of ẋ = f (x, 0) can
stay identically in S = {h(x, 0) = 0}, other than the zero solution x(t) ≡ 0. The
property (5.18) can be interpreted as an observability condition. Recall that for the
linear system
ẋ = Ax, y = Cx
observability is equivalent to
y(t) = CeAt x(0) ≡ 0 ⇔ x(0) = 0 ⇔ x(t) ≡ 0
For easy reference, we define (5.18) as an observability property of the system.
116 CHAPTER 5. PASSIVITY

Definition 5.5 The system (5.17) is said to be zero-state observable if no solution

of ẋ = f (x, 0) can stay identically in S = {h(x, 0) = 0}, other than the zero solution
x(t) ≡ 0.

Lemma 5.6 Consider the system (5.17). The origin of ẋ = f (x, 0) is asymptoti-
cally stable if the system is
• strictly passive or
• output strictly passive and zero-state observable.
Furthermore, if the storage function is radially unbounded, the origin will be globally
asymptotically stable. 3

Proof: Suppose the system is strictly passive and let V (x) be its storage function.
Then, with u = 0, V̇ satisfies the inequality V̇ ≤ −ψ(x), where ψ(x) is positive
definite. We can use this inequality to show that V (x) is positive definite. In
particular, for any x ∈ Rn , the equation ẋ = f (x, 0) has a solution φ(t; x), starting
from x at t = 0 and defined on some interval [0, δ]. Integrating the inequality
V̇ ≤ −ψ(x) yields
 τ
V (φ(τ, x)) − V (x) ≤ − ψ(φ(t; x)) dt, ∀ τ ∈ [0, δ]
0

Using V (φ(τ, x)) ≥ 0, we obtain

 τ
V (x) ≥ ψ(φ(t; x)) dt
0

Suppose there is x̄ = 0 such that V (x̄) = 0. The foregoing inequality implies

 τ
ψ(φ(t; x̄)) dt = 0, ∀ τ ∈ [0, δ] ⇒ ψ(φ(t; x̄)) ≡ 0 ⇒ φ(t; x̄) ≡ 0 ⇒ x̄ = 0
0

which contradicts the claim that x̄ = 0. Thus, V (x) > 0 for all x = 0. This qualifies
V (x) as a Lyapunov function candidate, and since V̇ (x) ≤ −ψ(x), we conclude that
the origin is asymptotically stable.
Suppose now that the system is output strictly passive and let V (x) be its
storage function. Then, with u = 0, V̇ satisfies the inequality V̇ ≤ −y T ρ(y), where
y T ρ(y) > 0 for all y = 0. By repeating the preceding argument, we can use this
inequality to show that V (x) is positive definite. In particular, if there is x̄ = 0
such that V (x̄) = 0, then
 τ
0 = V (x̄) ≥ hT (φ(t; x̄), 0)ρ(h(φ(t; x̄), 0)) dt, ∀ τ ∈ [0, δ] ⇒ h(φ(t; x̄), 0) ≡ 0
0

Due to zero-state observability,

h(φ(t; x̄), 0) ≡ 0 ⇒ φ(t; x̄) ≡ 0 ⇒ x̄ = 0

5.4. CONNECTION WITH STABILITY 117

Hence, V (x) > 0 for all x = 0. Since V̇ (x) ≤ −y T ρ(y) and y(t) ≡ 0 ⇒ x(t) ≡ 0,
we conclude by the invariance principle that the origin is asymptotically stable.
Finally, if V (x) is radially unbounded, we can infer global asymptotic stability from
Theorems 3.3 and 3.5, respectively. 2

Example 5.7 Consider the m-input–m-output system

ẋ = f (x) + G(x)u, y = h(x)

where f and G are locally Lipschitz, h is continuous, f (0) = 0, and h(0) = 0.

Suppose there is a continuously differentiable positive semidefinite function V (x)
such that
∂V ∂V
f (x) ≤ 0, G(x) = hT (x)
∂x ∂x
Then,
∂V ∂V ∂V
uT y − [f (x) + G(x)u] = uT h(x) − f (x) − hT (x)u = − f (x) ≥ 0
∂x ∂x ∂x
which shows that the system is passive. If V (x) is positive definite, we can conclude
that the origin of ẋ = f (x) is stable. If we have the stronger condition
∂V ∂V
f (x) ≤ −khT (x)h(x), G(x) = hT (x)
∂x ∂x
for some k > 0, then
∂V
uT y − [f (x) + G(x)u] ≥ ky T y
∂x
and the system is output strictly passive with ρ(y) = ky. If, in addition, it is zero-
state observable, then the origin of ẋ = f (x) is asymptotically stable. Furthermore,
if V (x) is radially unbounded, the origin will be globally asymptotically stable. 

Example 5.8 Consider the single-input–single-output system6

ẋ1 = x2 , ẋ2 = −ax31 − kx2 + u, y = x2

where a and k are positive constants. Let V (x) = 14 ax41 + 12 x22 be a storage function
candidate.
V̇ = ax31 x2 + x2 (−ax31 − kx2 + u) = −ky 2 + yu
Therefore, the system is output strictly passive. Moreover, when u = 0,

y(t) ≡ 0 ⇔ x2 (t) ≡ 0 ⇒ ax31 (t) ≡ 0 ⇒ x1 (t) ≡ 0

Hence, the system is zero-state observable. It follows from Lemma 5.6 that the
origin of the unforced system is globally asymptotically stable. 
6 Lyapunov stability of this system was studied in Examples 3.8 and 3.9.
118 CHAPTER 5. PASSIVITY

5.5 Exercises
5.1 Show that the parallel connection of two passive (respectively, input strictly
passive, output strictly passive, strictly passive) dynamical systems is passive (re-
spectively, input strictly passive, output strictly passive, strictly passive).

5.2 Consider the system

ẋ1 = −ax1 + x2 , ẋ2 = −h(x1 ) − bx2 + u, y = x2

where a ≥ 0, b ≥ 0, h is locally Lipschitz, h(0) = 0 and x1 h(x1 ) > 0 for all

x
x1 = 0. Using a storage function of the form V (x) = k1 0 1 h(σ) dσ + 12 k2 x22 , with
appropriately chosen positive constants k1 and k2 , show that the system is

(a) Passive.

(b) Lossless when a = b = 0.

(c) Output strictly passive when b > 0.

(d) Strictly passive when a > 0 and b > 0.

5.3 Consider the system

ẋ1 = x2 , ẋ2 = −g(x1 )(x1 + x2 ) + u, y = x1 + 2x2

 x and g(y) ≥ 1 for all y. Using the storage function

where gis locally Lipschitz
x
V (x) = 0 1 σg(σ) dσ + 0 2 (x1 + 2σ) dσ, show that the system is passive.

5.4 Consider the system

ẋ1 = x2 , ẋ2 = −h(x1 ) − ax2 + u, y = bx1 + cx2

where a, b, c are positive constants, h is locally Lipschitz, h(0) = 0 and x1 h(x1 ) > 0
x
for all x1 = 0. Using a storage function of the form V (x) = k 0 1 h(σ) dσ + 12 xT P x,
with appropriately chosen positive constant k and positive definite matrix P , show
that the system is strictly passive if b < ac.

5.5 Consider the system

ẋ1 = x2 , ẋ2 = −h1 (x1 ) − x2 − h2 (x3 ) + u, ẋ3 = x2 − x3 , y = x2

where h1 and h2 are locally Lipschitz functions that satisfy hi (0)

 x = 0 and yhi (y) > 0
for all y = 0. Using a storage function of the form V (x) = k1 0 1 h1 (σ) dσ + k2 x22 +
x
k3 0 3 h2 (σ) dσ, with appropriately chosen positive constants k1 to k3 , show that
the system is output strictly passive.
5.5. EXERCISES 119

5.6 Consider the system

M ẋ = −Kx + ah(x) + u, y = h(x)

where x, u, and y are n-dimensional vectors, M and Kare positive definite sym-
x
metric matrices, h is locally Lipschitz, h ∈ [0, K], and 0 hT (σ)M dσ ≥ 0 for all
x. Show that the system is passive when a = 1 and output strictly passive when
a < 1.

5.7 For each of the following systems, show the stated passivity property for the
given output.
(1) The mass-spring system (A.6) with y = x2 is output strictly passive.
(2) The tunnel-diode circuit (A.7) with y = x2 is strictly passive.
(3) The boost converter (A.16) with y = k(x1 − αkx2 ) is passive.

5.8 An m-link robot manipulator is represented by equation (A.34). Suppose

g(q) = 0 has an isolated solution q = q ∗ and the potential energy P (q) has a
minimum at q ∗ ; that is, P (q) ≥ P (q ∗ ) for all q.
(a) Using V = 12 q̇ T M (q)q̇ + P (q) − P (q ∗ ) as a storage function candidate, show
that the system, with input u and output q̇, is passive.
(b) Let u = −Kd q̇ + v, where Kd is a positive definite symmetric matrix. Show
that the system, with input v and output q̇, is output strictly passive.

5.9 ([105]) Euler equations for a rotating rigid spacecraft are given by

J1 ω̇1 = (J2 −J3 )ω2 ω3 +u1 , J2 ω̇2 = (J3 −J1 )ω3 ω1 +u2 , J3 ω̇3 = (J1 −J2 )ω1 ω2 +u3

where ω = col(ω1 , ω2 , ω3 ) is the angular velocity vector along the principal axes,
u = col(u1 , u2 , u3 ) is the vector of torque inputs applied about the principal axes,
and J1 to J3 are the principal moments of inertia.
(a) Show that the system, with input u and output ω, is lossless.
(b) Let u = −Kω + v, where K is a positive definite symmetric matrix. Show that
the system, with input v and output ω, is strictly passive.

5.10 Consider the TORA system of Section A.12.

(a) Using the sum of the potential energy 12 kx2c and kinetic energy 12 v T Dv, where

θ̇ J + mL2 mL cos θ
v= and D =
ẋc mL cos θ m + M

as the storage function, show that the system with input u and output θ̇ is
passive.
120 CHAPTER 5. PASSIVITY

(b) Let u = −φ1 (θ) + w where φ1 is locally Lipschitz, φ1 (0) = 0 and yφ1 (y) > 0
for all y = 0. Using
 θ
1 T 1
V = v Dv + kx2c + φ1 (λ) dλ
2 2 0

as the storage function, show that the system with input w and output θ̇ is
passive.

5.11 Show that the transfer function (b1 s + b2 )/(s2 + a1 s + a2 ) is strictly positive
real if and only if all coefficients are positive and b2 < a1 b1 .

5.12 Show that the transfer function d + b1 s/(s2 + a1 s + a2 ) is strictly positive

real if a1 , a2 , d, and a1 d + b1 are positive.

5.13 Consider the system

0 1 0  
ẋ = x+ u, y= 1 2 x
−1 −1 1

(a) Find the transfer function and show that it is strictly positive real.

(b) Choose ε such that A + (ε/2)I is Hurwitz and solve equations (5.13) through
(5.15) for P , L and W .

5.14 Consider equations (5.13) through (5.15) and suppose that (D + DT ) is non-
singular. Choose ε such that A + (ε/2)I is Hurwitz. Show that P satisfies the
Riccati equation7
P A0 + AT0 P − P B0 P + C0 = 0
where A0 = −(/2)I − A + B(D + DT )−1 C, B0 = B(D + DT )−1 B T , and C0 =
−C T (D + DT )−1 C, and L is given by

LT L = (C T − P B)(D + DT )−1 (C − B T P )

5.15 Investigate zero-state observability of the systems in the following exercises:

5.2 5.3 5.4 5.5

5.6 5.7(1) 5.7(2) 5.7(3)
5.8(a) 5.8(b) 5.10(a) 5.10(b)

7 The Riccati equation can be solved by the MATLAB command “are(A0 , B0 , C0 )”.
Chapter 6

Input-Output Stability

input-output stability In most of this book, we use the state-space approach to

model nonlinear dynamical systems and place a lot of emphasis on the behavior
of the state variables. An alternative approach to the mathematical modeling of
dynamical systems is the input-output approach. An input-output model relates
the output of the system directly to the input, with no knowledge of the internal
structure that is represented by the state equation. The system is viewed as a black
box that can be accessed only through its input and output terminals. The input-
output approach can handle systems that are not representable by state models
such as systems with time delay, which cannot be represented by state models unless
the system is approximated. This chapter gives a brief introduction to the notion
of input-output stability and how to show it for state models.1 In Section 6.1,
we introduce input-output models and define L stability, a notion of stability in
the input-output sense. In Section 6.2, we study L stability of nonlinear systems
represented by state models. Finally, in Section 6.3 we discuss the calculation of
the L2 gain for a special class of nonlinear systems.

6.1 L Stability
We consider a system whose input–output relation is represented by

y = Hu

where H is some mapping or operator that specifies y in terms of u. The input u

belongs to a space of signals that map the time interval [0, ∞) into the Euclidean
space Rm ; that is, u : [0, ∞) → Rm . Examples are the space of piecewise con-
tinuous, bounded functions; that is, supt≥0 u(t) < ∞, and the space of piecewise
1 For comprehensive treatment of input-output stability, the reader may consult [35, 116, 147].

121
122 CHAPTER 6. INPUT-OUTPUT STABILITY

∞
continuous, square-integrable functions; that is, 0 uT (t)u(t) dt < ∞. To mea-
sure the size of a signal, we introduce the norm function u , which satisfies three
properties:
• The norm of a signal is zero if and only if the signal is identically zero and is
strictly positive otherwise.
• Scaling a signal results in a corresponding scaling of the norm; that is, au =
a u for any positive constant a and every signal u.
• The norm satisfies the triangle inequality u1 + u2 ≤ u1 + u2 for any
signals u1 and u2 .
For the space of piecewise continuous, bounded functions, the norm is defined as

u L∞ = sup u(t) < ∞

t≥0

and the space is denoted by Lm ∞ . For the space of piecewise continuous, square-
integrable functions, the norm is defined by
 ∞
u L2 = uT (t)u(t) dt < ∞
0

and the space is denoted by Lm 2 . More generally, the space Lp for 1 ≤ p < ∞ is
m

defined as the set of all piecewise continuous functions u : [0, ∞) → Rm such that
 ∞ 1/p
u Lp = u(t) p dt <∞
0

The subscript p in Lm p refers to the type of p-norm used to define the space, while
the superscript m is the dimension of the signal u. If they are clear from the context,
we may drop one or both of them and refer to the space simply as Lp , Lm , or L.
To distinguish the norm of u as a vector in the space L from the norm of u(t) as a
vector in Rm , we write the first norm as · L .
If we think of u ∈ Lm as a “well-behaved” input, the question to ask is whether
the output y will be “well behaved” in the sense that y ∈ Lq , where Lq is the same
space as Lm , except that q (the number of output variables) could be different
from m (the number of input variables). A system having the property that any
“well-behaved” input will generate a “well-behaved” output will be defined as a
stable system. However, we cannot define H as a mapping from Lm to Lq , because
we have to deal with systems which are unstable, in that an input u ∈ Lm may
generate an output y that does not belong to Lq . Therefore, H is usually defined
as a mapping from an extended space Lm e to an extended space Le , where Le is
q m

defined by
e = {u | uτ ∈ L , ∀ τ ∈ [0, ∞)}
Lm m
6.1. L STABILITY 123

and uτ is a truncation of u defined by


u(t), 0≤t≤τ
uτ (t) =
0, t>τ

The extended space Lm e is a linear space that contains the unextended space L
m
as
a subset. It allows us to deal with unbounded “ever-growing” signals. For example,
the signal u(t) = t does not belong to the space L∞ , but its truncation

t, 0 ≤ t ≤ τ
uτ (t) =
0, t>τ

belongs to L∞ for every finite τ . Hence, u(t) = t belongs to the extended space
L∞e .
A mapping H : Lme → Le is said to be causal if the value of the output (Hu)(t)
q

at any time t depends only on the values of the input up to time t. This is equivalent
to
(Hu)τ = (Huτ )τ
Causality is an intrinsic property of dynamical systems represented by state models.
With the space of input and output signals defined, we proceed to define input–
output stability. We start by defining the concept of a gain function.
Definition 6.1 A scalar continuous function g(r), defined for r ∈ [0, a), is a gain
function if it is nondecreasing and g(0) = 0.
Note that a class K function is a gain function but not the other way around.
By not requiring the gain function to be strictly increasing we can have g = 0 or
g(r) = sat(r).
Definition 6.2 A mapping H : Lm e → Le is L stable if there exist a gain function
q

g, defined on [0, ∞), and a nonnegative constant β such that

(Hu)τ L ≤ g ( uτ L ) + β (6.1)

e and τ ∈ [0, ∞). It is finite-gain L stable if there exist nonnegative

for all u ∈ Lm
constants γ and β such that

(Hu)τ L ≤ γ uτ L + β (6.2)

e and τ ∈ [0, ∞).

for all u ∈ Lm
The constant β in (6.1) or (6.2) is called the bias term. It is included in the definition
to allow for systems where Hu does not vanish at u = 0. When inequality (6.2) is
satisfied, we say that the system has L gain ≤ γ.
For causal, L stable systems, it can be shown by a simple argument that

u ∈ Lm ⇒ Hu ∈ Lq
124 CHAPTER 6. INPUT-OUTPUT STABILITY

and
Hu L ≤ g ( u L ) + β, ∀ u ∈ Lm
For causal, finite-gain L stable systems, the foregoing inequality takes the form

Hu L ≤ γ u L + β, ∀ u ∈ Lm

The definition of L∞ stability is the familiar notion of bounded-input–bounded-

output stability; namely, if the system is L∞ stable, then for every bounded input
u(t), the output Hu(t) is bounded.

Example 6.1 a real-valued function of a real argument h(u) can be viewed as an

operator H that assigns to every input signal u(t) the output signal y(t) = h(u(t)).
We use this simple operator to illustrate the definition of L stability. Suppose

|h(u)| ≤ a + b|u|, ∀ u ∈ R

Then, H is finite-gain L∞ stable with γ = b and β = a. If a = 0, then for each

p ∈ [1, ∞),  ∞  ∞
|h(u(t))|p dt ≤ (b)p |u(t)|p dt
0 0

Thus, for each p ∈ [1, ∞], the operator H is finite-gain Lp stable with zero bias and
γ = b. Finally, for h(u) = u2 , H is L∞ stable with zero bias and g(r) = r2 . It is
not finite-gain L∞ stable because |h(u)| = u2 cannot be bounded γ|u| + β for all
u ∈ R. 

Example 6.2 Consider a single-input–single-output system defined by the causal

convolution operator
 t
y(t) = h(t − σ)u(σ) dσ
0

where h(t) = 0 for t < 0. Suppose h ∈ L1e ; that is, for every τ ∈ [0, ∞),
 ∞  τ
hτ L1 = |hτ (σ)| dσ = |h(σ)| dσ < ∞
0 0

If u ∈ L∞e and τ ≥ t, then

 t
|y(t)| ≤ |h(t − σ)| |u(σ)| dσ
0
 t  t
≤ |h(t − σ)| dσ sup |u(σ)| = |h(s)| ds sup |u(σ)|
0 0≤σ≤τ 0 0≤σ≤τ

Consequently,
yτ L∞ ≤ hτ L1 uτ L∞ , ∀ τ ∈ [0, ∞)
6.1. L STABILITY 125

This inequality resembles (6.2), but it is not the same as (6.2) because the constant
γ in (6.2) is required to be independent of τ . While hτ L1 is finite for every finite τ ,
it may not be bounded uniformly in τ . For example, h(t) = et has hτ L1 = (eτ −1),
which is finite for all τ ∈ [0, ∞) but not uniformly bounded in τ . Inequality (6.2)
will be satisfied if h ∈ L1 ; that is,
 ∞
h L1 = |h(σ)| dσ < ∞
0

Then, the inequality

yτ L∞ ≤ h L1 uτ L∞ , ∀ τ ∈ [0, ∞)

shows that the system is finite-gain L∞ stable. The condition h L1 < ∞ actually
guarantees finite-gain Lp stability for each p ∈ [1, ∞]. Consider first the case p = 1.
For t ≤ τ < ∞, we have
 τ  τ  t   τ t
 
|y(t)| dt =  
h(t − σ)u(σ) dσ  dt ≤ |h(t − σ)| |u(σ)| dσ dt

0 0 0 0 0

Reversing the order of integration yields

 τ  τ  τ  τ
|y(t)| dt ≤ |u(σ)| |h(t − σ)| dt dσ ≤ |u(σ)| h L1 dσ ≤ h L1 uτ L1
0 0 σ 0

Thus,
yτ L1 ≤ h L1 uτ L1 , ∀ τ ∈ [0, ∞)
Consider now the case p ∈ (1, ∞) and let q ∈ (1, ∞) be defined by 1/p + 1/q = 1.
For t ≤ τ < ∞, we have
 t  t
|y(t)| ≤ |h(t − σ)| |u(σ)| dσ = |h(t − σ)|1/q |h(t − σ)|1/p |u(σ)| dσ
0 0

By Hölder’s inequality,2
 t 1/q  t 1/p
|y(t)| ≤ |h(t − σ)| dσ |h(t − σ)| |u(σ)| dσ p
0 0
 t 1/p
1/q
≤ ( hτ L1 ) |h(t − σ)| |u(σ)| dσ p
0
2 Hölder’s inequality states that if f ∈ L
pe and g ∈ Lqe , where p ∈ (1, ∞) and 1/p + 1/q = 1,
then  τ  τ 1/p  τ 1/q
|f (t)g(t)| dt ≤ |f (t)|p dt |g(t)|q dt
0 0 0
for every τ ∈ [0, ∞). (See [7].)
126 CHAPTER 6. INPUT-OUTPUT STABILITY

Thus,

 p τ
yτ Lp = |y(t)|p dt
0
 τ  t 
p/q
≤ ( hτ L1 ) |h(t − σ)| |u(σ)|p dσ dt
0 0
 τ  t
p/q
= ( hτ L1 ) |h(t − σ)| |u(σ)|p dσ dt
0 0

By reversing the order of integration, we obtain

 τ  τ
 p p/q
yτ Lp ≤ ( hτ L1 ) |u(σ)|p |h(t − σ)| dt dσ
0 σ
p/q  p p  p
≤ ( hτ L1 ) hτ L1 uτ Lp = ( hτ L1 ) uτ Lp

Hence,
yτ Lp ≤ h L1 uτ Lp
In summary, if h L1 < ∞, then for each p ∈ [1, ∞], the causal convolution operator
is finite-gain Lp stable and (6.2) is satisfied with γ = h L1 and β = 0. 

One drawback of Definition 6.2 is the implicit requirement that inequality (6.1)
or (6.2) be satisfied for all signals in the input space Lm . This excludes systems
where the input–output relation may be defined only for a subset of the input space.
The next example explores this point and motivates the definition of small-signal
L stability that follows the example.
Example 6.3 Consider a single-input–single-output system defined by y = tan u.
The output y(t) is defined only when the input signal is restricted to |u(t)| < π/2
for all t ≥ 0. Thus, the system is not L∞ stable in the sense of Definition 6.2.
However, if we restrict u(t) to the set |u| ≤ r < π/2, then
 
tan r
|y| ≤ |u|
r
and the system will satisfy the inequality
 
tan r
y Lp ≤ u Lp
r
for every u ∈ Lp such that |u(t)| ≤ r for all t ≥ 0, where p could be any number
in [1, ∞]. In the space L∞ , the requirement |u(t)| ≤ r implies that u L∞ ≤ r,
showing that the foregoing inequality holds only for input signals of small norm.
However, for other Lp spaces with p < ∞ the instantaneous bound on |u(t)| does
not necessarily restrict the norm of the input signal. For example, the signal

u(t) = re−rt/a , a > 0

6.2. L STABILITY OF STATE MODELS 127

which belongs to Lp for each p ∈ [1, ∞], satisfies the instantaneous bound |u(t)| ≤ r
while its Lp norm
 1/p
a
u Lp = r , 1≤p<∞
rp
can be arbitrarily large. 

Definition 6.3 A mapping H : Lm e → Le is small-signal L stable (respectively,

q

small-signal finite-gain L stable ) if there is a positive constant r such that inequality

(6.1) [respectively, (6.2)] is satisfied for all u ∈ Lme with sup0≤t≤τ u(t) ≤ r.

6.2 L Stability of State Models

The notion of input–output stability is intuitively appealing. This is probably
why most of us were introduced to dynamical system stability in the framework
of bounded-input–bounded-output stability. Since, in Lyapunov stability, we put a
lot of emphasis on studying the stability of equilibrium points and the asymptotic
behavior of state variables, one may wonder: What can we see about input–output
stability starting from the formalism of Lyapunov stability? In this section, we
show how Lyapunov stability tools can be used to establish L stability of nonlinear
systems represented by state models.
Consider the system

ẋ = f (x, u), y = h(x, u) (6.3)

where x(0) = x0 , f is locally Lipschitz and h is continuous for all x ∈ D and

u ∈ Du , in which D ⊂ Rn is a domain that contains x = 0, and Du ⊂ Rm is a
domain that contains u = 0. The dimension q of the output y could be different
from the dimension m of the input u. For each fixed x0 ∈ D, the state model
(6.3) defines an operator H that assigns to each input signal u(t) the corresponding
output signal y(t). Suppose x = 0 is an equilibrium point of the unforced system

ẋ = f (x, 0) (6.4)

The theme of this section is that if the origin of (6.4) is asymptotically (or expo-
nentially) stable, then, under some assumptions on f and h, the system (6.3) will
be L stable or small-signal L stable for a certain signal space L. We pursue this
idea first in the case of exponentially stability, and then for the more general case
of asymptotic stability.

Theorem 6.1 Consider the system (6.3) and take r > 0 and ru > 0 such that
{ x ≤ r} ⊂ D and { u ≤ ru } ⊂ Du . Suppose the following assumptions hold for
all x ∈ D and u ∈ Du :
128 CHAPTER 6. INPUT-OUTPUT STABILITY

• x = 0 is an exponentially stable equilibrium point of ẋ = f (x, 0), and there is

a Lyapunov function V(x) that satisfies

∂V ∂V
c1 x 2 ≤ V (x) ≤ c2 x 2 , f (x, 0) ≤ −c3 x 2 ,
∂x ≤ c4 x (6.5)
∂x
for some positive constants c1 to c4 .
• f and h satisfy the inequalities
f (x, u) − f (x, 0) ≤ L u (6.6)
h(x, u) ≤ η1 x + η2 u (6.7)
for some nonnegative constants L, η1 , and η2 .

Then, for each x0 with x0 ≤ r c1 /c2 , the system (6.3) is small-signal finite-gain
Lp stable for each p ∈ [1, ∞]. In particular, for each u ∈ Lpe with sup0≤t≤τ u(t) ≤
min{ru , c1 c3 r/(c2 c4 L)}, the output y(t) satisfies
yτ Lp ≤ γ uτ Lp + β (6.8)
for all τ ∈ [0, ∞), with
⎧
 ⎪
⎨ 1, if p = ∞
η1 c2 c4 L c2
γ = η2 + , β = η1 ρ x0 , where ρ =  1/p
c1 c3 c1 ⎪
⎩ 2c2
c3 p , if p ∈ [1, ∞)

Furthermore, if the origin is globally exponentially stable and the assumptions hold
globally (with D = Rn and Du = Rm ), then, for each x0 ∈ Rn , the system (6.3) is
finite-gain Lp stable for each p ∈ [1, ∞] and (6.8) is satisfied for each u ∈ Lpe . 3
Proof: The derivative of V along the trajectories of ẋ = f (x, u) satisfies
∂V ∂V
V̇ = f (x, 0) + [f (x, u) − f (x, 0)]
∂x ∂x
c3 c4 L √
≤ −c3 x 2 + c4 L x u ≤ − V + √ u V
c2 c1

Let W (x) = V (x). When V (x(t)) = 0, Ẇ is given by

V̇ c3 c4 L
Ẇ = √ ≤ − W + √ u (6.9)
2 V 2c2 2 c1
When V (x(t)) = 0, W has a directional derivative that satisfies3
1 c4 L
Ẇ = lim [W (x(t + δ)) − W (x(t))] ≤ √ u
δ→0 δ 2 c1
3 See Exercises 3.24 and 5.6 of [74] for how to show this inequality.
6.2. L STABILITY OF STATE MODELS 129

Thus, inequality (6.9) holds all the time. Rewrite it as

Ẇ ≤ −aW + b u
√
where a = c3 /(2c2 ) and b = c4 L/(2 c1 ). Set U (t) = eat W (x(t)). Then,

U̇ = eat Ẇ + aeat W ≤ −aeat W + beat u + aeat W = beat u

By integration, we obtain
 t
U (t) ≤ U (0) + beaτ u(τ ) dτ
0

Hence,  t
−at
W (x(t)) ≤ e W (x(0)) + e−a(t−τ ) b u(τ ) dτ
0
Recalling that
√ √
c1 x ≤ W (x) ≤ c2 x
and substituting for b we arrive at
 
c2 c4 L t −a(t−τ )
x(t) ≤ x0 e−at + e u(τ ) dτ (6.10)
c1 2c1 0
It can be verified that

c1 c1 c3 r
x0 ≤ r and sup u(σ) ≤
c2 0≤σ≤t c2 c4 L
ensure that x(t) ≤ r; hence, x(t) stays within the domain of validity of the
assumptions. Using (6.7), we have
 t
y(t) ≤ k1 e−at + k2 e−a(t−τ ) u(τ ) dτ + k3 u(t)
0

where 
c2 c4 Lη1
k1 = x0 η1 , k2 = , k3 = η2
c1 2c1
Set
 t
y1 (t) = k1 e−at , y2 (t) = k2 e−a(t−τ ) u(τ ) dτ, y3 (t) = k3 u(t)
0

Suppose now that u ∈ Lm pe for some p ∈ [1, ∞]. It is clear that y3τ Lp ≤ k3 uτ Lp
and using the results of Example 6.2, it can be verified that y2τ Lp ≤ (k2 /a) uτ Lp .
As for the first term, y1 (t), it can be verified that
⎧
⎪
⎨ 1, if p = ∞
y1τ Lp ≤ k1 ρ, where ρ =  1/p
⎪
⎩ 1
ap , if p ∈ [1, ∞)
130 CHAPTER 6. INPUT-OUTPUT STABILITY

Thus, by the triangle inequality, (6.8) is satisfied with

k2
γ = k3 + , β = k1 ρ
a
When all the assumptions hold globally, there is no need to restrict x0 or u(t) .
Therefore, (6.8) is satisfied for each x0 ∈ Rn and u ∈ Lpe . 2

The use of (the converse Lyapunov) Theorem 3.8 shows the existence of a Lya-
punov function satisfying (6.5). Consequently, we have the following corollary.

Corollary 6.1 Suppose that, in some neighborhood of (x = 0, u = 0), f is contin-

uously differentiable and h satisfies (6.7). If the origin x = 0 is an exponentially
stable equilibrium point of ẋ = f (x, 0), then there is a constant r0 > 0 such that
for each x0 with x0 < r0 , the system (6.3) is small-signal finite-gain Lp stable
for each p ∈ [1, ∞]. Furthermore, if all the assumptions hold globally, the origin is
globally exponentially stable, and f is globally Lipschitz, then for each x0 ∈ Rn , the
system (6.3) is finite-gain Lp stable for each p ∈ [1, ∞]. 3

Example 6.4 Consider the system

ẋ = −x − x3 + u, y = tanh x + u

with x(0) = x0 . The origin of ẋ = −x − x3 is globally exponentially stable, as

can be seen by the Lyapunov function V (x) = 12 x2 . The function V satisfies (6.5)
globally with c1 = c2 = 12 , c3 = c4 = 1. The functions f and h satisfy (6.6) and
(6.7) globally with L = η1 = η2 = 1. Hence, for each x0 ∈ R and each p ∈ [1, ∞],
the system is finite-gain Lp stable. 

Example 6.5 Consider the system

ẋ1 = x2 , ẋ2 = −x1 − x2 − a tanh x1 + u, y = x1

where a is a nonnegative constant. Use

V (x) = xT P x = p11 x21 + 2p12 x1 x2 + p22 x22

as a Lyapunov function candidate for the unforced system:

V̇ = −2p12 (x21 +ax1 tanh x1 )+2(p11 −p12 −p22 )x1 x2 −2ap22 x2 tanh x1 −2(p22 −p12 )x22

Choose p11 = p12 + p22 to cancel the cross-product term x1 x2 . Then, taking p22 =
2p12 = 1 makes P positive definite and results in

V̇ = −x21 − x22 − ax1 tanh x1 − 2ax2 tanh x1

6.2. L STABILITY OF STATE MODELS 131

Using the fact that x1 tanh x1 ≥ 0 and | tanh x1 | ≤ |x1 | for all x1 , we obtain

V̇ ≤ − x 2 + 2a|x1 | |x2 | ≤ −(1 − a) x 2

Thus, for all a < 1, V satisfies (6.5) globally with c1 = λmin (P ), c2 = λmax (P ),
c3 = 1 − a, and c4 = 2 P = 2λmax (P ). The functions f and h satisfy (6.6)
and (6.7) globally with L = η1 = 1, η2 = 0. Hence, for each x0 ∈ R2 and each
p ∈ [1, ∞], the system is finite-gain Lp stable and the Lp gain can be estimated by
γ = 2[λmax (P )]2 /[(1 − a)λmin (P )]. 

We turn now to the more general case when the origin of ẋ = f (x, 0) is asymp-
totically stable and restrict our attention to the study of L∞ stability. The next two
theorems give conditions for L∞ stability and small-signal L∞ stability, respectively.
Theorem 6.2 Consider the system (6.3) and suppose that, for all (x, u), f is locally
Lipschitz and h is continuous and satisfies the inequality

h(x, u) ≤ g1 ( x ) + g2 ( u ) + η (6.11)

for some gain functions g1 , g2 and a nonnegative constant η. If the system ẋ =

f (x, u) is input-to-state stable, then, for each x0 ∈ Rn , the system (6.3) is L∞
stable. 3
Proof: Input-to-state stability shows that x(t) satisfies the inequality
  
x(t) ≤ max β( x0 , t), γ sup u(σ) (6.12)
0≤σ≤t

for all t ≥ 0, where β and γ are class KL and class K functions, respectively. Using
(6.11), we obtain
   
y(t) ≤ g1 max β( x0 , t), γ sup u(σ) + g2 ( u(t) ) + η
0≤σ≤t
  
≤ g1 (β( x0 , t)) + g1 γ sup u(σ) + g2 ( u(t) ) + η
0≤σ≤t

where we used the property g1 (max{a, b}) ≤ g1 (a) + g1 (b), which holds for any gain
function. Thus,
yτ L∞ ≤ g ( uτ L∞ ) + β0 (6.13)
for all τ ∈ [0, ∞), where g = g1 ◦ γ + g2 and β0 = g1 (β( x0 , 0)) + η. 2

Theorem 6.3 Consider the system (6.3), where f is locally Lipschitz and h is
continuous in some neighborhood of (x = 0, u = 0). If the origin of ẋ = f (x, 0)
is asymptotically stable, then there is a constant k > 0 such that for any x0 with
x0 < k, the system (6.3) is small-signal L∞ stable. 3
132 CHAPTER 6. INPUT-OUTPUT STABILITY

Proof: By Lemma 4.7, the system (6.3) is locally input-to state stable. Hence,
there exist positive constants k and k1 such that inequality (6.12) is satisfied for
all x(0) ≤ k and u(t) ≤ k1 . By continuity of h, inequality (6.11) is satisfied in
some neighborhood of (x = 0, u = 0). The rest of the proof is the same as that of
Theorem 6.2. 2

Example 6.6 Consider the system

ẋ = −x − 2x3 + (1 + x2 )u2 , y = x2 + u

We saw in Example 4.13 that the state equation is input-to-state stable. The output
function h satisfies (6.11) globally with g1 (r) = r2 , g2 (r) = r, and η = 0. Thus, the
system is L∞ stable. 

Example 6.7 Consider the system

ẋ1 = −x31 + x2 , ẋ2 = −x1 − x32 + u, y = x1 + x2

Taking V = (x21 + x22 ), we have

V̇ = −2x41 − 2x42 + 2x2 u

Using x41 + x42 ≥ 12 x 4 yields

V̇ ≤ − x 4 + 2 x |u|
= −(1 − θ) x 4 − θ x 4 + 2 x |u|, 0 < θ < 1
 1/3
2|u|
≤ −(1 − θ) x , ∀ x ≥
4
θ

Thus, V satisfies inequalities (4.36) and (4.37) of Theorem 4.6 globally, with α1 (r) =
α2 (r) = r2 , W3 (x) = (1−θ) x 4 , and ρ(r) = (2r/θ)1/3 . Hence, the state equation is
input-to-state√stable. Furthermore, the function h = x1 +x2 satisfies (6.11) globally
with g1 (r) = 2r, g2 = 0, and η = 0. Thus, the system is L∞ stable. 

6.3 L2 Gain
L2 stability plays a special role in system analysis. It is natural to work with square-
integrable signals, which can be viewed as finite-energy signals.4 In many control
problems,5 the system is represented as an input–output map, from a disturbance
input u to a controlled output y, which is required to be small. With L2 input
signals, the control system is designed to make the input–output map finite-gain
4 If you think of u(t) as current or voltage, then uT (t)u(t) is proportional to the instantaneous

power of the signal, and its integral over time is a measure of the energy content of the signal.
5 See the literature on H
∞ control; for example, [14], [37], [41], [68], [143], or [154].
6.3. L2 GAIN 133

L2 stable and to minimize the L2 gain. In such problems, it is important not only
to be able to find out that the system is finite-gain L2 stable, but also to calculate
the L2 gain or an upper bound on it. In this section, we show how to calculate an
upper bound on the L2 gain for special cases. We start with linear systems.

Theorem 6.4 Consider the linear system

ẋ = Ax + Bu, y = Cx + Du

where A is Hurwitz. Let G(s) = C(sI − A)−1 B + D. Then, the L2 gain of the
system is less than or equal to supω∈R G(jω) .6 3

Proof: Due to linearity, we set x(0) = 0. From Fourier transform theory,7 we know
that for a causal signal u ∈ L2 , the Fourier transform U (jω) is given by
 ∞
U (jω) = u(t)e−jωt dt and Y (jω) = G(jω)U (jω)
0

Using Parseval’s theorem,8 we can write

 ∞  ∞
1
y L2 =
2 T
y (t)y(t) dt = Y ∗ (jω)Y (jω) dω
0 2π −∞
 ∞
1
= U ∗ (jω)GT (−jω)G(jω)U (jω) dω
2π −∞
 2  ∞
1
≤ sup G(jω) U ∗ (jω)U (jω) dω
ω∈R 2π −∞
 2
= sup G(jω) u 2L2
ω∈R

which shows that the L2 gain is less than or equal to supω∈R G(jω) . 2

It can be shown that supω∈R G(jω) is indeed the L2 gain and not just an
upper bound on it.9 For nonlinear systems, we have two results, Theorems 6.5 and
6.6, for estimating the L2 gain. Both theorems make use of the following lemma.
6
 This is the induced 2-norm of the complex matrix G(jω), which is equal to
λmax [GT (−jω)G(jω)] = σmax [G(jω)]. This quantity is known as the H∞ norm of G(jω) when
G(jω) is viewed as an element of the Hardy space H∞ . (See [41].)
7 See [35].
8 Parseval’s theorem [35] states that for a causal signal y ∈ L ,
2
 ∞  ∞
1
y T (t)y(t) dt = Y ∗ (jω)Y (jω) dω
0 2π −∞

9 See [74, Theorem 5.4].

134 CHAPTER 6. INPUT-OUTPUT STABILITY

Lemma 6.1 Consider the multiinput–multioutput system

ẋ = f (x, u), y = h(x, u) (6.14)
where x(0) = x0 , f (0, 0) = 0, h(0, 0) = 0, f is locally Lipschitz and h is continuous
for all x ∈ Rn and u ∈ Rm . Let V (x) be a continuously differentiable, positive
semidefinite function such that
∂V
f (x, u) ≤ k(γ 2 u 2 − y 2 )
V̇ = (6.15)
∂x
for all (x, u), for some positive constants k and γ. Then, for each x0 ∈ Rn , the
system is finite-gain L2 stable and its L2 gain is less than or equal to γ. 3
Proof: By integrating both sides of (6.15) over the interval [0, τ ], we obtain
 τ  τ
V (x(τ )) − V (x(0)) ≤ kγ 2
u(t) dt − k
2
y(t) 2 dt
0 0

Since V (x(τ )) ≥ 0,
 τ  τ
V (x(0))
y(t) 2 dt ≤ γ 2 u(t) 2 dt +
0 0 k
√
Taking the square roots and using the inequality a2 + b2 ≤ a + b for nonnegative
numbers a and b, we obtain

V (x(0))
yτ L2 ≤ γ uτ L2 + (6.16)
k
which completes the proof. 2

Theorem 6.5 Consider the m-input–m-output system

ẋ = f (x, u), y = h(x, u) (6.17)
where x(0) = x0 , f (0, 0) = 0, h(0, 0) = 0, f is locally Lipschitz, and h is continuous
for all x ∈ Rn and u ∈ Rm . If the system is output strictly passive with
uT y ≥ V̇ + δy T y, δ>0 (6.18)
then it is finite-gain L2 stable and its L2 gain is less than or equal to 1/δ. 3
Proof:
V̇ ≤ uT y − δy T y
1 1 δ
= − (u − δy)T (u − δy) + uT u − y T y
2δ  2δ 2
δ 1 T
≤ u u−y y T
2 δ2
Application of Lemma 6.1 completes the proof. 2
6.3. L2 GAIN 135

Theorem 6.6 Consider the multiinput–multioutput system

ẋ = f (x) + G(x)u, y = h(x) (6.19)
where x(0) = x0 , f (0) = 0, h(0) = 0, f and G are locally Lipschitz, and h is contin-
uous for all x ∈ Rn . Let γ be a positive number and suppose there is a continuously
differentiable, positive semidefinite function V (x) that satisfies the inequality
 T
∂V 1 ∂V ∂V 1
f (x) + 2 T
G(x)G (x) + hT (x)h(x) ≤ 0 (6.20)
∂x 2γ ∂x ∂x 2
for all x ∈ Rn . Then, for each x0 ∈ Rn , the system (6.19) is finite-gain L2 stable
and its L2 gain is less than or equal to γ. 3
Proof: By completing the squares, we have
 T
∂V ∂V 1 2 1 T ∂V 2 ∂V

f (x) + G(x)u = − γ u − 2 G (x) + f (x)
∂x ∂x 2 γ ∂x ∂x
 T
1 ∂V ∂V 1
+ 2 G(x)GT (x) + γ 2 u 2
2γ ∂x ∂x 2
Substituting (6.20) yields
 T
∂V ∂V 1 2 1 1 2
1 T ∂V 2

f (x) + G(x)u ≤ γ u − y − γ u − 2 G (x)
2 2

∂x ∂x 2 2 2 γ ∂x

Hence,  
V̇ ≤ 1
2 γ 2 u 2 − y 2
Application of Lemma 6.1 completes the proof. 2

Inequality (6.20) is known as the Hamilton–Jacobi inequality (or the Hamilton–

Jacobi equation when ≤ is replaced by =). The search for a function V (x) that
satisfies (6.20) requires basically the solution of a partial differential equation. Un-
like Theorem 6.1, the finite-gain L2 stability results of Theorems 6.5 and 6.6 do not
require the origin of the unforced system to be exponentially stable. This point is
illustrated by the next example.
Example 6.8 Consider the single-input–single-output system
ẋ1 = x2 , ẋ2 = −ax31 − kx2 + u, y = x2
where a and k are positive constants. The unforced system is a special case of
the system treated in Example 3.8, from which we obtain the Lyapunov function
V (x) = 14 ax41 + 12 x22 . With u = 0,

V̇ = ax31 x2 − ax31 x2 − kx22 + x2 u = −kx22 + x2 u

136 CHAPTER 6. INPUT-OUTPUT STABILITY

Hence
yu = V̇ + ky 2
and it follows from Theorem 6.5 that the system is finite-gain L2 stable and its
L2 gain is less than or equal to 1/k. The same conclusion can be arrived at by
application of Theorem 6.6 after verifying that V (x) = (1/k)( 14 ax41 + 12 x22 ) satisfies
the Hamilton–Jacobi inequality (6.20) with γ = 1/k. We note that the conditions
of Theorem 6.1 are not satisfied in this example because the origin of the unforced
system is not exponentially stable. Linearization at the origin yields the matrix

0 1
0 −k

which is not Hurwitz. 

Example 6.9 Consider the linear system

ẋ = Ax + Bu, y = Cx

Suppose P ≥ 0 is a solution of the Riccati equation

1
P A + AT P + P BB T P + C T C = 0 (6.21)
γ2

for some γ > 0. Taking V (x) = 12 xT P x and using the expression [∂V /∂x] = xT P ,
it can be easily seen that V (x) satisfies the Hamilton–Jacobi equation

1 T 1
xT P Ax + 2
x P B T BP x + xT C T Cx = 0
2γ 2

Hence, the system is finite-gain L2 stable and its L2 gain is ≤ γ. 

In Lemma 6.1, we assumed that inequality (6.15) holds globally. It is clear from
the proof that if the inequality holds only for x ∈ D ⊂ Rn and u ∈ Du ⊂ Rm ,
where D and Du are domains that contain x = 0 and u = 0, respectively, we will
still arrive at inequality (6.16) as long as x(t) and u(t) stay in D and Du for all
t ≥ 0. Ensuring that x(t) remains in some neighborhood of the origin, when both
x0 and sup0≤t≤τ u(t) are sufficiently small, follows from asymptotic stability of
the origin of ẋ = f (x, 0). This fact is used to show small-signal L2 stability in the
next lemma.

Lemma 6.2 Suppose the assumptions of Lemma 6.1 are satisfied for x ∈ D ⊂ Rn
and u ∈ Du ⊂ Rm , where D and Du are domains that contain x = 0 and u = 0,
respectively. Suppose further that x = 0 is an asymptotically stable equilibrium point
of ẋ = f (x, 0). Then, there is r > 0 such that for each x0 with x0 ≤ r, the system
(6.14) is small-signal finite-gain L2 stable with L2 gain less than or equal to γ. 3
6.4. EXERCISES 137

Proof: From Lemma 4.7, the system ẋ = f (x, u) is locally input-to-state-stable.

Hence, there exist positive constants k1 and k2 , a class KL function β, and a class
K function γ0 such that for any initial state x0 with x0 ≤ k1 and any input u(t)
with sup0≤t≤τ u(t) ≤ k2 , the solution x(t) satisfies
  
x(t) ≤ max β( x0 , t), γ0 sup u(σ)
0≤σ≤t

for all 0 ≤ t ≤ τ . Thus, by choosing k1 and k2 small enough, we can be sure that
x(t) ∈ D and u(t) ∈ Du for all 0 ≤ t ≤ τ . Proceeding as in the proof of Lemma 6.1
we arrive at (6.16). 2

With the help of Lemma 6.2, we can state small-signal versions of Theorems 6.5
and 6.6.
Theorem 6.7 Consider the system (6.17). Assume (6.18) is satisfied in some
neighborhood of (x = 0, u = 0) and the origin is an asymptotically stable equi-
librium point of ẋ = f (x, 0). Then, the system is small-signal finite-gain L2 stable
and its L2 gain is less than or equal to 1/δ. 3
Theorem 6.8 Consider the system (6.19). Assume (6.20) is satisfied in some
neighborhood of (x = 0, u = 0) and the origin is an asymptotically stable equi-
librium point of ẋ = f (x). Then, the system is small-signal finite-gain L2 stable
and its L2 gain is less than or equal to γ. 3
Example 6.10 As a variation on the theme of Example 6.8, consider the system
ẋ1 = x2 , ẋ2 = −a(x1 − 13 x31 ) − kx2 + u, y = x2
1 2 
where a, k > 0. The function V (x) = a 2 x1 − 12 1 4
x1 + 12 x22 is positive semidefinite
√
in the set {|x1 | ≤ 6} and satisfies inequality (6.18) since
V̇ = −kx22 + x2 u = −ky 2 + yu
With u = 0, V̇ = −kx22 ≤ 0 and asymptotic stability of the origin can be shown by
LaSalle’s invariance principle since
x2 (t) ≡ 0 ⇒ x1 (t)[3 − x21 (t)] ≡ 0 ⇒ x1 (t) ≡ 0
√
in the domain {|x1 | < 3}. Alternatively, we can show asymptotic stability by
linearization, which results in a Hurwitz matrix. By Theorem 6.7, we conclude that
the system is small-signal finite-gain L2 stable and its L2 gain is ≤ 1/k. 

6.4 Exercises
6.1 Show that the series connection of two L stable (respectively, finite-gain L
stable) systems is L stable (respectively, finite-gain L stable).
138 CHAPTER 6. INPUT-OUTPUT STABILITY

y y

u u

(a) On – off with hysteresis (b) On–off with dead zone and hysteresis

y y

u u

(c) Ideal on–off (d) On– off with dead zone

Figure 6.1: Relay characteristics

6.2 Show that the parallel connection of two L stable (respectively, finite-gain L
stable) systems is L stable (respectively, finite-gain L stable).

6.3 Consider the memoryless function y = h(u), where h belongs to the sector
[α, β] with β > α ≥ 0. Show that h is finite-gain Lp stable for every p ∈ [1, ∞] and
estimate the Lp gain.

6.4 For each of the relay characteristices shown in Figure 6.1, investigate L∞ and
L2 stability.

6.5 Consider the system

ẋ1 = −x1 + x2 , ẋ2 = −x1 − ψ(x2 ) + u, y = x2

where ψ is locally Lipschitz and belongs to the sector [k, ∞] for some k > 0. Show
that the system is finite-gain Lp stable for every p ∈ [1, ∞] and estimate the Lp
gain.
6.4. EXERCISES 139

6.6 Consider the system

ẋ1 = −x1 + x2 , ẋ2 = −ψ(x1 ) − x2 + u, y = x2

where ψ is locally Lipschitz and belongs to the sector [k, ∞] for some k > 0. Show
that the system is L∞ stable. Under what additional conditions on ψ will it be
finite-gain L∞ stable?

6.7 For each of the systems in Exercise 4.14, investigate L∞ and finite-gain L∞
stability when the output is defined as follows.

(1) y = x (2) y = sin x (3) y = x (4) y = x

(5) y = x21 (6) y = x1 + x2 (7) y = x1 + x2 (8) y = x2
(9) y = x2 (10) y = x1

6.8 For each of the systems in Exercise 4.15, investigate L∞ and small-signal L∞
stability when the output is x1 .

6.9 Investigate finite-gain L2 stability of the systems in the following exercises:

(1) Exercise 5.2
(2) Exercise 5.5
(3) Exercise 5.6
When it is stable, estimate the L2 gain.

6.10 Consider the tunnel-diode circuit (A.7) with output y = x2 .

(a) Show that the system is L∞ stable.
(b) Show that it is finite-gain L2 stable and estimate the L2 gain.

6.11 Consider the system

ẋ1 = −2x1 + x2 , ẋ2 = −2x1 − ψ(x1 ) + u, y = −x1 + x2

where ψ is a locally Lipschitz passive function.

x
(a) With u = 0, use V (x) = 0 1 ψ(σ)dσ + 12 x21 + 12 (x2 − x1 )2 to show that the
origin is globally asymptotically stable.
(b) Show that the system is L∞ stable.
(c) Show that it is finite-gain L2 stable and estimate the L2 gain.

6.12 Show that the system of Exercise 5.5 is finite-gain L2 stable and estimate
the L2 gain.
140 CHAPTER 6. INPUT-OUTPUT STABILITY

6.13 Consider the pendulum equation (A.2) with b > 0.

(a) Show that whether the output is taken as y = x1 or y = x2 , the system will be
small-signal finite-gain Lp stable for every p ∈ [1, ∞].

(b) If the output is y = x2 , show that the system will be finite-gain L2 stable and
estimate the L2 gain.

6.14 An m-link robot manipulator is represented by equation (A.35). Suppose

g(q) = 0 has an isolated solution q = q ∗ and the potential energy P (q) has a
minimum at q ∗ ; that is, P (q) ≥ P (q ∗ ) for all q. Let u = −Kd q̇ + v, where Kd is a
positive definite symmetric matrix. Show that the system, with input v and output
q̇ is finite-gain L2 stable and estimate the L2 gain. Hint: See Exercise 5.8.

6.15 Consider the system

ẋ1 = −x1 − x2 + u, ẋ2 = ψ1 (x1 ) − ψ2 (x2 ), y = x1

where ψ1 and ψ2 are locally Lipschitz functions and ψi belongs to the sector [αi , βi ]
with βi > αi > 0.

(a) Show that the system is finite-gain Lp stable for every p ∈ [1, ∞] and estimate
the Lp gain.

(b) Apply Theorem 6.6 to show that the L2 gain is ≤ β1 /α1 .

6.16 Consider the system

ẋ1 = x2 , ẋ2 = −ψ(x1 ) − x2 + u, y = x1 + 2x2

where ψ is locally Lipschitz, ψ(0) = 0, and x1 ψ(x1 ) ≥ αx21 ∀ x1 with α > 0. Apply
Theorem 6.6 to show that the system is finite-gain L2 stable  x and estimate the L2
gain. Hint: Use a function of the form V (x) = k(xT P x + 0 1 ψ(σ) dσ).

6.17 Consider the TORA system (A.49)–(A.52) and let u = −βx2 + w, where
β > 0. In this exercise we study the L2 stability of the system with input w and
output x2 . Let E(x) be the sum of the potential energy 12 kx23 and the kinetic energy
1 T
2 v Dv, where

x J + mL2 mL cos x1
v= 2 and D =
x4 mL cos x1 m+M

(a) Using V (x) = αE(x), show that α can be chosen such that V (x) satisfies (6.20).

(b) Show that the system is finite-gain L2 stable and the L2 gain is ≤ 1/β.
Chapter 7

Stability of Feedback Systems

Many systems can be represented by the negative feedback connection of Figure 7.1.
This feedback structure brings in new challenges and opportunities. The challenge
is exemplified by the fact that the feedback connection of two stable systems could
be unstable. In the face of this challenge comes the opportunity to take advantage
of the feedback structure itself to come up with conditions that guard against insta-
bility. For linear time-invariant systems, where H1 and H2 are transfer functions,
we know from the Nyquist criterion that the feedback loop will be stable if the
loop phase is less than 180 degrees or the loop gain is less than one. The passivity
theorems of Section 7.1 and the small-gain theorem of Section 7.3 extend these two
conditions, respectively, to nonlinear systems.
The main passivity theorem states that the feedback connection of two passive
systems is passive. Under additional conditions, we can show that the feedback
connection has an asymptotically stable equilibrium point at the origin, or that the
input-output map is finite-gain L2 stable. The connection between passivity and
the phase of a transfer function comes from the frequency-domain characterization
of positive real transfer functions, given in Section 5.3, which shows that the phase

u1 e1 y1
H1
+
−

y2 e2 +
H2 + u2

Figure 7.1: Feedback connection.

141
142 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

of a positive real transfer function cannot exceed 90 degrees. Hence, the loop phase
cannot exceed 180 degrees. If one of the two transfer functions is strictly positive
real, the loop phase will be strictly less than 180 degrees. The small-gain theorem
deals with a feedback connection where each component is finite-gain L-stable and
requires the loop gain to be less than one.1
Another opportunity that arises in feedback systems is the possibility of mod-
eling the system in such a way that the component in the forward path is a linear
time-invariant system while the one in the feedback path is a memoryless nonlin-
earity. The absolute stability results of Section 7.3 take advantage of this structure
and provide the circle and Popov criteria, which allow for unknown nonlinearities
that belong to a given sector and require frequency-domain conditions on the linear
system that build on the classical Nyquist criterion and Nyquist plot.

7.1 Passivity Theorems

Consider the feedback connection of Figure 7.1 where each of the feedback compo-
nents H1 and H2 is either a time-invariant dynamical system represented by the
state model
ẋi = fi (xi , ei ), yi = hi (xi , ei ) (7.1)
or a (possibly time-varying) memoryless function represented by

yi = hi (t, ei ) (7.2)

We are interested in using passivity properties of the feedback components H1

and H2 to analyze stability of the feedback connection. We require the feedback
connection to have a well-defined state model. When both components H1 and H2
are dynamical systems, the closed-loop state model takes the form

ẋ = f (x, u), y = h(x, u) (7.3)

where x = col(x1 , x2 ), u = col(u1 , u2 ), and y = col(y1 , y2 ). We assume that f is

locally Lipschitz, h is continuous, f (0, 0) = 0, and h(0, 0) = 0. It can be easily
verified that the feedback connection will have a well-defined state model if the
equations
e1 = u1 − h2 (x2 , e2 ), e2 = u2 + h1 (x1 , e1 ) (7.4)
have a unique solution (e1 , e2 ) for every (x1 , x2 , u1 , u2 ). The properties f (0, 0) = 0
and h(0, 0) = 0 follow from fi (0, 0) = 0 and hi (0, 0) = 0. It is also easy to see
that (7.4) will always have a unique solution if h1 is independent of e1 or h2 is
independent of e2 . In this case, the functions f and h of the closed-loop state
model inherit smoothness properties of the functions fi and hi of the feedback
components. In particular, if fi and hi are locally Lipschitz, so are f and h. For
1 Nonlinear extensions of this classical small-gain theorem can be found in [69, 90, 138].
7.1. PASSIVITY THEOREMS 143

linear systems, requiring hi to be independent of ei is equivalent to requiring the

transfer function of Hi to be strictly proper.
When one component, H1 say, is a dynamical system, while the other one is a
memoryless function, the closed-loop state model takes the form

ẋ = f (t, x, u), y = h(t, x, u) (7.5)

where x = x1 , u = col(u1 , u2 ), and y = col(y1 , y2 ). We assume that f is piecewise

continuous in t and locally Lipschitz in (x, u), h is piecewise continuous in t and
continuous in (x, u), f (t, 0, 0) = 0, and h(t, 0, 0) = 0. The feedback connection will
have a well-defined state model if the equations

e1 = u1 − h2 (t, e2 ), e2 = u2 + h1 (x1 , e1 ) (7.6)

have a unique solution (e1 , e2 ) for every (x1 , t, u1 , u2 ). This will be always the case
when h1 is independent of e1 .
The starting point of our analysis is the following fundamental property:
Theorem 7.1 The feedback connection of two passive systems is passive.
Proof: Let V1 (x1 ) and V2 (x2 ) be the storage functions for H1 and H2 , respectively.
If either component is a memoryless function, take Vi = 0. Then, eTi yi ≥ V̇i . From
the feedback connection of Figure 7.1, we see that

eT1 y1 + eT2 y2 = (u1 − y2 )T y1 + (u2 + y1 )T y2 = uT1 y1 + uT2 y2

Hence,
uT y = uT1 y1 + uT2 y2 ≥ V̇1 + V̇2
With V (x) = V1 (x1 ) + V2 (x2 ) as the storage function for the feedback connection,
we obtain uT y ≥ V̇ . 2

Now we want to use passivity to study stability and asymptotic stability of the
origin of the closed-loop system when the input u = 0. Stability of the origin fol-
lows trivially from Theorem 7.1 and Lemma 5.5. Therefore, we focus our attention
on studying asymptotic stability. The next theorem is an immediate consequence
of Theorem 7.1 and Lemma 5.6.

Theorem 7.2 Consider the feedback connection of two time-invariant dynamical

systems of the form (7.1). The origin of the closed-loop system (7.3) (when u = 0)
is asymptotically stable if one of the following conditions is satisfied:
• both feedback components are strictly passive;
• both feedback components are output strictly passive and zero-state observable;
• one component is strictly passive and the other one is output strictly passive
and zero-state observable.
144 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

Furthermore, if the storage function for each component is radially unbounded, the
origin is globally asymptotically stable. 3

Proof: Let V1 (x1 ) and V2 (x2 ) be the storage functions for H1 and H2 , respectively.
As in the proof of Lemma 5.6, we can show that V1 (x1 ) and V2 (x2 ) are positive
definite functions. Take V (x) = V1 (x1 ) + V2 (x2 ) as a Lyapunov function candidate
for the closed-loop system. In the first case, the derivative V̇ satisfies

V̇ ≤ uT y − ψ1 (x1 ) − ψ2 (x2 ) = −ψ1 (x1 ) − ψ2 (x2 )

since u = 0. Hence, the origin is asymptotically stable. In the second case,

V̇ ≤ −y1T ρ1 (y1 ) − y2T ρ2 (y2 )

where yiT ρi (yi ) > 0 for all yi = 0. Here V̇ is only negative semidefinite and V̇ = 0 ⇒
y = 0. To apply the invariance principle, we need to show that y(t) ≡ 0 ⇒ x(t) ≡ 0.
Note that y2 (t) ≡ 0 ⇒ e1 (t) ≡ 0. By zero-state observability of H1 , y1 (t) ≡ 0 ⇒
x1 (t) ≡ 0. Similarly, y1 (t) ≡ 0 ⇒ e2 (t) ≡ 0 and, by zero-state observability of H2 ,
y2 (t) ≡ 0 ⇒ x2 (t) ≡ 0. Thus, the origin is asymptotically stable. In the third case
(with H1 as the strictly passive component),

V̇ ≤ −ψ1 (x1 ) − y2T ρ2 (y2 )

and V̇ = 0 implies x1 = 0 and y2 = 0. Once again, y2 (t) ≡ 0 ⇒ e1 (t) ≡ 0, which

together with x1 (t) ≡ 0 imply that y1 (t) ≡ 0. Hence, e2 (t) ≡ 0 and, by zero-state
observability of H2 , y2 (t) ≡ 0 ⇒ x2 (t) ≡ 0. Thus, the origin is asymptotically
stable. Finally, if V1 (x1 ) and V2 (x2 ) are radially unbounded, so is V (x), and we can
conclude global asymptotic stability. 2

The proof uses the sum of the storage functions for the feedback components as a
Lyapunov function candidate for the feedback connection. The analysis is restrictive
because, to show that V̇ = V̇1 + V̇2 ≤ 0, we insist that both V̇1 ≤ 0 and V̇2 ≤ 0.
Clearly, this is not necessary. One term, V̇1 say, could be positive over some region
as long as the sum V̇ ≤ 0 over the same region. This idea is exploited in Examples
7.2 and 7.3, while Example 7.1 is a straightforward application of Theorem 7.2.
Example 7.1 Consider the feedback connection of
⎧ ⎧
⎨ ẋ1 = x2 ⎨ ẋ3 = x4
H1 : ẋ2 = −ax31 − kx2 + e1 and H2 : ẋ4 = −bx3 − x34 + e2
⎩ ⎩
y1 = x2 y2 = x4
1
where a, b, and k are positive constants. Using V1 = 4
4 ax1 + 12 x22 as a storage
function for H1 , we obtain

V̇1 = ax31 x2 − ax31 x2 − kx22 + x2 e1 = −ky12 + y1 e1

7.1. PASSIVITY THEOREMS 145

Hence, H1 is output strictly passive. Besides, with e1 = 0, we have

y1 (t) ≡ 0 ⇔ x2 (t) ≡ 0 ⇒ x1 (t) ≡ 0

which shows that H1 is zero-state observable. Using V2 = 12 bx23 + 12 x24 as a storage

function for H2 , we obtain

V̇2 = bx3 x4 − bx3 x4 − x44 + x4 e2 = −y24 + y2 e2

Therefore, H2 is output strictly passive. Moreover, with e2 = 0, we have

y2 (t) ≡ 0 ⇔ x4 (t) ≡ 0 ⇒ x3 (t) ≡ 0

which shows that H2 is zero-state observable. Thus, by the second case of Theo-
rem 7.2 and the fact that V1 and V2 are radially unbounded, we conclude that the
origin is globally asymptotically stable. 

Example 7.2 Reconsider the feedback connection of the previous example, but
change the output of H1 to y1 = x2 + e1 . From the expression

V̇1 = −kx22 + x2 e1 = −k(y1 − e1 )2 − e21 + y1 e1

we can conclude that H1 is passive, but we cannot conclude strict passivity or

output strict passivity. Therefore, we cannot apply Theorem 7.2. Using

V = V1 + V2 = 14 ax41 + 12 x22 + 12 bx23 + 12 x24

as a Lyapunov function candidate for the closed-loop system, we obtain

V̇ = −kx22 + x2 e1 − x44 + x4 e2 = −kx22 − x2 x4 − x44 + x4 (x2 − x4 )

= −kx22 − x44 − x24 ≤ 0

Moreover, V̇ = 0 implies that x2 = x4 = 0 and

x2 (t) ≡ 0 ⇒ ax31 (t) − x4 (t) ≡ 0 ⇒ x1 (t) ≡ 0

x4 (t) ≡ 0 ⇒ −bx3 (t) + x2 (t) − x4 (t) ≡ 0 ⇒ x3 (t) ≡ 0

Thus, by the invariance principle and the fact that V is radially unbounded, we
conclude that the origin is globally asymptotically stable. 

Example 7.3 Reconsider the system

ẋ1 = x2 , ẋ2 = −h1 (x1 ) − h2 (x2 )

from Examples 3.8 and 3.9, where h1 and h2 are locally Lipschitz and belong to
the sector (0, ∞). The system can be viewed as the state model of the feedback
connection of Figure 7.2, where H1 consists of a negative feedback loop around the
146 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

H1
+ + x2
− −

h2 (.)

x1
h1(.)
H2

Figure 7.2: Example 7.3.

integrator x2 with h2 in the feedback path, and H2 consists of a cascade connection

of the integrator x1 with h1 . We saw in Example 5.2, that H1 is output strictly
1 2
passive with the storage function x1V1 = 2 x2 and, in Example 5.3, that H2 is lossless
with the storage function V2 = 0 h1 (σ) dσ. We cannot apply Theorem 7.2 because
H2 is neither
 x strictly passive nor output strictly passive. However, using V =
V1 + V2 = 0 1 h1 (σ) dσ + 12 x22 as a Lyapunov function candidate, we proceed to
investigate asymptotic stability of the origin. This is already done in Examples 3.8
and 3.9, where it is shown that the origin is asymptotically stable and will be
y
globally asymptotically stable if 0 h1 (z) dz → ∞ as |y| → ∞. We will not
repeat the analysis of these two examples here, but note that if h1 (y) and h2 (y)
belong to the sector (0, ∞) only for y ∈ (−a, a), then the Lyapunov analysis can
be limited to some region around the origin, leading to a local asymptotic stability
conclusion, as in Example 3.8. This shows that passivity remains useful as a tool
for Lyapunov analysis even when it holds only on a finite region. 

When the feedback connection has a dynamical system as one component and
a memoryless function as the other component, we can perform Lyapunov analysis
by using the storage function of the dynamical system as a Lyapunov function
candidate. It is important, however, to distinguish between time-invariant and
time-varying memoryless functions, for in the latter case the closed-loop system
will be time varying and we cannot apply the invariance principle as we did in the
proof of Theorem 7.2. We treat these two cases separately in the next two theorems.

Theorem 7.3 Consider the feedback connection of a strictly passive, time-invariant,

dynamical system of the form (7.1) with a passive (possibly time-varying) memo-
ryless function of the form (7.2). Then, the origin of the closed-loop system (7.5)
(when u = 0) is uniformly asymptotically stable. Furthermore, if the storage func-
tion for the dynamical system is radially unbounded, the origin will be globally uni-
7.1. PASSIVITY THEOREMS 147

formly asymptotically stable. 3

Proof: As in the proof of Lemma 5.6, it can be shown that V1 (x1 ) is positive
definite. Its derivative is given by
∂V1
V̇1 = f1 (x1 , e1 ) ≤ eT1 y1 − ψ1 (x1 ) = −eT2 y2 − ψ1 (x1 ) ≤ −ψ1 (x1 )
∂x1
The conclusion follows from Theorem 4.2. 2

Theorem 7.4 Consider the feedback connection of a time-invariant dynamical sys-

tem H1 of the form (7.1) with a time-invariant memoryless function H2 of the form
(7.2). Suppose that H1 is zero-state observable and has a positive definite storage
function V1 (x), which satisfies

eT1 y1 ≥ V̇1 + y1T ρ1 (y1 ) (7.7)

and that H2 satisfies

eT2 y2 ≥ eT2 ϕ2 (e2 ) (7.8)
Then, the origin of the closed-loop system (7.5) (when u = 0) is asymptotically
stable if
v T [ρ1 (v) + ϕ2 (v)] > 0, ∀ v = 0 (7.9)
Furthermore, if V1 is radially unbounded, the origin will be globally asymptotically
stable. 3

Proof: Use V1 (x1 ) as a Lyapunov function candidate, to obtain

∂V1
V̇1 = f1 (x1 , e1 ) ≤ eT1 y1 − y1T ρ1 (y1 )
∂x1
= −eT2 y2 − y1T ρ1 (y1 ) ≤ −[y1T ϕ2 (y1 ) + y1T ρ1 (y1 )]

Inequality (7.9) shows that V̇1 ≤ 0 and V̇1 = 0 ⇒ y1 = 0. Noting that y1 (t) ≡
0 ⇒ e2 (t) ≡ 0 ⇒ e1 (t) ≡ 0, we see that zero-state observability of H1 implies that
x1 (t) ≡ 0. The conclusion follows from the invariance principle. 2

Example 7.4 Consider the feedback connection of a strictly positive real transfer
function and a passive time-varying memoryless function. From Lemma 5.4, we
know that the dynamical system is strictly passive with a positive definite storage
function of the form V (x) = 12 xT P x. From Theorem 7.3, we conclude that the
origin of the closed-loop system is globally uniformly asymptotically stable. This is
a version of the circle criterion of Section 7.3. 

Example 7.5 Consider the feedback connection of


ẋ = f (x) + G(x)e1
H1 : and H2 : y2 = σ(e2 )
y1 = h(x)
148 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

where ei , yi ∈ Rm , σ(0) = 0, and eT2 σ(e2 ) > 0 for all e2 = 0. Suppose there is a
radially unbounded positive definite function V1 (x) such that

∂V1 ∂V1
f (x) ≤ 0, G(x) = hT (x), ∀ x ∈ Rn
∂x ∂x
and H1 is zero-state observable. Both components are passive. Moreover, H2
satisfies eT2 y2 = eT2 σ(e2 ). Thus, (7.7) is satisfied with ρ1 = 0, and (7.8) is satisfied
with ϕ2 = σ. Since vσ(v) > 0 for all v = 0, (7.9) is satisfied. It follows from
Theorem 7.4 that the origin of the closed-loop system is globally asymptotically
stable. 

The utility of Theorems 7.3 and 7.4 can be extended by loop transformations.
Suppose H1 is a time-invariant dynamical system, while H2 is a (possibly time-
varying) memoryless function that belongs to the sector [K1 , K2 ], where K = K2 −
K1 is a positive definite symmetric matrix. We saw in Section 5.1 that a function in
the sector [K1 , K2 ] can be transformed into a function in the sector [0, ∞] by input
feedforward followed by output feedback, as shown in Figure 5.5. Input feedforward
on H2 can be nullified by output feedback on H1 , as shown in Figure 7.3(b), resulting
in an equivalent feedback connection, as far as asymptotic stability of the origin
is concerned. Similarly, premultiplying the modified H2 by K −1 can be nullified
by postmultiplying the modified H1 by K, as shown in Figure 7.3(c). Finally,
output feedback on the component in the feedback path can be nullified by input
feedforward on the component in the forward path, as shown in Figure 7.3(d).
The reconfigured feedback connection has two components H̃1 and H̃2 , where H̃2
is a memoryless function that belongs to the sector [0, ∞]. We can now apply
Theorem 7.3 if H̃2 is time varying or Theorem 7.4 if it is time invariant.
Example 7.6 Consider the feedback connection of
⎧
⎨ ẋ1 = x2
H1 : ẋ2 = −h(x1 ) + cx2 + e1 and H2 : y2 = σ(e2 )
⎩
y1 = x2

where σ ∈ [α, β], h ∈ [α1 , ∞], c > 0, α1 > 0, and b = β − α > 0. Applying the loop
transformation of Figure 7.3(d) (with K1 = α and K2 = β) results in the feedback
connection of
⎧
⎨ ẋ1 = x2
H̃1 : ẋ2 = −h(x1 ) − ax2 + ẽ1 and H̃2 : ỹ2 = σ̃(ẽ2 )
⎩
ỹ1 = bx2 + ẽ1

where σ̃ ∈ [0, ∞] and a = α − c. When α > c, it is shown in Example 5.4 that H̃1 is
strictly passive with a radially unbounded storage function. Thus, we conclude from
Theorem 7.3 that the origin of the feedback connection is globally asymptotically
stable. 
7.1. PASSIVITY THEOREMS 149

+ + +
H1 H1
− − −

K1

+
H2 H2
−

K1

(a) (b)

+ + + + + H̃1
H1 K H1 K
− − − − +

K1 K1

+ + +
H2 K−1 H2 K−1
− − +

K1 K1
H̃2

(c) (d)

Figure 7.3: Loop transformation with constant gains. A memoryless function H2 in

the sector [K1 , K2 ] is transformed into a memoryless function H̃2 in the sector [0, ∞].

Passivity theorems can also establish finite-gain L2 stability of the feedback

connection of Figure 7.1, where H1 : Lm 2e → L2e and H2 : L2e → L2e . Define
m m m

e = col(e1 , e2 ). Suppose that the feedback system has a well-defined state model of
2e there exist unique outputs e, y ∈ L2e .
the form (7.3) or (7.5), and for every u ∈ L2m 2m

The question of interest here is whether the feedback connection, when viewed as a
mapping from the input u to the output e or the output y, is finite-gain L2 stable.
It is not hard to see that the mapping from u to e is finite-gain L2 stable if and
only if the same is true for the mapping from u to y. Therefore, we can simply say
that the feedback connection is finite-gain L2 stable if either mapping is finite-gain
L2 stable. The next theorem is an immediate consequence of Theorem 6.5.
150 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

Theorem 7.5 The feedback connection of two output strictly passive systems with

eTi yi ≥ V̇i + δi yiT yi , δi > 0

is finite-gain L2 stable and its L2 gain is less than or equal to 1/ min{δ1 , δ2 }.

Proof: With V = V1 + V2 and δ = min{δ1 , δ2 }, we have

uT y = eT1 y1 + eT2 y2 ≥ V̇1 + δ1 y1T y1 + V̇2 + δ2 y2T y2

≥ V̇ + δ(y1T y1 + y2T y2 ) = V̇ + δy T y

Application of Theorem 6.5 completes the proof. 2

The proof of Theorem 6.5 uses the inequality

uT y ≥ V̇ + δy T y (7.10)

to arrive at
1 T δ
u u − yT y
V̇ ≤ (7.11)
2δ 2
which is then used to show finite-gain L2 stability. However, even if (7.10) does not
hold for the feedback connection, we may still be able to show an inequality of the
form (7.11). This idea is used in the next theorem to prove a more general result
that includes Theorem 7.5 as a special case.
Theorem 7.6 Consider the feedback connection of Figure 7.1 and suppose each
feedback component satisfies the inequality

eTi yi ≥ V̇i + εi eTi ei + δi yiT yi , for i = 1, 2 (7.12)

for some storage function Vi (xi ). Then, the closed-loop map from u to y is finite-
gain L2 stable if
ε1 + δ2 > 0 and ε2 + δ1 > 0 (7.13)
3
Proof: Adding inequalities (7.12) for i = 1, 2 and using

eT1 y1 + eT2 y2 = uT1 y1 + uT2 y2

T
e1 e1 = uT1 u1 − 2uT1 y2 + y2T y2
eT2 e2 = uT2 u2 + 2uT2 y1 + y1T y1

we obtain
V̇ ≤ −y T Ly − uT M u + uT N y
where

(ε2 + δ1 )I 0 ε1 I 0 I 2ε1 I
L= , M= , N=
0 (ε1 + δ2 )I 0 ε2 I −2ε2 I I
7.1. PASSIVITY THEOREMS 151

and V (x) = V1 (x1 ) + V2 (x2 ). Let a = min{ε2 + δ1 , ε1 + δ2 } > 0, b = N ≥ 0, and

c = M ≥ 0. Then

V̇ ≤ −a y 2 + b u y + c u 2
1 b2 a
= − (b u − a y )2 + u 2 − y 2 + c u 2
2a 2a 2
k2 a
≤ u − y
2 2
2a 2

where k 2 = b2 + 2ac. Application of Lemma 6.1 completes the proof. 2

Theorem 7.6 reduces to Theorem 7.5 when (7.12) is satisfied with ε1 = ε2 = 0,

δ1 > 0, and δ2 > 0. However, condition (7.13) is satisfied in several other cases.
For example, it is satisfied when both H1 and H2 are input strictly passive with
eTi yi ≥ V̇i + εi uTi ui for some εi > 0. It is also satisfied when one component (H1 , for
example) is passive, while the other component satisfies (7.12) with positive ε2 and
δ2 . What is more interesting is that (7.13) can be satisfied even when some of the
constants εi and δi are negative. For example, a negative ε1 can be compensated
for by a positive δ2 and vice versa.

Example 7.7 Consider the feedback connection of


ẋ = f (x) + G(x)e1
H1 : and H2 : y2 = σ(e2 )
y1 = h(x)

where ei , yi ∈ Rm and σ is a decoupled nonlinearity as in (5.1) with each component

σk in the sector [α, β] with β > α > 0. Suppose there is a positive semidefinite
function V1 (x) such that

∂V1 ∂V1
f (x) ≤ 0, G(x) = hT (x), ∀ x ∈ Rn
∂x ∂x
Both components are passive. Moreover, H2 satisfies

1 T 1
αeT2 e2 ≤ eT2 y2 ≤ βeT2 e2 and y y2 ≤ eT2 y2 ≤ y2T y2
β 2 α

Hence,

1
eT2 y2 = γeT2 y2 + (1 − γ)eT2 y2 ≥ γαeT2 e2 + (1 − γ) y2T y2 , 0<γ<1
β

Thus, (7.9) is satisfied with ε1 = δ1 = 0, ε2 = γα, and δ2 = (1 − γ)/β. This shows

that (7.13) is satisfied, and we conclude that the closed-loop map from u to y is
finite-gain L2 stable. 
152 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

Example 7.8 Consider the feedback connection of

⎧
⎨ ẋ1 = x2
H1 : ẋ2 = −ax31 − σ(x2 ) + e1 and H2 : y2 = ke2
⎩
y1 = x2

where σ ∈ [−α, ∞], a > 0, α > 0, and k > 0. If σ was in the sector [0, ∞], we could
have shown that H1 is passive with the storage function V1 (x) = 14 ax41 + 12 x22 . For
σ ∈ [−α, ∞], we have

V̇1 = ax31 x2 − ax31 x2 − x2 σ(x2 ) + x2 e1 ≤ αx22 + x2 e1 = αy12 + y1 e1

Hence, (7.9) is satisfied for H1 with ε1 = 0 and δ1 = −α. Since

(1 − γ) 2
e2 y2 = ke22 = γke22 + y2 , 0<γ<1
k
(7.9) is satisfied for H2 with ε2 = γk and δ2 = (1 − γ)/k. If k > α, we can choose
γ such that γk > α. Then, ε1 + δ2 > 0 and ε2 + δ1 > 0. We conclude that the
closed-loop map from u to y is finite-gain L2 stable if k > α.. 

7.2 The Small-Gain Theorem

Consider the feedback connection of Figure 7.1 where H1 : Lme → Le and H2 : Le →
q q

Le . Suppose both systems are finite-gain L stable; that is,

m

y1τ L ≤ γ1 e1τ L + β1 , ∀ e1 ∈ Lm
e , ∀ τ ∈ [0, ∞) (7.14)
y2τ L ≤ γ2 e2τ L + β2 , ∀ e2 ∈ Lqe , ∀ τ ∈ [0, ∞) (7.15)

Suppose further that the feedback system is well defined in the sense that for every
pair of inputs u1 ∈ Lm e and u2 ∈ Le , there exist unique outputs e1 , y2 ∈ Le
q m

and e2 , y1 ∈ Le . If H1 and H2 are represented by state models, we assume that

q

the feedback connection has a well-defined state model. Define u = col(u1 , u2 ),

y = col(y1 , y2 ), and e = col(e1 , e2 ). The next (small-gain) theorem gives a sufficient
condition for the feedback connection to be finite-gain L stable; that is, the mapping
from u to e, or equivalently from u to y, is finite-gain L stable.

Theorem 7.7 Under the preceding assumptions, the feedback connection is finite-
gain L stable if γ1 γ2 < 1. 3

Proof: Throughout the proof, · stands for · L . Assuming existence of the

solution, we can write

e1τ = u1τ − (H2 e2 )τ , e2τ = u2τ + (H1 e1 )τ

7.2. THE SMALL-GAIN THEOREM 153

Then,

e1τ ≤ u1τ + (H2 e2 )τ ≤ u1τ + γ2 e2τ + β2

≤ u1τ + γ2 ( u2τ + γ1 e1τ + β1 ) + β2
= γ1 γ2 e1τ + ( u1τ + γ2 u2τ + β2 + γ2 β1 )

Since γ1 γ2 < 1,
1
e1τ ≤ ( u1τ + γ2 u2τ + β2 + γ2 β1 ) (7.16)
1 − γ1 γ2
for all τ ∈ [0, ∞). Similarly,
1
e2τ ≤ ( u2τ + γ1 u1τ + β1 + γ1 β2 ) (7.17)
1 − γ1 γ2
for all τ ∈ [0, ∞). The proof is completed by noting that e ≤ e1 + e2 , which
follows from the triangle inequality. 2

The feedback connection of Figure 7.1 provides a convenient setup for studying
robustness issues in feedback systems. Quite often, feedback systems subject to
model uncertainties can be represented in the form of a feedback connection with
H1 , say, as a stable nominal system and H2 a stable perturbation. Then, the
requirement γ1 γ2 < 1 is satisfied whenever γ2 is small enough. Therefore, the
small-gain theorem provides a conceptual framework for understanding many of
the robustness results that arise in the study of feedback systems.

Example 7.9 Consider the feedback connection of Figure 7.1, with m = q. Let
H1 be a linear time-invariant system with a Hurwitz transfer function G(s). Let
H2 be a memoryless function y2 = ψ(t, e2 ) that satisfies

ψ(t, y) ≤ γ2 y , ∀ t ≥ 0, ∀ y ∈ Rm

From Theorem 6.4, we know that H1 is finite-gain L2 stable and its L2 gain is
≤ γ1 = supw∈R G(jω) . It is easy to verify that H2 is finite-gain L2 stable and its
L2 gain is ≤ γ2 . Assuming the feedback connection is well defined, we conclude by
Theorem 7.7 that it is finite-gain L2 stable if γ1 γ2 < 1.2 

Example 7.10 Consider the system3

ẋ = f (t, x, v + d1 (t)), εż = Az + B[u + d2 (t)], v = Cz

where f is a smooth function of its arguments, A is a Hurwitz matrix, −CA−1 B = I,

ε is a small positive parameter, and d1 , d2 are disturbance signals. The linear
2 The same condition results from applying the circle criterion; see Example 7.11.
3 The example is taken from [139].
154 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

part of this model represents actuator dynamics that are, typically, much faster
than the plant dynamics represented here by the nonlinear equation ẋ = f . The
disturbance signals d1 and d2 enter the system at the inputs of the plant and
actuator, respectively. Suppose the disturbance signals d1 and d2 belong to a signal
space L, where L could be any Lp space, and the goal is to attenuate the effect of
this disturbance on the state x. This goal can be met if feedback control can be
designed so that the closed-loop input–output map from (d1 , d2 ) to x is finite-gain
L stable and the L gain is less than some given tolerance δ > 0. To simplify the
design problem, it is common to neglect the actuator dynamics by setting ε = 0
and substituting v = −CA−1 B(u + d2 ) = u + d2 in the plant equation to obtain the
reduced-order model
ẋ = f (t, x, u + d)
where d = d1 +d2 . Assuming that the state variables are available for measurement,
we use this model to design state feedback control to meet the design objective.
Suppose we have designed a smooth state feedback control u = φ(t, x) such that

xL ≤ γdL + β (7.18)

for some γ < δ. Will the control meet the design objective when applied to the
actual system with the actuator dynamics included? This is a question of robustness
of the controller with respect to the unmodeled actuator dynamics. When the
control is applied to the actual system, the closed-loop equation is given by

ẋ = f (t, x, Cz + d1 (t)), εż = Az + B[φ(t, x) + d2 (t)]

Let us assume that d2 (t) is differentiable and d˙2 ∈ L. The change of variables

η = z + A−1 B[φ(t, x) + d2 (t)]

brings the closed-loop system into the form

ẋ = f (t, x, φ(t, x) + d(t) + Cη), εη̇ = Aη + εA−1 B[φ̇ + d˙2 (t)]

where
∂φ ∂φ
φ̇ = + f (t, x, φ(t, x) + d(t) + Cη)
∂t ∂x
It is not difficult to see that the closed-loop system can be represented in the form
of Figure 7.1 with H1 defined by

∂φ ∂φ
ẋ = f (t, x, φ(t, x) + e1 ), y1 = φ̇ = + f (t, x, φ(t, x) + e1 )
∂t ∂x
H2 defined by
1
η̇ = Aη + A−1 Be2 , y2 = −Cη
ε
7.3. ABSOLUTE STABILITY 155

u1 = d1 + d2 = d, and u2 = d˙2 . In this representation, the system H1 is the nominal

reduced-order closed-loop system, while H2 represents the effect of the unmodeled
dynamics. Setting ε = 0 opens the loop and the overall closed-loop system reduces
to the nominal one. Let us assume that the feedback function φ(t, x) satisfies the
inequality
∂φ ∂φ
+ f (t, x, φ(t, x) + e )
1 ≤ c1 x + c2 e1  (7.19)
∂t ∂x
for all (t, x, e1 ), where c1 and c2 are nonnegative constants. Using (7.18) and (7.19),
it can be shown that

y1 L ≤ γ1 e1 L + β1 where γ1 = c1 γ + c2 , β1 = c1 β

Since H2 is a linear time-invariant system and A is Hurwitz, we apply Theorem 6.1

with V (η) = εη T P η, where P is the solution of the Lyapunov equation P A+AT P =
−I, to show that H2 is finite-gain Lp stable for every p ∈ [1, ∞] and

def
y2 L ≤ γ2 e2 L + β2 = εγf e2 L + β2

where γf = 2λ2max (P )A−1 BC/λmin (P ). Thus, assuming the feedback connec-

tion is well defined, we conclude from Theorem 7.7 that the input–output map from
u to e is finite-gain L stable if εγf γ1 < 1. From (7.16),

1
e1 L ≤ [u1 L + εγf u2 L + εγf β1 + β2 ]
1 − εγ1 γf

Using
xL ≤ γe1 L + β
which follows from (7.18), and the definition of u1 and u2 , we obtain
γ
xL ≤ [dL + εγf d˙2 L + εγf β1 + β2 ] + β (7.20)
1 − εγ1 γf

The right-hand side of (7.20) approaches γdL + β + γβ2 as ε → 0, which shows

that for sufficiently small ε the upper bound on the L gain of the map from d to
x, under the actual closed-loop system, will be close to the corresponding quantity
under the nominal closed-loop system. 

7.3 Absolute Stability

Consider the feedback connection of Figure 7.4. We assume that the external input
r = 0 and study the behavior of the unforced system, represented by

ẋ = Ax + Bu, y = Cx + Du, u = −ψ(t, y) (7.21)

156 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

r + u y
G(s)
−

ψ(.)

Figure 7.4: Feedback connection.

where x ∈ Rn , u, y ∈ Rm , (A, B) is controllable, (A, C) is observable, and ψ is

a memoryless, possibly time-varying, nonlinearity, which is piecewise continuous
in t and locally Lipschitz in y. We assume that the feedback connection has a
well-defined state model, which is the case when

u = −ψ(t, Cx + Du) (7.22)

has a unique solution u for every (t, x) in the domain of interest. This is always the
case when D = 0. The transfer function matrix of the linear system

G(s) = C(sI − A)−1 B + D (7.23)

is square and proper. The controllability and observability assumptions ensure that
{A, B, C, D} is a minimal realization of G(s). The nonlinearity ψ is required to
satisfy a sector condition per Definition 5.2. The sector condition may be satisfied
globally, that is, for all y ∈ Rm , or satisfied only for y ∈ Y , a subset of Rm , whose
interior is connected and contains the origin.
For all nonlinearities satisfying the sector condition, the origin x = 0 is an
equilibrium point of the system (7.21). The problem of interest here is to study
the stability of the origin, not for a given nonlinearity, but rather for a class of
nonlinearities that satisfy a given sector condition. If we succeed in showing that
the origin is uniformly asymptotically stable for all nonlinearities in the sector, the
system is said to be absolutely stable. The problem was originally formulated by
Lure and is sometimes called Lure’s problem.4 Traditionally, absolute stability has
been defined for the case when the origin is globally uniformly asymptotically stable.
To keep up this tradition, we will use the phrase “absolute stability” when the sector
condition is satisfied globally and the origin is globally uniformly asymptotically
stable. Otherwise, we will use the phrase “absolute stability with finite domain.”

Definition 7.1 Consider the system (7.21), where ψ satisfies a sector condition
per Definition 5.2. The system is absolutely stable if the origin is globally uniformly
asymptotically stable for any nonlinearity in the given sector. It is absolutely stable
with finite domain if the origin is uniformly asymptotically stable.
4 For a historical perspective and further reading, see [63, 96, 97, 125].
7.3. ABSOLUTE STABILITY 157

We will investigate asymptotic stability of the origin by using Lyapunov analy-

sis. A Lyapunov function candidate can be chosen by using the passivity tools of
Section 7.1. In particular, if the closed-loop system can be represented as a feed-
back connection of two passive systems, then the sum of the two storage functions
can be used as a Lyapunov function candidate for the closed-loop system. The use
of loop transformations allows us to cover various sectors and Lyapunov function
candidates, leading to the circle and Popov criteria.

7.3.1 Circle Criterion

Theorem 7.8 The system (7.21) is absolutely stable if
• ψ ∈ [K1 , ∞] and G(s)[I + K1 G(s)]−1 is strictly positive real, or
• ψ ∈ [K1 , K2 ], with K = K2 − K1 = K T > 0, and [I + K2 G(s)][I + K1 G(s)]−1
is strictly positive real.
If the sector condition is satisfied only on a set Y ⊂ Rm , then the foregoing condi-
tions ensure absolute stability with finite domain. 3

We refer to this theorem as the multivariable circle criterion, although the reason
for using this name will not be clear until we specialize it to the single-input–single-
output case. A necessary condition for equation (7.22) to have a unique solution u
for every ψ ∈ [K1 , ∞] or ψ ∈ [K1 , K2 ] is the nonsingularity of the matrix (I +K1 D).
This can be seen by taking ψ = K1 y in (7.22). Therefore, the transfer function
[I + K1 G(s)]−1 is proper.

Proof of Theorem 7.8: We prove the theorem first for the sector [0, ∞] and re-
cover the other cases by loop transformations. If ψ ∈ [0, ∞] and G(s) is strictly pos-
itive real, we have a feedback connection of two passive systems. From Lemma 5.4,
we know that the storage function for the linear dynamical system is V (x) = 12 xT P x,
where P = P T > 0 satisfies the Kalman–Yakubovich–Popov equations (5.13)–
(5.15). Using V (x) as a Lyapunov function candidate, we obtain

V̇ = 12 xT P ẋ + 12 ẋT P x = 12 xT (P A + AT P )x + xT P Bu

Using (5.13) and (5.14) yields

V̇ = − 12 xT LT Lx − 1 T
2 εx P x + xT (C T − LT W )u
= − 12 xT LT Lx − 1 T
2 εx P x + (Cx + Du)T u − uT Du − xT LT W u

Using (5.15) and the fact that uT Du = 12 uT (D + DT )u, we obtain

V̇ = − 12 εxT P x − 1
2 (Lx + W u)T (Lx + W u) − y T ψ(t, y)

Since y T ψ(t, y) ≥ 0, we have

V̇ ≤ − 12 εxT P x
158 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

which shows that the origin is globally exponentially stable. If ψ satisfies the sector
condition only for y ∈ Y , the foregoing analysis will be valid in some neighborhood
of the origin, showing that the origin is exponentially stable. The case ψ ∈ [K1 , ∞]
can be transformed to a case where the nonlinearity belongs to [0, ∞] via the loop
transformation of Figure 7.3(b). Hence, the system is absolutely stable if G(s)[I +
K1 G(s)]−1 is strictly positive real. The case ψ ∈ [K1 , K2 ] can be transformed
to a case where the nonlinearity belongs to [0, ∞] via the loop transformation of
Figure 7.3(d). Hence, the system is absolutely stable if
I + KG(s)[I + K1 G(s)]−1 = [I + K2 G(s)][I + K1 G(s)]−1
is strictly positive real. 2

Example 7.11 Consider the system (7.21) and suppose G(s) is Hurwitz and strictly
proper. Let
γ1 = sup σmax [G(jω)] = sup G(jω)
ω∈R ω∈R

where σmax [·] denotes the maximum singular value of a complex matrix. The con-
stant γ1 is finite since G(s) is Hurwitz. Suppose ψ satisfies the inequality
ψ(t, y) ≤ γ2 y, ∀ t ≥ 0, ∀ y ∈ Rm (7.24)
then it belongs to the sector [K1 , K2 ] with K1 = −γ2 I and K2 = γ2 I. To apply
Theorem 7.8, we need to show that
Z(s) = [I + γ2 G(s)][I − γ2 G(s)]−1
is strictly positive real. We note that det[Z(s) + Z T (−s)] is not identically zero
because Z(∞) = I. We apply Lemma 5.1. Since G(s) is Hurwitz, Z(s) will be
Hurwitz if [I − γ2 G(s)]−1 is Hurwitz. Noting that5
σmin [I − γ2 G(jω)] ≥ 1 − γ1 γ2
we see that if γ1 γ2 < 1, the plot of det[I − γ2 G(jω)] will not go through nor encircle
the origin. Hence, by the multivariable Nyquist criterion,6 [I −γ2 G(s)]−1 is Hurwitz;
consequently, Z(s) is Hurwitz. Next, we show that Z(jω) + Z T (−jω) > 0, ∀ ω ∈ R.
Z(jω) + Z T (−jω) = [I + γ2 G(jω)][I − γ2 G(jω)]−1
+ [I − γ2 GT (−jω)]−1 [I + γ2 GT (−jω)]
 
= [I − γ2 GT (−jω)]−1 2I − 2γ22 GT (−jω)G(jω)
× [I − γ2 G(jω)]−1
5 The following properties of singular values of a complex matrix are used:
det G
= 0 ⇔ σmin [G] > 0, σmax [G−1 ] = 1/σmin [G], if σmin [G] > 0
σmin [I + G] ≥ 1 − σmax [G], σmax [G1 G2 ] ≤ σmax [G1 ]σmax [G2 ]

6 See [21, pp. 160–161] for a statement of the multivariable Nyquist criterion.
7.3. ABSOLUTE STABILITY 159

Hence, Z(jω) + Z T (−jω) is positive definite for all ω ∈ R if and only if

 
σmin I − γ22 GT (−jω)G(jω) > 0, ∀ ω ∈ R

Now, for γ1 γ2 < 1, we have

σmin [I − γ22 GT (−jω)G(jω)] ≥ 1 − γ22 σmax [GT (−jω)]σmax [G(jω)]

≥ 1 − γ12 γ22 > 0

Finally, Z(∞) + Z T (∞) = 2I. Thus, all the conditions of Lemma 5.1 are satisfied
and we conclude that Z(s) is strictly positive real and the system is absolutely stable
if γ1 γ2 < 1. This is a robustness result, which shows that closing the loop around
a Hurwitz transfer function with a nonlinearity satisfying (7.24), with a sufficiently
small γ2 , does not destroy the stability of the system.7 

In the case m = 1, the conditions of Theorem 7.8 can be verified graphically

by examining the Nyquist plot of G(s). For ψ ∈ [α, β], with β > α, the system is
absolutely stable if the scalar transfer function

1 + βG(s)
Z(s) =
1 + αG(s)

is strictly positive real. Lemma 5.1 states that Z(s) is strictly positive real if it is
Hurwitz and

1 + βG(jω)
Re > 0, ∀ ω ∈ [0, ∞] (7.25)
1 + αG(jω)
To relate condition (7.25) to the Nyquist plot of G(s), we distinguish between three
cases, depending on the sign of α. Consider first the case β > α > 0, where (7.25)
can be rewritten as
'1 (
β + G(jω)
Re 1 > 0, ∀ ω ∈ [0, ∞] (7.26)
α + G(jω)

For a point q on the Nyquist plot of G(s), the two complex numbers (1/β) + G(jω)
and (1/α) + G(jω) can be represented by the lines connecting q to −(1/β) + j0 and
−(1/α) + j0, respectively, as shown in Figure 7.5. The real part of the ratio of two
complex numbers is positive when the angle difference between the two numbers is
less than π/2; that is, (θ1 − θ2 ) < π/2 in Figure 7.5. If we define D(α, β) to be the
closed disk in the complex plane whose diameter is the line segment connecting the
points −(1/α) + j0 and −(1/β) + j0, then it is simple to see that the (θ1 − θ2 ) < π/2
when q is outside the disk D(α, β). Since (7.26) is required to hold for all ω, all
points on the Nyquist plot of G(s) must be strictly outside the disk D(α, β). On the
other hand, Z(s) is Hurwitz if G(s)/[1 + αG(s)] is Hurwitz. The Nyquist criterion
7 The inequality γ1 γ2 < 1 can be derived also from the small-gain theorem. (See Example 7.9.)
160 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

D (α, β) q

θ2 θ1

−1/α −1/β

Figure 7.5: Graphical representation of the circle criterion.

states that G(s)/[1 + αG(s)] is Hurwitz if and only if the Nyquist plot of G(s)
does not intersect the point −(1/α) + j0 and encircles it exactly p times in the
counterclockwise direction, where p is the number of poles of G(s) in the open
right-half complex plane.8 Therefore, the conditions of Theorem 7.8 are satisfied if
the Nyquist plot of G(s) does not enter the disk D(α, β) and encircles it p times
in the counterclockwise direction. Consider, next, the case when β > 0 and α = 0.
For this case, Theorem 7.8 requires 1 + βG(s) to be strictly positive real. This is
the case if G(s) is Hurwitz and

1
Re[1 + βG(jω)] > 0 ⇐⇒ Re[G(jω)] > − , ∀ ω ∈ [0, ∞]
β

which is equivalent to the graphical condition that the Nyquist plot of G(s) lies to
the right of the vertical line defined by Re[s] = −1/β. Finally, consider the case
α < 0 < β, where (7.25) is equivalent to
'1 (
β + G(jω)
Re 1 < 0, ∀ ω ∈ [0, ∞] (7.27)
α + G(jω)

The inequality sign is reversed because, as we go from (7.25) to (7.27), we multiply

by α/β, which is now negative. Repeating previous arguments, it can be seen
that for (7.27) to hold, the Nyquist plot of G(s) must lie inside the disk D(α, β).
Consequently, the Nyquist plot cannot encircle the point −(1/α) + j0. Therefore,
from the Nyquist criterion, we see that G(s) must be Hurwitz for G(s)/[1 + αG(s)]
to be so. The stability criteria for the three cases are summarized in the following
theorem, which is known as the circle criterion.

8 When G(s) has poles on the imaginary axis, the Nyquist path is indented in the right-half
plane, as usual.
7.3. ABSOLUTE STABILITY 161

Theorem 7.9 Consider a single-input–single-output system of the form (7.21),

where {A, B, C, D} is a minimal realization of G(s) and ψ ∈ [α, β]. Then, the
system is absolutely stable if one of the following conditions is satisfied, as appro-
priate:

1. If 0 < α < β, the Nyquist plot of G(s) does not enter the disk D(α, β) and
encircles it p times in the counterclockwise direction, where p is the number
of poles of G(s) with positive real parts.

2. If 0 = α < β, G(s) is Hurwitz and the Nyquist plot of G(s) lies to the right of
the vertical line Re[s] = −1/β.

3. If α < 0 < β, G(s) is Hurwitz and the Nyquist plot of G(s) lies in the interior
of the disk D(α, β).

If the sector condition is satisfied only on an interval [a, b], then the foregoing con-
ditions ensure absolute stability with finite domain. 3

The circle criterion allows us to investigate absolute stability by using only the
Nyquist plot of G(s). This is important because the Nyquist plot can be determined
directly from experimental data. Given the Nyquist plot of G(s), we can determine
permissible sectors for which the system is absolutely stable. The next two examples
illustrate the use of the circle criterion.

Example 7.12 The Nyquist plot of

24
G(s) =
(s + 1)(s + 2)(s + 3)

is shown in Figure 7.6. Since G(s) is Hurwitz, we can allow α to be negative and
apply the third case of the circle criterion. So, we need to determine a disk D(α, β)
that encloses the Nyquist plot. Clearly, the choice of the disk is not unique. Suppose
we locate the center of the disk at the origin of the complex plane and work with a
disk D(−γ2 , γ2 ), where the radius (1/γ2 ) > 0 is to be chosen. The Nyquist plot will
be inside this disk if |G(jω)| < 1/γ2 . In particular, if we set γ1 = supω∈R |G(jω)|,
then γ2 must be chosen to satisfy γ1 γ2 < 1. This is the same condition we found in
Example 7.11. It is not hard to see that |G(jω)| is maximum at ω = 0 and γ1 = 4.
Thus, γ2 must be less than 0.25. Hence, we conclude that the system is absolutely
stable for all nonlinearities in the sector (−0.25, 0.25). Inspection of the Nyquist
plot and the disk D(−0.25, 0.25) in Figure 7.6 suggests that locating the center at
the origin may not be the best choice. By locating the center at another point,
we might be able to obtain a disk that encloses the Nyquist plot more tightly. For
example, let us locate the center at the point 1.5 + j0. The maximum distance
from this point to the Nyquist plot is 2.834. Hence, choosing the radius of the disk
to be 2.9 ensures that the Nyquist plot is inside the disk D(−1/4.4, 1/1.4), and
we conclude that the system is absolutely stable for all nonlinearities in the sector
162 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

6 Im G
Im G 0.4
4
0.2
2
0
0 Re G
Re G
−2 −0.2

−4 −0.4
−5 0 5 −4 −2 0

Figure 7.6: Nyquist plot for Exam- Figure 7.7: Nyquist plot for Exam-
ple 7.12. ple 7.13.

[−0.227, 0.714]. Comparing this sector with the previous one shows that by giving in
a little bit on the lower bound of the sector, we achieve a significant improvement in
the upper bound. Clearly, there is still room for optimizing the choice of the center
of the disk. The point we wanted to show is that the graphical representation of
the circle criterion gives a closer look at the problem, compared with the use of
norm inequalities as in Example 7.11, which allows us to obtain less conservative
estimates of the sector. Another direction we can pursue is to restrict α to zero and
apply the second case of the circle criterion. The Nyquist plot lies to the right of the
vertical line Re[s] = −0.857. Hence, we can conclude that the system is absolutely
stable for all nonlinearities in the sector [0, 1.166]. It gives the best estimate of β,
which is achieved at the expense of limiting the nonlinearity to the first and third
quadrants. To appreciate how this flexibility in using the circle criterion could be
useful in applications, let us suppose that we are interested in studying the stability
of the system when ψ(y) = sat(y). The saturation nonlinearity belongs to the
sector [0, 1]. Therefore, it is included in the sector [0, 1.166], but not in the sector
(−0.25, 0.25) or [−0.227, 0.714]. 

Example 7.13 The transfer function

24
G(s) =
(s − 1)(s + 2)(s + 3)
is not Hurwitz, since it has a pole in the open right-half plane. So, we must restrict
α to be positive and apply the first case of the circle criterion. The Nyquist plot of
G(s) is shown in Figure 7.7. From the circle criterion, we know that the Nyquist plot
must encircle the disk D(α, β) once in the counterclockwise direction. Therefore,
the disk has to be totally inside the left lobe of the plot. Let us locate the center of
the disk at the point −3.2 + j0, about halfway between the two ends of the lobe on
the real axis. The minimum distance from this center to the Nyquist plot is 0.1688.
Hence, choosing the radius to be 0.168, we conclude that the system is absolutely
stable for all nonlinearities in the sector [0.2969, 0.3298]. 
7.3. ABSOLUTE STABILITY 163

1
Im G

y/a
ψ (y) 0.5
1

−a −1 1 a y 0
Re G

−0.5
−2 −1.5 −1 −0.5 0
Figure 7.8: Sector for Exam-
ple 7.14.
Figure 7.9: Nyquist plot for Exam-
ple 7.14.

In Examples 7.11 through 7.13, we have considered cases where the sector con-
dition is satisfied globally. In the next example, the sector condition is satisfied only
on a finite interval.
Example 7.14 Consider the feedback connection of Figure 7.4, where the linear
system is represented by the transfer function
s+2
G(s) =
(s + 1)(s − 1)
and the nonlinear element is ψ(y) = sat(y). The nonlinearity belongs globally to
the sector [0, 1]. However, since G(s) is not Hurwitz, we must apply the first case
of the circle criterion, which requires the sector condition to hold with a positive
α. Thus, we cannot conclude absolute stability by using the circle criterion.9 The
best we can hope for is to show absolute stability with finite domain. Figure 7.8
shows that on the interval [−a, a], the nonlinearity ψ belongs to the sector [α, β]
with α = 1/a and β = 1.
Since G(s) has a pole with positive real part, its Nyquist plot, shown in Fig-
ure 7.9, must encircle the disk D(α, 1) once in the counterclockwise direction. It
can be verified, analytically, that (7.25) is satisfied for α > 0.5359. Thus, choos-
ing α = 0.55, the sector condition is satisfied on the interval [−1.818, 1.818] and
the disk D(0.55, 1) is encircled once by the Nyquist plot in the counterclockwise
direction. From the first case of the circle criterion, we conclude that the system is
absolutely stable with finite domain. We can also use a quadratic Lyapunov func-
tion V (x) = xT P x to estimate the region of attraction. Let G(s) be realized by the
state model
ẋ1 = x2 , ẋ2 = x1 + u, y = 2x1 + x2
To use the Lyapunov function V (x) = xT P x as in the proof of Theorem 7.8, we
apply the loop transformation of Figure 7.3(d) to transform the nonlinearity ψ into
9 In
fact, the origin is not globally asymptotically stable because the system has three equilib-
rium points.
164 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

a passive nonlinearity. The loop transformation is given by

u = −αy + ũ = −0.55y + ũ, ỹ = (β − α)y + ũ = 0.45y + ũ

and the transformed linear system is

ẋ = Ax + B ũ, ỹ = Cx + Dũ

where

0 1 0  
A= , B= , C= 0.9 0.45 , and D = 1
−0.1 −0.55 1

The matrix P is obtained by solving the

√ Kalman–Yakubovich–Popov equations (5.13)–
(5.15). From equation (5.15), W = 2. Taking ε = 0.02, it can be verified that
equations (5.13)–(5.14) have two solutions, given by10

0.4946 0.4834  
P1 = , L1 = 0.2946 −0.4436
0.4834 1.0774

and

0.7595 0.4920  
P2 = , L2 = −0.2885 1.0554
0.4920 1.9426
Thus V1 (x) = xT P1 x and V2 (x) = xT P2 x are two different Lyapunov functions for
the system. For each Lyapunov function, we estimate the region of attraction by a
set of the form {V (x) ≤ c} in the interior of {|y| ≤ 1.8181}. Noting that11

(1.818)2 (1.818)2
min V1 (x) = = 0.3445 and min V2 (x) = = 0.6212
{|y|=1.818} bT P1−1 b {|y|=1.818} bT P2−1 b
 
where bT = 2 1 , we estimate the region of attraction by the sets {V1 (x) ≤ 0.34}
and {V2 (x) ≤ 0.62}, shown in Figure 7.10. According to Remark 3.1, the union of
these two sets is an estimate of the region of attraction, which is bigger than the
estimates obtained by the individual Lyapunov functions. 

7.3.2 Popov Criterion

Consider a special case of the system (7.21), given by

ẋ = Ax + Bu, y = Cx ui = −ψi (yi ), 1 ≤ i ≤ m (7.28)

where x ∈ Rn , u, y ∈ Rm , (A, B) is controllable, (A, C) is observable, and ψi is

a locally Lipschitz memoryless nonlinearity that belongs to the sector [0, ki ]. In
10 The value of ε is chosen such that H(s−ε/2) is positive real and Hurwitz, where H(s) = C(sI −

A)−1 B + D. Then, P is calculated by solving a Riccati equation, as described in Exercise 5.14.

11 See equation (B.3).
7.3. ABSOLUTE STABILITY 165

1
V2(x) = 0.62 x2
y = 1.818
0.5
V1(x) = 0.34

0
x1

−0.5
y = −1.818
1
−1.5 −1 −0.5 0 0.5 1 1.5

Figure 7.10: Region of attraction for Example 7.14.

this special case, the transfer function G(s) = C(sI − A)−1 B is strictly proper
and ψ is time invariant and decoupled; that is, ψi (y) = ψi (yi ). Since D = 0, the
feedback connection has a well-defined state model. The following theorem, known
as the multivariable Popov criterion,)is proved
y by using a (Lure-type) Lyapunov
function of the form V = 12 xT P x + γi 0 i ψi (σ) dσ, which is motivated by the
application of a loop transformation that transforms the system (7.28) into the
feedback connection of two passive dynamical systems.

Theorem 7.10 The system (7.28) is absolutely stable if, for 1 ≤ i ≤ m, ψi ∈

[0, ki ], 0 < ki ≤ ∞, and there exists a constant γi ≥ 0, with (1 + λk γi ) = 0
for every eigenvalue λk of A, such that M + (I + sΓ)G(s) is strictly positive real,
where Γ = diag(γ1 , . . . , γm ) and M = diag(1/k1 , . . . , 1/km ). If the sector condition
ψi ∈ [0, ki ] is satisfied only on a set Y ⊂ Rm , then the foregoing conditions ensure
absolute stability with finite domain. 3

Proof: The loop transformation of Figure 7.11 results in a feedback connection of

H̃1 and H̃2 , where H̃1 is a linear system whose transfer function is

M + (I + sΓ)G(s) = M + (I + sΓ)C(sI − A)−1 B

= M + C(sI − A)−1 B + ΓCs(sI − A)−1 B
= M + C(sI − A)−1 B + ΓC(sI − A + A)(sI − A)−1 B
= M + (C + ΓCA)(sI − A)−1 B + ΓCB

Thus, M + (I + sΓ)G(s) can be realized by the state model {A, B, C, D}, where
A = A, B = B, C = C + ΓCA, and D = M + ΓCB. Let λk be an eigenvalue of A
and vk be the associated eigenvector. Then

(C + ΓCA)vk = (C + ΓCλk )vk = (I + λk Γ)Cvk

The condition (1+λk γi ) = 0 implies that (A, C) is observable; hence, the realization
{A, B, C, D} is minimal. If M + (I + sΓ)G(s) is strictly positive real, we can apply
166 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

M
H̃1
+ + +
G(s) G(s) (I + sΓ)
− − +

+
ψ(·) ψ(·) (I + sΓ)−1
+
H̃2

Figure 7.11: Loop transformation.

the Kalman–Yakubovich–Popov lemma to conclude that there are matrices P =

P T > 0, L, and W , and a positive constant ε that satisfy

P A + AT P = −LT L − εP (7.29)
PB = (C + ΓCA) − L W T T
(7.30)
T
W W = 2M + ΓCB + B T C T Γ (7.31)

and V = 12 xT P x is a storage function for H̃1 . One the other hand, it can be verified
)m y
that H̃2 is passive with the storage function i=1 γi 0 i ψi (σ) dσ. Thus, the storage
function for the transformed feedback connection of Figure 7.11 is


m  yi
V = 12 xT P x + γi ψi (σ) dσ
i=1 0

We use V as a Lyapunov function candidate for the original feedback connection

(7.28). The derivative V̇ is given by
1 T 1 T T
V̇ = 2 x P ẋ + 2 ẋ P x + ψ (y)Γẏ
1 T T T
= 2 x (P A + A P )x + x P Bu + ψ T (y)ΓC(Ax + Bu)

Using (7.29) and (7.30) yields

V̇ = − 12 xT LT Lx − 1 T
2 εx P x + xT (C T + AT C T Γ − LT W )u
+ ψ T (y)ΓCAx + ψ (y)ΓCBu
T

Substituting u = −ψ(y) and using (7.31), we obtain

V̇ = − 12 εxT P x − 1
2 (Lx + W u)T (Lx + W u) − ψ(y)T [y − M ψ(y)] ≤ − 12 εxT P x
7.3. ABSOLUTE STABILITY 167

slope = 1/γ ωIm[G(jω)] 0.2 ω Im G slope = 1

0
Re G
−1/k −0.2
Re[G(jω)]
−0.4
−0.6
−0.8
−1
−0.5 0 0.5 1

Figure 7.12: Popov plot. Figure 7.13: Popov plot for Exam-
ple 7.15.

which shows that the origin is globally asymptotically stable. If ψ satisfies the sector
condition only for y ∈ Y , the foregoing analysis will be valid in some neighborhood
of the origin, showing that the origin is asymptotically stable. 2

For M + (I + sΓ)G(s) to be strictly positive real, G(s) must be Hurwitz. As

we have done in the circle criterion, this restriction on G(s) may be removed by
performing a loop transformation that replaces G(s) by G(s)[I + K1 G(s)]−1 . We
will not repeat this idea in general, but will illustrate it by an example. In the
case m = 1, we can test strict positive realness of Z(s) = (1/k) + (1 + sγ)G(s)
graphically. By Lemma 5.1, Z(s) is strictly positive real if G(s) is Hurwitz and
1
+ Re[G(jω)] − γωIm[G(jω)] > 0, ∀ ω ∈ [0, ∞] (7.32)
k
where G(jω) = Re[G(jω)] + jIm[G(jω)]. If we plot Re[G(jω)] versus ωIm[G(jω)]
with ω as a parameter, then (7.32) is satisfied if the plot lies to the right of the line
that intercepts the point −(1/k) + j0 with slope 1/γ. (See Figure 7.12.) Such a
plot is known as the Popov plot, in contrast to the Nyquist plot, which is a plot
of Re[G(jω)] versus Im[G(jω)]. If (7.32) is satisfied only for ω ∈ [0, ∞), while the
left-hand side approaches zero as ω tends to ∞, then we need to analytically verify
that  
1
lim ω 2 + Re[G(jω)] − γωIm[G(jω)] > 0
ω→∞ k
With γ = 0, condition (7.32) reduces to the circle criterion condition Re[G(jω)] >
−1/k, which shows that, for the system (7.28), the conditions of the Popov criterion
are weaker than those of the circle criterion. In other words, with γ > 0, absolute
stability can be established under less stringent conditions.
Example 7.15 The two-dimensional system
ẋ1 = x2 , ẋ2 = −x2 − h(y), y = x1
168 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

would fit the form (7.28) if we took ψ = h, but the matrix A would not be Hurwitz.
Adding and subtracting the term αy to the right-hand side of the second state
equation, where α > 0, and defining ψ(y) = h(y) − αy, the system takes the form
(7.28), with

0 1 0  
A= , B= , C= 1 0
−α −1 1
Assume that h belongs to a sector [α, β], where β > α. Then, ψ belongs to the
sector [0, k], where k = β − α. Condition (7.32) takes the form
1 α − ω 2 + γω 2
+ > 0, ∀ ω ∈ [0, ∞]
k (α − ω 2 )2 + ω 2
For all finite positive values of α and k, this inequality is satisfied by choosing γ > 1.
Even with k = ∞, the foregoing inequality is satisfied for all ω ∈ [0, ∞) and
ω 2 (α − ω 2 + γω 2 )
lim =γ−1>0
ω→∞ (α − ω 2 )2 + ω 2

Hence, the system is absolutely stable for all nonlinearities h in the sector [α, ∞],
where α can be arbitrarily small. Figure 7.13 shows the Popov plot of G(jω) for
α = 1. The plot is drawn only for ω ≥ 0, since Re[G(jω)] and ωIm[G(jω)] are
even functions of ω. The Popov plot asymptotically approaches the line through
the origin of unity slope from the right side. Therefore, it lies to the right of any
line of slope less than one that intersects the real axis at the origin and approaches
it asymptotically as ω tends to ∞. To see the advantage of having γ > 0, let us
take γ = 0 and apply the circle criterion. From the second case of Theorem 7.9,
the system is absolutely stable if the Nyquist plot of G(s) lies to the right of the
vertical line defined by Re[s] = −1/k. Since a portion of the Nyquist plot lies in
the left-half plane, k cannot be arbitrarily large. The maximum permissible value
of k can be determined analytically from the condition
1 α − ω2
+ > 0, ∀ ω ∈ [0, ∞]
k (α − ω 2 )2 + ω 2
√
which yields k < 1 + 2 α. Thus, using the circle criterion, we can√only conclude
absolutely stability for nonlinearities h in the sector [α, 1 + α + 2 α − ε], where
α > 0 and ε > 0 can be arbitrarily small. 

7.4 Exercises
7.1 Consider the feedback connection of Figure 7.1 with
H1 : ẋ1 = −x1 + x2 , ẋ2 = −x31 − x2 + e1 , y1 = x2
H2 : ẋ3 = −x33 + e2 , y2 = x33
Let u2 = 0, u = u1 be the input, and y = y1 be the output.
7.4. EXERCISES 169

(a) Show that the origin of the unforced system is globally asymptotically stable.

(b) Show that the mapping from u to y is finite-gain L2 stable.

7.2 Repeat the previous exercise with

H1 : ẋ1 = x2 , ẋ2 = −x1 − ψ1 (x2 ) + e1 , y1 = x2

H2 : ẋ3 = −aψ2 (x3 ) + e2 , y2 = ψ2 (x3 )

where ψ1 and ψ2 are locally Lipschitz functions, which satisfy zψ1 (z) > bz 2 and
zψ2 (z) > 0 for all z, and a and b are positive constants.

7.3 Consider the system

ẋ1 = −x1 − x2 + u, ẋ2 = ψ1 (x1 ) − ψ2 (x2 ), y = x1

where ψi belong to the sector [αi , βi ] with βi > αi > 0, for i = 1, 2.

(a) Represent the state equation as the feedback connection of two strictly passive
dynamical systems.

(b) Show that the origin of the unforced system is globally exponentially stable.

(c) Find V (x) ≥ 0 that satisfies the Hamilton Jacobi inequality with γ = β1 /α1 .

(d) Let u = −ψ3 (y) + r, where ψ3 belongs to the sector [−δ, δ], with δ > 0. Find an
upper bound on δ such that the mapping from r to y is finite-gain L2 stable.

7.4 Consider the system

ẋ1 = −x1 + x2 , ẋ2 = −ψ1 (x1 ) − x2 + u, y = x2

where ψ1 belongs to the sector [k, ∞] for some k > 0, and let u = −ψ2 (t, y) + r, 
where ψ2 is a time-varying memoryless function that belongs to the sector − 12 , 12 .

(a) Show that the origin of the unforced closed-loop system is globally uniformly
asymptotically stable.

(b) Show that the mapping from r to y is finite-gain L2 stable.

7.5 Consider the system

ẋ1 = −x1 − x2 + u, ẋ2 = x1 − x2 − x32 , y = x1

(a) Show that the origin of the unforced system is globally exponentially stable.
√
(b) Show that the system is finite-gain L∞ stable and the L∞ gain is ≤ 2.
170 CHAPTER 7. STABILITY OF FEEDBACK SYSTEMS

(c) Let u = −ψ(y)+r, where ψ is locally Lipschitz and belongs to the sector [−δ, δ]
with δ > 0. Find an upper bound on δ such that the system with input r and
output y is finite-gain L∞ stable.
7.6 Consider the two-dimensional system
ẋ = x/(1 + x2 ) + z, εż = −z + u, y=x
where the ż-equation represents fast actuator dynamics. The feedback control is
taken as u = −2x + r. Find an upper bound on ε such that the mapping from r to
y is finite-gain Lp stable for every p ∈ [1, ∞].
7.7 Consider the linear time-varying system ẋ = [A + BE(t)C]x, where A is
Hurwitz and E(t) ≤ k ∀ t ≥ 0. Let γ = supω∈R σmax [C(jωI − A)−1 B]. Find
an upper bound on k such that the origin is uniformly asymptotically stable.
7.8 Consider the feedback connection of Figure 7.1 with
H1 : ẋ1 = −x1 +e1 , ẋ2 = x1 −x2 +e1 , y1 = x1 +x2 and H2 : y2 = ψ(e2 )
where ψ is a locally Lipschitz function that satisfies yψ(y) ≥ 0 and |ψ(y)| ≤ 2 for all
y, and ψ(y) = 0 for |y| ≤ 1. Show that the origin of the unforced system is globally
exponentially stable.
7.9 Consider the single-input system ẋ = Ax+BM sat(u/M ) in which the control
input saturates at ±M . Let u = −F x, where A − BF is Hurwitz. We use the circle
criterion to investigate asymptotic stability of the origin of the closed-loop system.
(a) Show that the closed-loop system can be represented in the form of Figure 7.4
with G(s) = −F (sI − A + BF )−1 B and ψ(y) = y − M sat(y/M ).
(b) Show that the origin is globally asymptotically stable if Re[G(jω)] > −1 ∀ ω.
(c) If Re[G(jω)] > −β ∀ ω with β > 1, show that the origin is asymptotically stable
and its region of attraction is estimated by {xT P x ≤ c}, where P satisfies
√
P (A − BF ) + (A − BF )T P = −LT L − εP, P B = −βF T − 2LT
and c > 0 is chosen such that {xT P x ≤ c} ⊂ {|F x| ≤ βM/(β − 1)}.

0 1 0
(d) Estimate the region of attraction when A = ,B= ,
  1 0 1
F = 2 1 , and M = 1.
7.10 Consider the feedback connection of Figure 7.4. Use the circle and Popov
criteria to find a sector [α, β] for absolute stability when
(1) G(s) = 1/[(s + 2)(s − 1)] (2) G(s) = 1/[(s + 2)(s + 1)]
(3) G(s) = 2s/(s2 − 2s + 1) (4) G(s) = 2s/(s2 + 2s + 1)
(5) G(s) = 1/(s + 1)3 (6) G(s) = (s − 1)/[(s + 1)(s + 2)]
(7) G(s) = (s + 2)/[(s2 − 1)(s + 4)] (8) G(s) = (s + 1)2 /[(s + 2)(s − 1)2 ]
Chapter 8

Special Nonlinear Forms

The design of feedback control for nonlinear systems is simplified when the system
takes one of certain special forms. Three such forms are presented in this chapter:
the normal, controller, and observer forms.1 The normal form of Section 8.1 plays a
central role in the forthcoming chapters. It enables us to extend to nonlinear systems
some important notions in feedback control of linear systems, such as the relative
degree of a transfer function, its zeros, and the minimum phase property when
all the zeros have negative real parts. One of the fundamental tools in feedback
control is the use of high-gain feedback to achieve disturbance rejection and low
sensitivity to model uncertainty.2 For linear systems, root locus analysis shows
that a minimum phase, relative-degree-one system can be stabilized by arbitrarily
large feedback gain because, as the gain approaches infinity, one branch of the root
locus approaches infinity on the negative real axis while the remaining branches
approach the zeros of the system. The normal form enables us to show the same
property for nonlinear systems, as we shall see in Section 12.5.1.
The controller form of Section 8.2 is a special case of the normal form, but sig-
nificant in its own sake because a nonlinear system in the controller form can be
converted into an equivalent controllable linear system by state feedback. There-
fore, systems that can be transformed into the controller form are called feedback
linearizable. The observer form of Section 8.3 reduces the design of a nonlinear
observer to a linear design problem.

8.1 Normal Form

Consider the n-dimensional, single-input–single-output system

ẋ = f (x) + g(x)u, y = h(x) (8.1)

1 This chapter is based mostly on [66].
2 See [34].

171
172 CHAPTER 8. SPECIAL NONLINEAR FORMS

where f , g, and h are sufficiently smooth in a domain D ⊂ Rn . The mappings

f : D → Rn and g : D → Rn are called vector fields on D. The derivative ẏ is given
by
∂h def
ẏ = [f (x) + g(x)u] = Lf h(x) + Lg h(x) u
∂x
where
∂h
Lf h(x) = f (x)
∂x
is called the Lie Derivative of h with respect to f or along f . This is the familiar
notion of the derivative of h along the trajectories of the system ẋ = f (x). The new
notation is convenient when we repeat the calculation of the derivative with respect
to the same vector field or a new one. For example, the following notation is used:
∂(Lf h)
Lg Lf h(x) = g(x)
∂x
∂(Lf h)
L2f h(x) = Lf Lf h(x) = f (x)
∂x
∂(Lk−1
f h)
Lkf h(x) = Lf Lk−1
f h(x) = f (x)
∂x
L0f h(x) = h(x)

If Lg h(x) = 0, then ẏ = Lf h(x) is independent of u. If we continue to calculate the

second derivative of y, denoted by y (2) , we obtain
∂(Lf h)
y (2) = [f (x) + g(x)u] = L2f h(x) + Lg Lf h(x) u
∂x
Once again, if Lg Lf h(x) = 0, then y (2) = L2f h(x) is independent of u. Repeating
this process, we see that if h(x) satisfies
ρ−1
f h(x) = 0, for i = 1, 2, . . . , ρ − 1, and Lg Lf
Lg Li−1 h(x) = 0

for some integer ρ, then u does not appear in the equations of y, ẏ, . . . , y (ρ−1) and
appears in the equation of y (ρ) with a nonzero coefficient:

y (ρ) = Lρf h(x) + Lg Lρ−1

f h(x) u

Such integer ρ has to be less than or equal to n,3 and is called the relative degree
of the system, according to the following definition.

Definition 8.1 The nonlinear system (8.1) has relative degree ρ, 1 ≤ ρ ≤ n, in an

open set R ⊂ D if, for all x ∈ R,

f h(x) = 0, for i = 1, 2, . . . , ρ − 1;
Lg Li−1 Lg Lρ−1
f h(x) = 0 (8.2)
3 See [74, Lemma C9] or [66, Lemma 4.1.1] for the proof that ρ ≤ n.
8.1. NORMAL FORM 173

If a system has relative degree ρ, then its input-output map can be converted into
a chain of integrators y (ρ) = v by the state feedback control

1  
u= −Lρf h(x) + v
Lg Lρ−1
f h(x)

Example 8.1 Consider the controlled van der Pol equation (A.13):

ẋ1 = x2 /ε, ẋ2 = ε[−x1 + x2 − 13 x32 + u]

with output y = x1 . Calculating the derivatives of the output, we obtain

ẏ = ẋ1 = x2 /ε, ÿ = ẋ2 /ε = −x1 + x2 − 13 x32 + u

Hence, the system has relative degree two in R2 . For the output y = x2 ,

ẏ = ε[−x1 + x2 − 13 x32 + u]

and the system has relative degree one in R2 . For the output y = 12 (ε2 x21 + x22 ),

ẏ = ε2 x1 ẋ1 + x2 ẋ2 = εx1 x2 + εx2 [−x1 + x2 − 13 x32 + u] = εx22 − (ε/3)x42 + εx2 u

and the system has relative degree one in {x2 = 0}. 

Example 8.2 The field-controlled DC motor (A.25) with δ = 0 and f (x3 ) = bx3
is modeled by the state equation

ẋ1 = d1 (−x1 − x2 x3 + Va ), ẋ2 = d2 [−fe (x2 ) + u], ẋ3 = d3 (x1 x2 − bx3 )

where d1 to d3 are positive constants. For speed control, we choose the output as
y = x3 . The derivatives of the output are given by

ẏ = ẋ3 = d3 (x1 x2 − bx3 )

ÿ = d3 (x1 ẋ2 + ẋ1 x2 − bẋ3 ) = (· · · ) + d2 d3 x1 u

The system has relative degree two in {x1 = 0}. 

Example 8.3 Consider a linear system represented by the transfer function

bm sm + bm−1 sm−1 + · · · + b0 def N (s)

H(s) = =
sn + an−1 sn−1 + · · · + a0 D(s)

where m < n and bm = 0. A state model of the system can be taken as

ẋ = Ax + Bu, y = Cx
174 CHAPTER 8. SPECIAL NONLINEAR FORMS

where
⎡ ⎤ ⎡ ⎤
0 1 0 ... ... 0 0
⎢ 0 0 1 ... ... 0 ⎥ ⎢ ⎥ 0
⎢ ⎥ ⎢ ⎥
⎢ .. .. .. ⎥ ⎢ ⎥ ..
⎢ . . . ⎥ ⎢ ⎥ .
⎢ ⎥ ⎢ ⎥
⎢ .. ⎥ ⎢ ⎥
⎢ . ⎥ ⎢ ⎥
A = ⎢ ⎥, B = ⎢ ⎥
⎢ .. .. ⎥ ⎢ ⎥
⎢ . . ⎥ ⎢ ⎥
⎢ ⎥ ⎢ ⎥
⎢ .. .. ⎥ ⎢ . ⎥
⎢ . . 0 ⎥ ⎢ .. ⎥
⎢ ⎥ ⎢ ⎥
⎣ 0 0 1 ⎦ ⎣ 0 ⎦
−a0 −a1 ... . . . −am ... . . . −an−1 1
 
C = b0 b1 ... ... bm 0 ... 0
This linear state model is a special case of (8.1), where f (x) = Ax, g = B, and
h(x) = Cx. To check the relative degree of the system, we calculate the derivatives
of the output. The first derivative is
ẏ = CAx + CBu
If m = n − 1, then CB = bn−1 = 0 and the system has relative degree one.
Otherwise, CB = 0 and we continue to calculate the second derivative y (2) . Noting
that CA is a row vector obtained by shifting the elements of C one position to the
right, while CA2 is obtained by shifting the elements of C two positions to the right,
and so on, we see that
CAi−1 B = 0, for i = 1, 2, . . . , n − m − 1, and CAn−m−1 B = bm = 0
Thus, u appears first in the equation of y (n−m) , given by
y (n−m) = CAn−m x + CAn−m−1 Bu
and the relative degree ρ = n − m.4 
To probe further into the structure of systems with well-defined relative degrees,
let us start with the linear system of the foregoing example. By Euclidean division,
we can write D(s) = Q(s)N (s) + R(s), where Q(s) and R(s) are the quotient and
remainder polynomials, respectively. From Euclidean division rules, we know that
deg Q = n − m = ρ, deg R < m, and the leading coefficient of Q(s) is 1/bm . With
this representation of D(s), we can rewrite H(s) as
1
N (s) Q(s)
H(s) = = R(s)
Q(s)N (s) + R(s) 1+ 1
Q(s) N (s)

Thus, H(s) can be represented as a negative feedback connection with 1/Q(s) in the
forward path and R(s)/N (s) in the feedback path. (See Figure 8.1.) The ρth-order
4 The“relative degree” definition is consistent with its use in linear control theory, namely, the
difference between the degrees of the denominator and numerator polynomials of H(s).
8.1. NORMAL FORM 175

u + e 1 y
Q(s)
−

w R(s)
N(s)

Figure 8.1: Feedback representation of H(s).

transfer function 1/Q(s)

 has no zeros  and can be realized by the ρth-dimensional
state vector ξ = col y, ẏ, . . . , y (ρ−1) to obtain the state model

ξ˙ = (Ac + Bc λT )ξ + Bc bm e, y = Cc ξ

where (Ac , Bc , Cc ) is a canonical form representation of a chain of ρ integrators,

⎡ ⎤ ⎡ ⎤
0 1 0 ... 0 0
⎢ 0 0 1 ... 0 ⎥ ⎢ 0 ⎥
⎢ ⎥ ⎢ ⎥
⎢ .. .. ..⎥ ⎢ ⎥  
Ac = ⎢
⎢ . . .⎥
⎥ , Bc = ⎢ ... ⎥ , Cc = 1 0 ... 0 0
⎢ .. ⎥ ⎢ ⎥
⎣ ⎣ 0 ⎦
. 0 1 ⎦
0 ... ... 0 0 ρ×ρ 1

and λ ∈ Rρ . Let
η̇ = A0 η + B0 y, w = C0 η

be a (minimal realization) state model of R(s)/N (s). The eigenvalues of A0 are the
zeros of the polynomial N (s), which are the zeros of the transfer function H(s).
From the feedback connection, we see that H(s) can be realized by the state model

η̇ = A0 η + B0 Cc ξ (8.3)
ξ˙ = Ac ξ + Bc (λT ξ − bm C0 η + bm u) (8.4)
y = Cc ξ (8.5)

Our next task is to develop a nonlinear version of the state model (8.3)–(8.5) for
the system (8.1) when it has relative degree ρ. Equivalently, we would like to find
a change of variables z = T (x) such that the new state z can be partitioned into a
ρ-dimensional vector ξ and (n − ρ)-dimensional vector η, where the components of
ξ comprise the output and its derivatives up to y (ρ−1) , while η satisfies a differential
equation whose right-hand side depends on η and ξ, but not on u. We take ξ as in
176 CHAPTER 8. SPECIAL NONLINEAR FORMS

the linear case: ⎡ ⎤ ⎡ ⎤

y h(x)
⎢ y (1) ⎥ ⎢ Lf h(x) ⎥
⎢ ⎥ ⎢ ⎥
ξ=⎢ .. ⎥=⎢ .. ⎥
⎣ . ⎦ ⎣ . ⎦
y (ρ−1) Lρ−1
f h(x)
When ρ = n, there is no η variable and the change of variables is given by
 
z = T (x) = col h(x), Lf h(x), . . . , Ln−1
f h(x) (8.6)

When ρ < n, the change of variables is taken as

⎡ ⎤
φ1 (x)
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥ ⎡ ⎤ ⎡ ⎤
⎢ φn−ρ (x) ⎥ φ(x) η
⎢ ⎥ def
z = T (x) = ⎢
⎢ −−− ⎥ =
⎥ ⎣ − − − ⎦ def
= ⎣ −−− ⎦ (8.7)
⎢ h(x) ⎥ H(x) ξ
⎢ ⎥
⎢ .. ⎥
⎣ . ⎦
Lρ−1
f h(x)

where φ1 to φn−ρ are chosen such that T (x) is a diffeomorphism on a domain

Dx ⊂ R and
∂φi
g(x) = 0, for 1 ≤ i ≤ n − ρ, ∀ x ∈ Dx (8.8)
∂x
so that the u term cancels out in
∂φ
η̇ = [f (x) + g(x)u]
∂x
The next theorem shows that the change of variables (8.6) or (8.7) is well defined,
at least locally.5
Theorem 8.1 Consider the system (8.1) and suppose it has relative degree ρ ≤ n
in R. If ρ = n, then for every x0 ∈ R, a neighborhood N of x0 exists such that
the map T (x) of (8.6), restricted to N , is a diffeomorphism on N . If ρ < n, then,
for every x0 ∈ R, a neighborhood N of x0 and continuously differentiable functions
φ1 (x), . . . , φn−ρ (x) exist such that (8.8) is satisfied for all x ∈ N and the map T (x)
of (8.7), restricted to N , is a diffeomorphism on N . 3
When ρ < n, the change of variables (8.7) transforms (8.1) into
η̇ = f0 (η, ξ) (8.9)
ξ˙ = Ac ξ + Bc [Lρf h(x) + Lg Lρ−1
f h(x)u] (8.10)
y = Cc ξ (8.11)
5 See [74, Theorem 13.1] or [66, Proposition 4.1.3] for the proof of Theorem 8.1. See [66, Section
9.1] for a global version and [66, Section 5.1] for a multiinput-multioutput version.
8.1. NORMAL FORM 177

where ξ ∈ Rρ , η ∈ Rn−ρ , (Ac , Bc , Cc ) is a canonical form representation of a chain

of ρ integrators, and 
∂φ 
f0 (η, ξ) = f (x) (8.12)
∂x x=T −1 (z)

We have kept Lρf h and Lg Lρ−1

f h in (8.10) as functions of x because they are inde-
pendent of the choice of φ. When expressed in the new coordinates, (8.10) is given
by
ξ˙ = Ac ξ + Bc [ψ̃(η, ξ) + γ̃(η, ξ)u]
where
 
 
ψ̃(η, ξ) = Lρf h(x) , γ̃(η, ξ) = Lg Lρ−1
f h(x)
x=T −1 (z) x=T −1 (z)

If x∗ is an open-loop equilibrium point of (8.1), then (η ∗ , ξ ∗ ), defined by

 
η ∗ = φ(x∗ ), ξ ∗ = h(x∗ ) 0 ··· 0

is an equilibrium point of (8.9)–(8.10). If h(x∗ ) = 0, we can transform x∗ into the

origin point (η = 0, ξ = 0) by choosing φ(x) such that φ(x∗ ) = 0, which is always
possible because adding a constant to a function φi that satisfies (8.8) does not
alter this condition, nor the property that T (x) is a diffeomorphism.
Equations (8.9) through (8.11) are said to be in the normal form. This form
decomposes the system into internal dynamics (8.9) and external dynamics (8.10)–
(8.11). The state feedback control

1  
u= −Lρf h(x) + v
Lg Lρ−1
f h(x)

converts the external dynamics into a chain of ρ integrators, y (ρ) = v, and makes
the internal dynamics unobservable from the output y. When y(t) is identically
zero, so is ξ(t). Setting ξ = 0 in (8.9) results in

η̇ = f0 (η, 0) (8.13)

which is called the zero dynamics. The system is said to be minimum phase if
(8.13) has an asymptotically stable equilibrium point in the domain of interest. In
particular, if T (x) is chosen such that the origin (η = 0, ξ = 0) is an equilibrium
point of (8.9)–(8.11), then the system is minimum phase if the origin of the zero
dynamics (8.13) is asymptotically stable. It is useful to know that the zero dynamics
can be characterized in the original coordinates. Noting that

Lρf h(x(t))
y(t) ≡ 0 ⇒ ξ(t) ≡ 0 ⇒ u(t) ≡ −
Lg Lρ−1
f h(x(t))
178 CHAPTER 8. SPECIAL NONLINEAR FORMS

we see that if the output is identically zero, the solution of the state equation must
be confined to the set

Z ∗ = {x ∈ R | h(x) = Lf h(x) = · · · = Lρ−1

f h(x) = 0}

and the input must be



Lρf h(x)
∗ def 
u = u (x) = − 
Lg Lρ−1
f h(x) 
x∈Z ∗

The restricted motion of the system is described by

' ρ
(
L h(x)
ẋ = f ∗ (x) = f (x) − g(x)
def f
Lg Lρ−1
f h(x) x∈Z ∗

In the special case ρ = n, the normal form reduces to

ż = Ac z + Bc [Lnf h(x) + Lg Ln−1

f h(x)u], y = Cc z (8.14)

In this case, the system has no zero dynamics and, by default, is said to be minimum
phase.

Example 8.4 Consider the controlled van der Pol equation (A.13):

ẋ1 = x2 /ε, ẋ2 = ε[−x1 + x2 − 13 x32 + u], y = x2

We have seen in Example 8.1 that the system has relative degree one in R2 . Taking
ξ = y and η = x1 , we see that the system is already in the normal form. The zero
dynamics are given by the equation ẋ1 = 0, which does not have an asymptotically
stable equilibrium point. Hence, the system is not minimum phase. 

Example 8.5 The system

2 + x23
ẋ1 = −x1 + u, ẋ2 = x3 , ẋ3 = x1 x3 + u, y = x2
1 + x23

has an open-loop equilibrium point at the origin. The derivatives of the output are

ẏ = ẋ2 = x3 , ÿ = ẋ3 = x1 x3 + u

Therefore, the system has relative degree two in R3 . To characterize the zero
dynamics, restrict x to Z ∗ = {x2 = x3 = 0} and take
 
 
L 2
h(x)  
∗
= −x1 x3 
f
u = u (x) = −  =0
Lg Lf h(x)  
x∈Z ∗ ∗ x∈Z
8.2. CONTROLLER FORM 179

This process yields ẋ1 = −x1 , which shows that the system is minimum phase. To
transform the system into the normal form, we want to choose φ(x) such that

∂φ
φ(0) = 0, g(x) = 0
∂x
and T (x) = col (φ(x), x2 , x3 ) is a diffeomorphism on some domain containing the
origin. The partial differential equation

∂φ 2 + x23 ∂φ
· + = 0, φ(0) = 0
∂x1 1 + x23 ∂x3

can be solved by separation of variables to obtain φ(x) = x1 − x3 − tan−1 x3 .

The mapping T (x) is a global diffeomorphism because it is proper and [∂T /∂x] is
nonsingular for all x. Thus, the normal form
 
  2 + ξ22
η̇ = − η + ξ2 + tan−1 ξ2 1 + ξ2
1 + ξ22
ξ˙1 = ξ2
 
ξ˙2 = η + ξ2 + tan−1 ξ2 ξ2 + u
y = ξ1

is defined globally. 

Example 8.6 The field-controlled DC motor of Example 8.2 has relative degree
two in R = {x1 = 0} with h = x3 and Lf h = d3 (x1 x2 − bx3 ). To characterize the
zero dynamics, restrict x to

Z ∗ = {x ∈ R | x3 = 0 and x1 x2 − bx3 = 0} = {x ∈ R | x2 = x3 = 0}

and take u = u∗ (x), to obtain ẋ1 = d1 (−x1 + Va ). The zero dynamics have an
asymptotically stable equilibrium point at x1 = Va . Hence, the system is min-
imum phase. To transform it into the normal form, we want to find a function
φ(x) such that [∂φ/∂x]g = d2 ∂φ/∂x2 = 0 and T = col (φ(x), x3 , d3 (x1 x2 − bx3 ))
is a diffeomorphism on some domain Dx ⊂ R. The choice φ(x) = x1 − Va satis-
fies ∂φ/∂x2 = 0, makes T (x) a diffeomorphism on {x1 > 0}, and transforms the
equilibrium point of the zero dynamics to the origin. 

8.2 Controller Form

A nonlinear system is in the controller form if its state model is given by

ẋ = Ax + B[ψ(x) + γ(x)u] (8.15)

180 CHAPTER 8. SPECIAL NONLINEAR FORMS

where (A, B) is controllable and γ(x) is a nonsingular matrix for all x in the domain
of interest. The system (8.15) can be converted into the controllable linear system
ẋ = Ax + Bv
by the state feedback control
u = γ −1 (x)[−ψ(x) + v]
where γ −1 (x) is the inverse matrix of γ(x). Therefore, any system that can be
represented in the controller form is said to be feedback linearizable.
Example 8.7 An m-link robot is modeled by (A.35):
ẋ1 = x2 , ẋ2 = M −1 (x1 )[u − C(x1 , x2 )x2 − Dx2 − g(x1 )]
Taking x = col (x1 , x2 ), the state model takes the form (8.15) with

0 Im 0
A= , B= , ψ = −M −1 (Cx2 + Dx2 + g), γ = M −1
0 0 Im

Even if the state model is not in the controller form, the system will be feedback
linearizable if it can be transformed into that form. In this section, we characterize
the class of feedback linearizable single-input systems6
ẋ = f (x) + g(x)u (8.16)
where f and g are sufficiently smooth vector fields on a domain D ⊂ Rn . From
the previous section we see that the system (8.16) is feedback linearizable in a
neighborhood of x0 ∈ D if a (sufficiently smooth) function h exists such that the
system
ẋ = f (x) + g(x)u, y = h(x) (8.17)
has relative degree n in a neighborhood of x0 . Theorem 8.1 ensures that the map
 
T (x) = col h(x), Lf h(x), . . . , Ln−1
f h(x)

restricted to a neighborhood N of x0 , is a diffeomorphism on N and the change of

variables z = T (x) transforms the system into the normal form
ż = Ac z + Bc [ψ̃(z) + γ̃(z)u], y = Cc z
where (Ac , Bc , Cc ) is a canonical form representation of a chain of n integrators. On
the other hand, if (8.16) is feedback linearizable in a neighborhood N of x0 ∈ D,
then there is a change of variables ζ = S(x) that transforms the system into
ζ̇ = Aζ + B[ψ̄(ζ) + γ̄(ζ)u]
6 See [66, Chapter 5] for multi-input systems.
8.2. CONTROLLER FORM 181

where (A, B) is controllable and γ̄(ζ) = 0 in S(N ). For any controllable pair (A, B),
we can find a nonsingular matrix M that transforms (A, B) into the controllable
canonical form:7 M AM −1 = Ac + Bc λT ; M B = Bc . The change of variables
def
z = M ζ = M S(x) = T (x)
transforms (8.16) into
ż = Ac z + Bc [ψ̃(z) + γ̃(z)u]
−1
where γ̃(z) = γ̄(M z) and ψ̃(z) = ψ̄(M −1 z) + λT z.

In summary, the system (8.16) is feedback linearizable in a neighborhood of

x0 ∈ D if and only if a (sufficiently smooth) function h(x) exists such that the system
(8.17) has relative degree n in a domain Dx ⊂ D, with x0 ∈ Dx , or, equivalently, h
satisfies the partial differential equations
Lg Li−1
f h(x) = 0, i = 1, 2, . . . , n − 1 (8.18)
subject to the condition
Lg Ln−1
f h(x) = 0 (8.19)
for all x ∈ Dx . The existence of h can be characterized by necessary and sufficient
conditions on the vector fields f and g. These conditions use the notions of Lie
brackets and involutive distributions, which we introduce next.

For two vector fields f and g on D ⊂ Rn , the Lie bracket [f, g] is a third vector
field defined by
∂g ∂f
[f, g](x) = f (x) − g(x)
∂x ∂x
where [∂g/∂x] and [∂f /∂x] are Jacobian matrices. The Lie bracket [f, g](x) is
invariant under the change of variables z = T (x) in the sense that if
 
∂T  ∂T 
f˜(z) = f (x) and g̃(z) = g(x)
∂x x=T −1 (z) ∂x x=T −1 (z)

then8 
∂T 
˜
[f , g̃](z) = [f, g](x)
∂x x=T −1 (z)
We may repeat bracketing of g with f . The following notation is used to simplify
this process:
ad0f g(x) = g(x)
adf g(x) = [f, g](x)
adkf g(x) = [f, adk−1
f g](x), k≥1
It is obvious that [f, g] = −[g, f ] and for constant vector fields f and g, [f, g] = 0.
7 See, for example, [114].
8 See [98, Proposition 2.30] for the proof of this property.
182 CHAPTER 8. SPECIAL NONLINEAR FORMS

Example 8.8 Let

x2 0
f (x) = , g=
− sin x1 − x2 x1

Then,

0 0 x2 0 1 0
[f, g] = −
1 0 − sin x1 − x2 − cos x1 −1 x1

−x1 def
= = adf g
x1 + x2

−1 0 x2 0 1 −x1
[f, adf g] = −
1 1 − sin x1 − x2 − cos x1 −1 x1 + x2

−x1 − 2x2 def

= = ad2f g
x1 + x2 − sin x1 − x1 cos x1

Example 8.9 If f (x) = Ax and g is a constant vector field, then adf g = [f, g] =
−Ag, ad2f g = [f, adf g] = −A(−Ag) = A2 g, and adkf g = (−1)k Ak g. 

For vector fields f1 , f2 , . . . , fk on D ⊂ Rn , let

Δ(x) = span{f1 (x), f2 (x), . . . , fk (x)}

be the subspace of Rn spanned by the vectors f1 (x), f2 (x), . . . , fk (x) at any fixed
x ∈ D. The collection of all vector spaces Δ(x) for x ∈ D is called a distribution
and referred to by
Δ = span{f1 , f2 , . . . , fk }
The dimension of Δ(x), defined by

dim(Δ(x)) = rank [f1 (x), f2 (x), . . . , fk (x)]

may vary with x, but if Δ = span{f1 , . . . , fk }, where {f1 (x), . . . , fk (x)} are linearly
independent for all x ∈ D, then dim(Δ(x)) = k for all x ∈ D. In this case, we say
that Δ is a nonsingular distribution on D, generated by f1 , . . . , fk . A distribution
Δ is involutive if
g1 ∈ Δ and g2 ∈ Δ ⇒ [g1 , g2 ] ∈ Δ
If Δ is a nonsingular distribution on D, generated by f1 , . . . , fk , then Δ is involutive
if and only if
[fi , fj ] ∈ Δ, ∀ 1 ≤ i, j ≤ k
One dimensional distributions, Δ = span{f }, and distributions generated by con-
stant vector fields are always involutive.
8.2. CONTROLLER FORM 183

Example 8.10 Let D = R3 and Δ = span{f1 , f2 }, where

⎡ ⎤ ⎡ ⎤
2x2 1
f1 = ⎣ 1 ⎦ , f2 = ⎣ 0 ⎦
0 x2

It can be verified that dim(Δ(x)) = 2 for all x. The Lie bracket [f1 , f2 ] ∈ Δ if and
only if rank [f1 (x), f2 (x), [f1 , f2 ](x)] = 2, for all x ∈ D. However,
⎡ ⎤
2x2 1 0
rank [f1 (x), f2 (x), [f1 , f2 ](x)] = rank ⎣ 1 0 0 ⎦ = 3, ∀x∈D
0 x2 1
Hence, Δ is not involutive. 

Example 8.11 Let D = {x ∈ R3 | x21 + x23 = 0} and Δ = span{f1 , f2 }, where

⎡ ⎤ ⎡ ⎤
2x3 −x1
f1 = ⎣ −1 ⎦ , f2 = ⎣ −2x2 ⎦
0 x3

It can be verified that dim(Δ(x)) = 2 for all x ∈ D and

⎡ ⎤
2x3 −x1 −4x3
rank [f1 (x), f2 (x), [f1 , f2 ](x)] = rank ⎣ −1 −2x2 2 ⎦ = 2, ∀x∈D
0 x3 0

Therefore, [f1 , f2 ] ∈ Δ. Since [f2 , f1 ] = −[f1 , f2 ], Δ is involutive. 

We are now ready to characterize the class of feedback linearizable systems.9

Theorem 8.2 The system (8.16) is feedback linearizable in a neighborhood of x0 ∈

D if and only if there is a domain Dx ⊂ D, with x0 ∈ Dx , such that
1. the matrix G(x) = [g(x), adf g(x), . . . , adn−1
f g(x)] has rank n for all x ∈ Dx ;

2. the distribution D = span {g, adf g, . . . , adn−2

f g} is involutive in Dx . 3

The conditions of the theorem ensure that for each x0 ∈ D there is a function
h(x) satisfying (8.18)–(8.19) in some neighborhood N of x0 and the system is feed-
back linearizable in N . They do not guarantee feedback linearizability in a given
domain. However, as we solve the partial differential equations (8.18) for h(x), it
is usually possible to determine a domain Dx in which the system is feedback lin-
earizable. This point is illustrated in the next three examples. In all examples, we
assume that the system (8.16) has an equilibrium point x∗ when u = 0, and choose
h(x) such that h(x∗ ) = 0. Consequently, z = T (x) maps x = x∗ into z = 0.
9 See[74, Theorem 13.2] or [66, Theorem 4.2.3] for the proof of Theorem 8.2. See also [66,
Theorem 5.2.3] for the multiinput version of the theorem.
184 CHAPTER 8. SPECIAL NONLINEAR FORMS

Example 8.12 Consider the system

a sin x2 0 def
ẋ = + u = f (x) + gu
−x21 1

We have

∂f −a cos x2
adf g = [f, g] = − g=
∂x 0
The matrix

0 −a cos x2
[g, adf g] =
1 0
has rank two for all x such that cos x2 = 0. The distribution span{g} is involutive.
Hence, the conditions of Theorem 8.2 are satisfied in the domain Dx = {|x2 | < π/2}.
To find the change of variables that transforms the system into the form (8.15), we
want to find h(x) that satisfies

∂h ∂(Lf h)
g = 0, g = 0, and h(0) = 0
∂x ∂x
From the condition [∂h/∂x]g = 0, we have [∂h/∂x]g = [∂h/∂x2 ] = 0. Thus, h must
be independent of x2 . Therefore, Lf h(x) = [∂h/∂x1 ]a sin x2 . The condition

∂(Lf h) ∂(Lf h) ∂h
g= = a cos x2 = 0
∂x ∂x2 ∂x1
is satisfied in Dx if (∂h/∂x1 ) = 0. The choice h = x1 yields

x1
z = T (x) =
a sin x2

and the transformed system


ż1 = z2 , ż2 = a2 − z22 (−z12 + u)

is well defined for −a < z2 < a. 

Example 8.13 A single link manipulator with flexible joints and negligible damp-
ing can be represented by the four-dimensional model [135]

ẋ = f (x) + gu

where ⎡ ⎤ ⎡⎤
x2 0
⎢ −a sin x1 − b(x1 − x3 ) ⎥ ⎢ 0 ⎥
f (x) = ⎢
⎣
⎥,
⎦ g=⎢ ⎥
⎣ 0 ⎦
x4
c(x1 − x3 ) d
8.2. CONTROLLER FORM 185

and a, b, c, and d are positive constants. The unforced system has an equilibrium
point at x = 0. We have
⎡ ⎤ ⎡ ⎤
0 0
∂f ⎢ 0 ⎥ ∂f ⎢ bd ⎥
adf g = [f, g] = − g=⎢ ⎥ ⎢
⎣ −d ⎦ , adf g = [f, adf g] = − ∂x adf g = ⎣ 0 ⎦
2 ⎥
∂x
0 −cd
⎡ ⎤
−bd
∂f 2 ⎢ 0 ⎥
ad3f g = [f, ad2f g] = − ad g = ⎢
⎣ cd ⎦
⎥
∂x f
0
The matrix ⎡ ⎤
0 0 0 −bd
⎢ 0 0 bd 0 ⎥
[g, adf g, ad2f g, ad3f g] = ⎢
⎣ 0
⎥
−d 0 cd ⎦
d 0 −cd 0
has full rank for all x ∈ R4 . The distribution span(g, adf g, ad2f g) is involutive since
g, adf g, and ad2f g are constant vector fields. Thus, the conditions of Theorem 8.2
are satisfied for all x ∈ R4 . To transform the system into the controller form, we
want to find h(x) that satisfies

∂(Li−1
f h) ∂(L3f h)
g = 0, i = 1, 2, 3, g = 0, and h(0) = 0
∂x ∂x
From the condition [∂h/∂x]g = 0, we have (∂h/∂x4 ) = 0, so we must choose h
independent of x4 . Therefore,

∂h ∂h ∂h
Lf h(x) = x2 + [−a sin x1 − b(x1 − x3 )] + x4
∂x1 ∂x2 ∂x3

From the condition [∂(Lf h)/∂x]g = 0, we have

∂(Lf h) ∂h
=0 ⇒ =0
∂x4 ∂x3
So, we choose h independent of x3 . Therefore, Lf h simplifies to

∂h ∂h
Lf h(x) = x2 + [−a sin x1 − b(x1 − x3 )]
∂x1 ∂x2
and
∂(Lf h) ∂(Lf h) ∂(Lf h)
L2f h(x) = x2 + [−a sin x1 − b(x1 − x3 )] + x4
∂x1 ∂x2 ∂x3
186 CHAPTER 8. SPECIAL NONLINEAR FORMS

Finally,
∂(L2f h) ∂(Lf h) ∂h
=0 ⇒ =0 ⇒ =0
∂x4 ∂x3 ∂x2
and we choose h independent of x2 . Hence,
∂(L2f h) ∂(L2f h) ∂(L2f h)
L3f h(x) = x2 + [−a sin x1 − b(x1 − x3 )] + x4
∂x1 ∂x2 ∂x3
and the condition [∂(L3f h)/∂x]g = 0 is satisfied whenever (∂h/∂x1 ) = 0. Therefore,
we take h(x) = x1 . The change of variables

z1 = h(x) = x1
z2 = Lf h(x) = x2
z3 = L2f h(x) = −a sin x1 − b(x1 − x3 )
z4 = L3f h(x) = −ax2 cos x1 − b(x2 − x4 )

transforms the system into the controller form

ż1 = z2 , ż2 = z3 , ż3 = z4 , ż4 = −(a cos z1 + b + c)z3 + a(z22 − c) sin z1 + bdu

Unlike the previous example, in the current one the state equation in the z-coordinates
is valid globally because z = T (x) is a global diffeomorphism. 

Example 8.14 The field-controlled DC motor (A.25) with δ = 0 and f (x3 ) = bx3
is modeled by ẋ = f (x) + gu, where
⎡ ⎤ ⎡ ⎤
d1 (−x1 − x2 x3 + Va ) 0
f =⎣ −d2 fe (x2 ) ⎦ , g = ⎣d2 ⎦
d3 (x1 x2 − bx3 ) 0

in which d1 to d3 and Va are positive and b ≥ 0. Assume that fe (x2 ) is twice

continuously differentiable in an open interval J ⊂ R. We have
⎡ ⎤ ⎡ ⎤
d1 d2 x3 d1 d2 x3 (d1 + d2 fe (x2 ) − bd3 )
adf g = ⎣ d22 fe (x2 ) ⎦ ; ad2f g = ⎣ d32 (fe (x2 ))2 − d32 f2 (x2 )fe (x2 ) ⎦
−d2 d3 x1 d1 d2 d3 (x1 − Va ) − d22 d3 x1 fe (x2 ) − bd2 d23 x1

det G = det[g, adf g, ad2f g] = −2d21 d32 d3 x3 (x1 − a)(1 − bd3 /d1 )
where a = 12 Va /(1 − bd3 /d1 ). We assume that bd3 /d1 = bTa /Tm < 1 so that a > 0,
which is reasonable since Ta is typically much smaller than Tm . Hence, G has rank
three for x1 = a and x3 = 0. The distribution D = span{g, adf g} is involutive if
[g, adf g] ∈ D. We have
⎡ ⎤ ⎡ ⎤
0 0 d1 d2 0
∂(adf g)
[g, adf g] = g=⎣ 0 d22 fe (x2 ) 0 ⎦ ⎣ d2 ⎦ = d22 fe (x2 )g
∂x
−d2 d3 0 0 0
8.3. OBSERVER FORM 187

Hence, D is involutive and the conditions of Theorem 8.2 are satisfied in the domain
Dx = {x1 > a, x2 ∈ J, x3 > 0}. We proceed now to find a function h that satisfies
(8.18) and (8.19). We take the desired operating speed as ω0 > 0, which is achieved
at the equilibrium point x∗ = col (x∗1 , x∗2 , x∗3 ), where x∗1 = (Va + Va2 − 4bω02 )/2,
x∗2 = bω0 /x∗1 , and x∗3 = ω0 . We assume that Va2 > 4bω02 , x∗1 > a, and x∗2 ∈ J. We
want to find h(x) that satisfies

∂h ∂(Lf h) ∂(L2f h)
g = 0; g = 0; g = 0
∂x ∂x ∂x
with h(x∗ ) = 0. From the condition [∂h/∂x]g = 0 ⇔ ∂h/∂x2 = 0, we see that h
must be independent of x2 . Therefore,

∂h ∂h
Lf h(x) = d1 (−x1 − x2 x3 + Va ) + d3 (x1 x2 − bx3 )
∂x1 ∂x3

From the condition [∂(Lf h)/∂x]g = 0 ⇔ ∂(Lf h)/∂x2 = 0, obtain

∂h ∂h
−d1 x3 + d3 x1 =0
∂x1 ∂x3

which is satisfied by h = d3 x21 + d1 x23 + c and we choose c = −d3 (x∗1 )2 − d1 (x∗3 )2 to

satisfy the condition h(x∗ ) = 0. With this choice of h, Lf h and L2f h are given by

Lf h(x) = 2d1 d3 x1 (Va − x1 ) − 2bd1 d3 x23

L2f h(x) = 2d21 d3 (Va − 2x1 )(−x1 − x2 x3 + Va ) − 4bd1 d23 x3 (x1 x2 − bx3 )
Hence,
∂(L2f h) ∂(L2f h)
g = d2 = 4d21 d2 d3 (1 − bd3 /d1 )x3 (x1 − a)
∂x ∂x2
and the condition [∂(L2f h)/∂x]g = 0 is satisfied in Dx . 

8.3 Observer Form

A nonlinear system represented by

ẋ = Ax + ψ(u, y), y = Cx (8.20)

where (A, C) is observable, is said to be in the observer form. Suppose the input u
and output y are available on line, but not the state x. Assuming that the matrices
A and C and the nonlinear function ψ(u, y) are known, we can implement the
nonlinear observer
x̂˙ = Ax̂ + ψ(u, y) + H(y − C x̂)
188 CHAPTER 8. SPECIAL NONLINEAR FORMS

to estimate x by x̂. The estimation error x̃ = x − x̂ satisfies the linear equation

x̃˙ = (A − HC)x̃
Because (A, C) is observable, we can design H to assign the eigenvalues of A − HC
in the open left-half plane, ensuring that limt→∞ x̃(t) = 0. The next two examples
show physical systems whose models are in the observer form.
Example 8.15 A single link manipulator with flexible joints and negligible damp-
ing can be represented by the model [135]
⎡ ⎤ ⎡ ⎤
x2 0
⎢ −a sin x1 − b(x1 − x3 ) ⎥ ⎢ 0 ⎥
ẋ = ⎢
⎣
⎥+⎢ ⎥
⎦ ⎣ 0 ⎦u
x4
c(x1 − x3 ) d
where a, b, c, and d are positive constants. Taking y = x1 as the output, it can be
seen that the system is in the form (8.20) with
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
0 1 0 0 0 1
⎢ −b 0 b 0 ⎥ ⎢ −a sin y ⎥ ⎢ 0 ⎥
A=⎣ ⎢ ⎥ , ψ=⎣ ⎢ ⎥ C =⎣ ⎥
T ⎢
0 0 0 1 ⎦ 0 ⎦, 0 ⎦
c 0 −c 0 du 0
and that (A, C) is observable. 
Example 8.16 In equation (A.47), the inverted pendulum is modeled by
ẋ1 = x2 , ẋ2 = a(sin x1 + u cos x1 )
Taking y = x1 as the output, the system is in the observer form (8.20) with

0 1 0 T 1
A= , ψ= , C =
0 0 a(sin y + u cos y) 0

In general, we may have to perform a change of variables to bring a system
into the observer form, if possible. In this section we consider the n-dimensional
single-output nonlinear system

m
ẋ = f (x) + gi (x)ui , y = h(x) (8.21)
i=1

where f , h, and g1 to gm are smooth functions in a domain D ⊂ Rn , and study

the existence of a diffeomorphism T (x) such that the change of variables z = T (x)
transforms the system (8.21) into the observer form

m
ż = Ac z + φ(y) + γi (y)ui , y = Cc z (8.22)
i=1
8.3. OBSERVER FORM 189

where
⎡ ⎤
0 1 0 ... 0
⎢ 0 0 1 ... 0 ⎥
⎢ ⎥
⎢ .. .. ..⎥  
Ac = ⎢
⎢ . . .⎥,
⎥ Cc = 1 0 ... 0 0
⎢ .. ⎥
⎣ . 0 1 ⎦
0 ... ... 0 0

There is no loss of generality in taking (Ac , Cc ) in the observable canonical form

since any observable pair (A, C) can be transformed into the observable canonical
form by a similarity transformation. We start by setting u = 0 and study the
existence of a change of variables that transforms the system

ẋ = f (x), y = h(x) (8.23)

into the observer form

ż = Ac z + φ(y), y = Cc z (8.24)
Define the map Φ(x) by
 
Φ(x) = col h(x), Lf h(x), . . . , Ln−1
f h(x)

and note that the existence of a diffeomorphism that transforms the system (8.23)
into the form (8.24) is possible only if the Jacobian matrix [∂Φ/∂x] is nonsingular
over the domain of interest. This is so because, in the z-coordinates,
 
Φ̃(z) = Φ(x)|x=T −1 (z) = col h̃(z), Lf˜h̃(z), . . . , Ln−1
˜
f
h̃(z)

where f˜(z) = Ac z + φ(Cc z), h̃(z) = Cc z, and the Jacobian matrix

⎡ ⎤
1 0 ··· 0
⎢ ∗ 1 0 0 ⎥
∂ Φ̃ ⎢ ⎢ ..
⎥
.. ⎥
=⎢ . . ⎥
∂z ⎢ ⎥
⎣ ∗ ··· ∗ 1 0 ⎦
∗ ··· ∗ 1

is nonsingular. By the chain rule, [∂ Φ̃/∂z] = [∂Φ/∂x][∂T −1 /∂z] and the Jacobian
matrix [∂T −1 /∂z] is nonsingular because T is a diffeomorphism. Therefore, [∂Φ/∂x]
is nonsingular. This implies that Φ(x) is a diffeomorphism. Noting that

Φ(x) = col(y, ẏ, . . . , y (n−1) )

we see that the system (8.23) is instantaneously observable because its state x can
be recovered from the output and its derivatives using the inverse of Φ.
190 CHAPTER 8. SPECIAL NONLINEAR FORMS

Let τ (x) be the unique solution of

∂Φ
τ = b, where b = col (0, . . . , 0, 1) (8.25)
∂x
Equation (8.25) is equivalent to
Lτ Lkf h(x) = 0, for 0 ≤ k ≤ n − 2 and Lτ Ln−1
f h(x) = 1 (8.26)
10
From (8.26) it can be shown that
Ladkf τ h(x) = 0, for 0 ≤ k ≤ n − 2 and Ladn−1 τ h(x) = (−1)n−1 (8.27)
f

Define
τk = (−1)n−k adn−k f τ, for 1 ≤ k ≤ n
and suppose τ1 (x), · · · , τn (x) are linearly independent in the domain of interest.
Let T (x) be the solution of the partial differential equation
∂T  
τ1 τ2 · · · τn = I (8.28)
∂x
It is clear that T (x) is a local diffeomorphism in the domain of interest. We will
show that the change of variables z = T (x) transforms (8.23) into (8.24). From
(8.28) we see that
⎡ ⎤
0
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ 0 ⎥
∂T ∂T ⎢ ⎥
τ = ek = ⎢ 1 ⎥
def
∂x
τk =
∂x
(−1)n−k adn−k
f ⎢ ⎥ ← kth row
⎢ 0 ⎥
⎢ ⎥
⎢ . ⎥
⎣ .. ⎦
0
For 1 ≤ k ≤ n − 1, the foregoing expression yields
∂T ∂T
ek = (−1)n−k adn−k
f τ (x) = (−1)n−k [f, adn−k−1
f τ ](x)
∂x ∂x
∂ f˜
= (−1)n−k [f˜(z), (−1)n−k−1 ek+1 ] = ek+1
∂z
where we used the property that Lie brackets are invariant under the change of
variables z = T (x). Hence, the Jacobian matrix [∂ f˜/∂z] takes the form
⎡ ⎤
∗ 1 0 ... 0
⎢ ∗ 0 1 ... 0 ⎥
⎢ ⎥
∂ f˜ ⎢ ⎢
.. . .. ⎥
=⎢ . . . . ⎥
⎥
∂z ⎢ . ⎥
⎣ .. 0 1 ⎦
∗ ... ... 0 0
10 Apply Lemma C.8 of [74].
8.3. OBSERVER FORM 191

By integrating [∂ f˜/∂z], it can be seen that f˜(z) = Ac z + φ(z1 ), for some function
φ. Consider now h̃(z) = h(T −1 (z)). By the chain rule,

∂ h̃ ∂h ∂T −1
=
∂z ∂x ∂z
From (8.28), it is clear that
∂T −1  
= τ1 τ2 ··· τn x=T −1 (z)
∂z
Therefore

∂ h̃ ∂h  
= (−1)n−1 n−1
adf τ, (−1) n−2
adf τ, · · · τ 
n−2
∂z ∂x x=T −1 (z)
 
= (−1)n−1 Ladn−1 τ h, (−1)n−2 Ladn−2 τ h, · · · Lτ h 
f f −1 x=T (z)

Using (8.27), we obtain

∂ h̃  
= 1, 0, ··· 0
∂z
Hence h̃(z) = Cc z. The following theorem gives necessary and sufficient conditions
for the existence of a change of variables that transforms the system (8.23) into the
observer form (8.24).11
Theorem 8.3 There is a domain Dx ⊂ Rn and a diffeomorphism T (x) on Dx such
that the change of variables z = T (x) transforms the system (8.23) into the observer
form (8.24) if and only if there is a domain D0 ∈ Rn such that

∂Φ
rank (x) = n, ∀ x ∈ D0
∂x

where Φ = col(h, Lf h, . . . , Ln−1

f h), and the solution τ of (8.25) satisfies

[adif τ, adjf τ ] = 0, for 0 ≤ i, j ≤ n − 1

The diffeomorphism T (x) is the solution of (8.28). 3

)m
We turn now to the input term i=1 gi (x)u and study under what conditions
will 
∂T 
g̃i (z) = gi (x)
∂x x=T −1 (z)

be independent of z2 to zn . We note that

∂T ∂g̃i
[gi (x), adn−k−1
f τ (x)] = [g̃i (z), (−1)n−k−1 ek+1 ] = (−1)n−k
∂x ∂zk+1
11 See [66, Theorem 5.2.1] for the proofs of Theorem 8.3, Corollary 8.1, and a global version.
192 CHAPTER 8. SPECIAL NONLINEAR FORMS

Hence, [∂g̃i /∂zk+1 ] is zero if only if the Lie bracket [gi , adn−k−1
f τ ] is zero. Con-
sequently, the vector field g̃i (z) is independent of z2 to zn if and only if the Lie
brackets [gi , adkf τ ] are zero for 0 ≤ k ≤ n − 2. In fact, if [gi , adn−1
f τ ] = 0 as well,
then g̃i (z) will be independent of z1 ; that is, it will be a constant vector field and
the vector γi in (8.22) will be independent of y. These observations are summarized
in the following corollary.
Corollary 8.1 Suppose the assumptions of Theorem 8.3 are satisfied and let T (x)
be the solution of (8.28). Then, the change of variables z = T (x) transforms the
system (8.21) into the observer form (8.22) if and only if
[gi , adkf τ ] = 0, , for 0 ≤ k ≤ n − 2 and 1 ≤ i ≤ m
Moreover, if for some i the foregoing condition is strengthened to
[gi , adkf τ ] = 0, , for 0 ≤ k ≤ n − 1
then the vector field γi , of (8.22), is constant. 3
Example 8.17 Consider the system

β1 (x1 ) + x2 b1
ẋ = + u, y = x1
f2 (x1 , x2 ) b2
where β1 and f2 are smooth functions and b1 and b2 are constants.

h(x) x1 ∂Φ 1 0
Φ(x) = = ⇒ =
Lf h(x) β1 (x1 ) + x2 ∂x ∗ 1

0 0
Hence, rank [∂Φ/∂x] = 2 for all x. The solution of [∂Φ/∂x]τ = is τ = .
1 1
The conditions of Theorem 8.3 are satisfied if [τ, adf τ ] = 0.

∂f ∗ 1 0 1
adf τ = [f, τ ] = − τ = − ∂f2 = − ∂f2
∂x ∗ ∂x 2
1 ∂x2
' (

∂adf τ 0 0 0
[τ, adf τ ] = τ =− ∂ 2 f2 ∂ 2 f2
∂x ∂x1 ∂x2 ∂x 2 1
2

Hence
∂ 2 f2
[τ, adf τ ] = 0 ⇔ = 0 ⇔ f2 (x1 , x2 ) = β2 (x1 ) + x2 β3 (x1 )
∂x22
The conditions of Corollary 8.1 are satisfied if [g, τ ] = 0, which holds since both
g and τ are constant vector fields. To obtain a constant vector field γ in the z-
coordinates, we need the additional condition

0 0 b1
0 = [g, adf τ ] = ∂β3
− ∂x 1
0 b2
8.3. OBSERVER FORM 193

which is satisfied if β3 is constant or b1 = 0. To find the change of variables, we

need to solve the partial differential equation
⎡ ∂T1 ∂T1 ⎤
∂x1 ∂x2

⎣ ⎦ 1 0 1 0
=
∂T2 ∂T2 β3 (x1 ) 1 0 1
∂x1 ∂x2

Taking T (0) = 0, we have

∂T1 ∂T1
= 0 and = 1 ⇒ T1 = x1
∂x2 ∂x1
 x1
∂T2 ∂T2
= 1 and + β3 (x1 ) = 0 ⇒ T2 (x) = x2 − β3 (σ) dσ
∂x2 ∂x1 0
The map T (x) is a global diffeomorphism since it is proper and [∂T /∂x] is nonsin-
gular for all x. For any z ∈ R2 we can solve for x by using
 z1
x1 = z1 , x2 = z2 + β3 (σ) dσ
0
The state model in the z-coordinates is given by

y

0 1 β1 (y) + 0 β3 (σ) dσ b1
ż = z+ + u
0 0 β2 (y) − β1 (y)β3 (y) b2 − b1 β3 (y)
 
y = 1 0 z

We conclude the section by focusing on a special case of the observer form for
single-input systems. Consider the n-dimensional system
ẋ = f (x) + g(x)u, y = h(x) (8.29)
where f (0) = 0 and h(0) = 0. Suppose the system satisfies the assumptions of
Corollary 8.1 with
[g, adkf τ ] = 0, , for 0 ≤ k ≤ n − 1
Then, there is a change of variables z = T (x) that transforms the system into the
form
ż = Ac z + φ(y) + γu, y = Cc z (8.30)
where φ(0) = 0 and γ is constant. Suppose further that the system (8.29) has
relative degree ρ ≤ n. Then, by differentiation of y, it can be seen that γ takes the
form  T
γ = 0, . . . , 0, γρ , . . . , γn , with γρ = 0 (8.31)
A system in the form (8.30) with γ as in (8.31) is said to be in the output feedback
form. We leave it to the reader to verify that if ρ < n, the zero dynamics will be
linear and the system will be minimum phase if the polynomial
γρ sn−ρ + · · · + γn−1 s + γn (8.32)
is Hurwitz.
194 CHAPTER 8. SPECIAL NONLINEAR FORMS

8.4 Exercises
8.1 For each of the following systems,

(a) Find the relative degree and show that the system is minimum phase.

(b) Transform the system into the normal form. Is the change of variables valid
globally?

(c) Under what additional conditions is the system feedback linearizable?

(d) Under the conditions of part (c), find a change of variables that transforms
the system into the controller form. Under what additional conditions is the
change of variables a global diffeomorphism?

(1) ẋ1 = x2 , ẋ2 = −ϕ(x1 ) − x2 + u, y = x1 + 2x2 , where ϕ(0) = 0, x1 ϕ(x1 ) ≥

β1 x21 ∀ x1 , and β1 > 0.

(2) ẋ1 = −x1 − x2 + u, ẋ2 = ϕ1 (x1 ) − ϕ2 (x2 ), y = x1 + x2 , where ϕ1 and ϕ2

belong to the sectors [αi , βi ] with βi > αi > 0, for i = 1, 2.

(3) ẋ1 = −ϕ1 (x1 ) + ϕ2 (x2 ), ẋ2 = −x1 − x2 + x3 , ẋ3 = −x3 + u, y = x2 , where
ϕ1 and ϕ2 are smooth and belong to the sectors [αi , βi ] with βi > αi > 0, for
i = 1, 2.

8.2 For each of the following systems, find the relative degree and determine if the
system is minimum phase.

(1) The boost converter (A.16) when the output is x1 .

(2) The boost converter (A.16) when the output is x2 .

(3) The biochemical reactor (A.19), with (A.20), when the output is y = x1 − 1.

(4) The biochemical reactor (A.19), with (A.20), when the output is y = x2 − 1.

(5) The magnetic levitation system (A.30)–(A.32) when the output is x1 .

(6) The electrostatic microactuator (A.33) when the output is x1

(7) The electrostatic microactuator (A.33) when the output is x3

(8) The inverted pendulum on a cart (A.41)–(A.44) when the output is x1 .

(9) The inverted pendulum on a cart (A.41)–(A.44) when the output is x3 .

(10) The TORA system (A.49)–(A.52) when the output is x1 .

(11) The TORA system (A.49)–(A.52) when the output is x3 .

8.4. EXERCISES 195

8.3 For each of the following systems investigate whether the system is feedback
linearizable. If it is so, find a change of variables that transforms the system in the
controller form and determine its domain of validity.
(1) The boost converter (A.16).
(2) The biochemical reactor (A.19) with ν defined by (A.20).
(3) The magnetic levitation system (A.30)–(A.32).
(4) The electrostatic microactuator (A.33).
(5) The inverted pendulum (A.47).
8.4 Consider the field-controlled DC motor (A.25) with δ = 0.
(a) Show that neglecting the armature circuit dynamics by setting Ta = 0 results
in the two-dimensional model
ẋ1 = a1 [−fe (x1 ) + u], ẋ2 = a2 [x1 (Va − x1 x2 ) − f (x2 )]
where x2 and x3 of (A.25) are renamed x1 and x2 , respectively.
(b) Study the relative degree when y = x2 .
(c) Study feedback linearization.
8.5 Consider the system
x2
ẋ1 = x1 + , ẋ2 = −x2 + x3 , ẋ3 = x1 x2 + u
1 + x21
(a) If the output is x2 , find the relative degree of the system and determine if it is
minimum phase.
(b) Repeat (a) if the output is x1 .
(c) Show that the system is feedback linearizable.
(d) Find a change of variables that transforms the system into the controller form
and determine its domain of validity.
8.6 A semitrailer-like vehicle can be modeled by the state equation
ẋ1 = tan(x3 )
tan(x2 ) 1
ẋ2 = − + tan(u)
a cos(x3 ) b cos(x2 ) cos(x3 )
tan(x2 )
ẋ3 =
a cos(x3 )
where a and b are positive constants. Restrict u to Du = {|u| < π/2} and x to
Dx = {|x2 | < π/2, |x3 | < π/2}. Then, cos(x2 ) = 0, cos(x3 ) = 0, and tan(u) is
invertible.
196 CHAPTER 8. SPECIAL NONLINEAR FORMS

(a) Show that the change of variables w = tan(u)/(b cos(x2 ) cos(x3 )) transforms
the system into the form
⎡ ⎤ ⎡ ⎤
tan(x3 ) 0
ẋ = f (x) + gw, where f = ⎣ − tan(x2 )/(a cos(x3 )) ⎦ , g = ⎣ 1 ⎦
tan(x2 )/(a cos(x3 )) 0

(b) Show that the system is feedback linearizable.

(c) Find a change of variables that transforms the system into the controller form
and determine its domain of validity.

8.7 Show that the electrostatic microactuator (A.33) with output y = col(x1 , x3 ).
is in the observer form.

8.8 For each of the following systems, determine if the system is transformable into
the observer form. If yes, find the change of variables and determine its domain of
validity.

(1) ẋ1 = x2 , ẋ2 = −3x21 x2 − x1 − x31 − x2 + u, y = x1 .

(2) ẋ1 = x2 , ẋ2 = −x22 + x3 , ẋ3 = x21 − 2x32 + 2x2 x3 + u, y = x1

(3) The van der Pol equation: ẋ1 = x2 , ẋ2 = −x1 + x2 − 13 x32 + u, y = x1

(4)

ẋ1 = β1 (x1 ) + x2 + b1 u
ẋ2 = β2 (x1 ) + β3 (x1 )x2 + x3 + b2 u
ẋ3 = β4 (x1 ) + β5 (x1 )x2 + ax3 + b3 u
y = x1

The functions β1 to β5 are smooth.

8.9 Consider an n-dimensional system of the form (8.30), which has relative degree
ρ < n. Show that the vector γ takes the form (8.31) and the system is minimum
phase if the polynomial (8.32) is Hurwitz.
Chapter 9

State Feedback Stabilization

Stabilization of equilibrium points is the core task in feedback control of nonlinear

systems. The problem is important in its own sake and is at the heart of other
control problems such as tracking and regulation. Stabilization is treated in three
chapters, starting with state feedback stabilization in this chapter, followed by ro-
bust stabilization in Chapter 10, and output feedback stabilization in Chapter 12.1

9.1 Basic Concepts

The state feedback stabilization problem for the system

ẋ = f (x, u)

is to design a feedback controller u = φ(x) such that the origin x = 0 is an asymp-

totically stable equilibrium point of the closed-loop system

ẋ = f (x, φ(x))

The control u = φ(x) is usually called “static feedback” because it is a memoryless

function of x. We may use a dynamic state feedback controller u = φ(x, z) where z
is the state of a dynamic system of the form

ż = g(x, z)

In this case the origin to be stabilized is (x = 0, z = 0). Examples of dynamic state

feedback control arise when we use integral or adaptive control.
While the standard problem is defined as stabilization of an equilibrium point
at the origin, we can use the same formulation to stabilize the system with respect
1 For further reading on stabilization, see [8, 19, 31, 32, 33, 36, 40, 43, 66, 67, 71, 75, 80, 81, 86,
91, 107, 115, 121, 137, 141].

197
198 CHAPTER 9. STATE FEEDBACK STABILIZATION

to an arbitrary point xss . For that we need the existence of a steady-state control
uss that can maintain equilibrium at xss ; namely,

0 = f (xss , uss )

The change of variables

xδ = x − xss , uδ = u − uss

results in
def
ẋδ = f (xss + xδ , uss + uδ ) = fδ (xδ , uδ )
where fδ (0, 0) = 0. We can now proceed to the solve the standard stabilization
problem for the system
ẋδ = fδ (xδ , uδ )
where uδ is designed as feedback control of xδ . The overall control u = uδ + uss has
a feedback component uδ and a feedforward component uss .
When a linear system is stabilized by linear feedback, the origin of the closed-
loop system is globally asymptotically stable. This is not the case for nonlinear
systems where different stabilization notions can be introduced. If the feedback
control guarantees asymptotic stability of the origin but does not prescribe a re-
gion of attraction, we say that the feedback control achieves local stabilization. If it
guarantees that a certain set is included in the region of attraction or an estimate
of the region of attraction is given, we say that the feedback control achieves re-
gional stabilization. Global stabilization is achieved if the origin of the closed-loop
system is globally asymptotically stable. If feedback control does not achieve global
stabilization, but can be designed such that any given compact set (no matter how
large) can be included in the region of attraction, we say that the feedback control
achieves semiglobal stabilization. It is typical in this case that the feedback control
depends on parameters that can be adjusted to enlarge the region of attraction.
These four stabilization notions are illustrated by the next example.
Example 9.1 Suppose we want to stabilize the scalar system

ẋ = x2 + u

by state feedback. Linearization at the origin results in the linear system ẋ = u,

which can be stabilized by u = −kx with k > 0. When this control is applied to
the nonlinear system, it results in

ẋ = −kx + x2

whose linearization at the origin is ẋ = −kx. Thus, by Theorem 3.2, the origin
is asymptotically stable, and we say that u = −kx achieves local stabilization. In
this example, it is not hard to see that the region of attraction is the set {x < k}.
With this information, we say that u = −kx achieves regional stabilization. By
9.2. LINEARIZATION 199

increasing k, we can expand the region of attraction. In fact, any compact set of
the form {−a ≤ x ≤ b}, with positive a and b, can be included it in the region of
attraction by choosing k > b. Hence, u = −kx achieves semiglobal stabilization. It
is important to notice that u = −kx does not achieve global stabilization. In fact,
for any finite k, there is a part of the state space (that is, x ≥ k), which is not in
the region of attraction. While semiglobal stabilization can include any compact
set in the region of attraction, the controller is dependent on the given set and will
not necessarily work with a bigger set. For a given b, we can choose k > b. Once k
is fixed and the controller is implemented, if the initial state happens to be in the
region {x > k}, the solution x(t) will diverge to infinity. Global stabilization can be
achieved by the nonlinear controller u = −x2 − kx, which cancels the nonlinearity
and yields the closed-loop system ẋ = −kx. 

9.2 Linearization
The stabilization problem is well understood for the linear system

ẋ = Ax + Bu

where the state feedback control u = −Kx preserves linearity of the open-loop
system, and the origin of the closed-loop system

ẋ = (A − BK)x

is asymptotically stable if and only if the matrix A − BK is Hurwitz. Thus, the

state feedback stabilization problem reduces to a problem of designing a matrix K to
assign the eigenvalues of A−BK in the open left-half complex plane. Linear control
theory2 confirms that the eigenvalues of A−BK can be arbitrarily assigned (subject
only to the constraint that complex eigenvalues are in conjugate pairs) provided
the pair (A, B) is controllable. Even if some eigenvalues of A are not controllable,
stabilization is still possible if the uncontrollable eigenvalues have negative real
parts. In this case, the pair (A, B) is called stabilizable, and the uncontrollable
(open-loop) eigenvalues of A will be (closed-loop) eigenvalues of A − BK.
For nonlinear systems, the problem is more difficult and less understood. A
simple approach is to use linearization. Consider the system

ẋ = f (x, u) (9.1)

where f (0, 0) = 0 and f (x, u) is continuously differentiable in a domain Dx × Du

of Rn × Rm containing the origin (x = 0, u = 0). We want to design a state
feedback controller u = φ(x) to stabilize the system. Linearization of (9.1) about
(x = 0, u = 0) results in the linear system

ẋ = Ax + Bu (9.2)
2 See, for example, [2], [22], [59], [82], or [114].
200 CHAPTER 9. STATE FEEDBACK STABILIZATION

where  
∂f  ∂f 
A= (x, u) ; B= (x, u)
∂x x=0,u=0 ∂u x=0,u=0

Assume the pair (A, B) is controllable, or at least stabilizable. Design a matrix K to

assign the eigenvalues of A − BK to desired locations in the open left-half complex
plane. Now apply the linear state feedback control u = −Kx to the nonlinear
system (9.1) to obtain the closed-loop system

ẋ = f (x, −Kx) (9.3)

Clearly, the origin x = 0 is an equilibrium point of the closed-loop system. The

linearization of (9.3) about the origin is given by

∂f ∂f
ẋ = (x, −Kx) + (x, −Kx) (−K) x = (A − BK)x
∂x ∂u x=0

Since A − BK is Hurwitz, it follows from Theorem 3.2 that the origin is an ex-
ponentially stable equilibrium point of the closed-loop system (9.3). Moreover, we
can find a quadratic Lyapunov function for the closed-loop system. Let Q be any
positive-definite symmetric matrix and solve the Lyapunov equation

P (A − BK) + (A − BK)T P = −Q

for P . Since (A − BK) is Hurwitz, the Lyapunov equation has a unique positive
definite solution (Theorem 3.7). The quadratic function V (x) = xT P x is a Lya-
punov function for the closed-loop system in the neighborhood of the origin. We
can use V (x) to estimate the region of attraction.

Example 9.2 Consider the pendulum equation (A.1):

θ̈ + sin θ + bθ̇ = cu

Suppose we want to design u to stabilize the pendulum at an angle θ = δ1 . For

the pendulum to maintain equilibrium at θ = δ1 , the steady-state control uss must
satisfy
sin δ1 = cuss
Choose the state variables as x1 = θ − δ1 , x2 = θ̇ and set uδ = u − uss . The state
equation
ẋ1 = x2 , ẋ2 = −[sin(x1 + δ1 ) − sin δ1 ] − bx2 + cuδ
is in the standard form (9.1), where f (0, 0) = 0. Linearization at the origin yields

0 1 0 1 0
A= = ; B=
− cos(x1 + δ1 ) −b x =0 − cos δ1 −b c
1
9.3. FEEDBACK LINEARIZATION 201

The pair (A, B) is controllable. Taking K = [k1 k2 ], it can be verified that A − BK

is Hurwitz for
cos δ1 b
k1 > − , k2 > −
c c
The control u is given by
sin δ1 sin δ1
u= − Kx = − k1 (θ − δ1 ) − k2 θ̇
c c


9.3 Feedback Linearization

Consider the nonlinear system
ẋ = f (x) + G(x)u
where f (0) = 0, x ∈ Rn , and u ∈ Rm . Let D be a domain of Rn containing the
origin and suppose there is a diffeomorphism T (x) on D, with T (0) = 0, such that
the change of variables z = T (x) transforms the system into the controller form
ż = Az + B[ψ(x) + γ(x)u]
where (A, B) is controllable and γ(x) is a nonsingular matrix for all x ∈ D. Let
γ −1 (x) be the inverse matrix of γ(x) for each x. The feedback control
u = γ −1 (x)[−ψ(x) + v]
cancels the nonlinearity and converts the system into
ż = Az + Bv
Hence, the origin z = 0 can be stabilized by v = −Kz, where A − BK is Hurwitz.
The closed-loop system, in the z-coordinates, is
ż = (A − BK)z
and the origin z = 0 is exponentially stable. In the x-coordinates, the control is
given by
u = γ −1 (x)[−ψ(x) − KT (x)]
and the closed-loop system is

ẋ = f (x) + G(x)γ −1 (x)[−ψ(x) − KT (x)] = fc (x)

def

The origin in the x-coordinates inherits exponential stability of the origin in the
z-coordinates because T (x) is a diffeomorphism in the neighborhood of x = 0. In
particular, since
∂T
ż = (x)ẋ = (A − BK)z
∂x
202 CHAPTER 9. STATE FEEDBACK STABILIZATION

we have
∂T
(x)fc (x) = (A − BK)T (x)
∂x
Calculating the Jacobian matrix of each side of the preceding equation at x = 0
and using the fact that fc (0) = 0, we obtain
∂fc ∂T
(0) = J −1 (A − BK)J, where J = (0)
∂x ∂x
The matrix J is nonsingular and the similarity transformation J −1 (A − BK)J pre-
serves the eigenvalues of A − BK. Hence, J −1 (A − BK)J is Hurwitz and x = 0 is
exponentially stable by Theorem 3.2. If D = Rn and T (x) is a global diffeomor-
phism, then x = 0 will be globally asymptotically stable. If T (x) is a diffeomor-
phism on a domain D ⊂ Rn so that the equation ż = (A − BK)z is valid in the
domain T (D), then the region of attraction can be estimated in the z coordinates
by Ωc = {z T P z ≤ c}, where P = P T > 0 is the solution of the Lyapunov equation

P (A − BK) + (A − BK)T P = −Q, for some Q = QT > 0 (9.4)

and c > 0 is chosen such that Ωc ⊂ T (D). The estimate in the x-coordinates is
given by T −1 (Ωc ) = {T T (x)P T (x) ≤ c}.
Example 9.3 We saw in Example 8.12 that the system

ẋ1 = a sin x2 , ẋ2 = −x21 + u

can be transformed into the controller form by the change of variables

x1
z = T (x) =
a sin x2

where T (x) is a diffeomorphism on D = {|x2 | < π/2} and T (D) = {|z2 | < a}. Take
K = σ 2 2σ , with σ > 0, to assign the eigenvalues of A − BK at −σ, −σ. The
Lyapunov equation (9.4) is satisfied with

3σ 2 σ 2σ 3 0
P = , Q=
σ 1 0 2σ
The region of attraction is estimated by

{3σ 2 x21 + 2σax1 sin x2 + a2 sin2 x2 ≤ c}

where3
a2 2a2
c < min z T P z =
=
|z2 |=a   0 3
0 1 P −1
1

3 See Section 3.6 and equation (B.3).
9.3. FEEDBACK LINEARIZATION 203

Feedback linearization is based on exact mathematical cancellation of the non-

linear terms γ and ψ, which requires knowledge of γ, ψ, and T . This is almost
impossible for several practical reasons such as model simplification, parameter un-
certainty, and computational errors. Most likely, the controller will be implementing
functions γ̂, ψ̂, and T̂ , which are approximations of γ, ψ, and T ; that is to say, the
actual control will be
u = γ̂ −1 (x)[−ψ̂(x) − K T̂ (x)]
The closed-loop system under this feedback control is given by

ż = Az + B{ψ(x) + γ(x)γ̂ −1 (x)[−ψ̂(x) − K T̂ (x)]}

By adding and subtracting the term BKz to the right-hand side we can rewrite the
closed-loop system as
ż = (A − BK)z + BΔ(z) (9.5)
where
! "
Δ(z) = ψ(x) + γ(x)γ̂ −1 (x)[−ψ̂(x) − K T̂ (x)] + KT (x)
x=T −1 (z)

The system (9.5) is a perturbation of the nominal system ż = (A − BK)z. The

effect of the perturbation term BΔ(z) can be investigated by using the techniques
of Section 4.2 when Δ(0) = 0 and those of Section 4.3 when Δ(0) = 0. The next
lemma summarizes the results of such investigation.

Lemma 9.1 Suppose the closed-loop system (9.5) is well defined in a domain Dz ⊂
Rn containing the origin, and Δ(z) ≤ kz for all z ∈ Dz . Let P be the solution
of the Lyapunov equation (9.4) with Q = I. Then, the origin z = 0 is exponentially
stable if
1
k< (9.6)
2P B
If Dz = Rn , the origin will be globally exponentially stable. Furthermore, suppose
Δ(z) ≤ kz+δ for all z ∈ Dz , where k satisfies (9.6) and δ > 0. Take r > 0 such
that Br ⊂ Dz . Then there exist positive constants c1 and c2 such that if δ < c1 r and
z(0) ∈ {z T P z ≤ λmin (P )r2 }, z(t) will be ultimately bounded by δc2 . If Dz = Rn ,
z(t) will be globally ultimately bounded by δc2 for any δ > 0.

Proof: With V (z) = z T P z, we have

V̇ = z T [P (A − BK) + (A − BK)T P ]z + 2z T P BΔ(z)

≤ −z2 + 2P B z Δ(z)

If Δ(z) ≤ kz + δ, then

V̇ ≤ −z2 + 2kP B z2 + 2δP B z

= −(1 − θ1 )z2 − θ1 z2 + 2kP B z2 + 2δP B z
204 CHAPTER 9. STATE FEEDBACK STABILIZATION

where θ1 ∈ (0, 1) is chosen close enough to one such that k < θ1 /(2P B). Conse-
quently,
V̇ ≤ −(1 − θ1 )z2 + 2δP B z
If Δ(z) ≤ kz, we set δ = 0 in the preceding inequality and conclude that the
origin z = 0 is exponentially stable, or globally exponentially stable if Dz = Rn . If
δ > 0,
2δP B def
V̇ ≤ −(1 − θ1 )(1 − θ2 )z2 , ∀ z ≥ = δc0
(1 − θ1 )θ2

where θ2 ∈ (0, 1). Theorem 4.5 shows that if δc0 < r λmin (P )/λ max (P ) and z(0) ∈
{z P z ≤ λmin (P )r }, then z(t) is ultimately bounded by δc0 λmax (P )/λmin (P ).
T 2

If Dz = Rn , the constant r can be chosen arbitrarily large; hence the ultimate bound
holds for any initial state z(0) with no restriction on how large δc0 is. 2

Example 9.4 Consider the pendulum equation

ẋ1 = x2 , ẋ2 = − sin(x1 + δ1 ) − bx2 + cu

from Example 9.2. A linearizing-stabilizing feedback control is given by

 
1
u= [sin(x1 + δ1 ) − (k1 x1 + k2 x2 )]
c

where k1 and k2 are chosen such that

0 1
A − BK =
−k1 −(k2 + b)

is Hurwitz. Suppose, due to uncertainties in the parameter c, the actual control is

 
1
u= [sin(x1 + δ1 ) − (k1 x1 + k2 x2 )]
ĉ

where ĉ is an estimate of c. The closed-loop system is given by

ẋ1 = x2 , ẋ2 = −k1 x1 − (k2 + b)x2 + Δ(x)

where  
c − ĉ
Δ(x) = [sin(x1 + δ1 ) − (k1 x1 + k2 x2 )]
ĉ
The error term Δ(x) satisfies the bound |Δ(x)| ≤ kx + δ globally with
     
 c − ĉ   c − ĉ 
k =   1 + k2 + k2 , δ =  
 ĉ  | sin δ1 |
ĉ  1 2
9.3. FEEDBACK LINEARIZATION 205

The constants k and δ are measures of the size of the error in estimating c. Let P
be the solution of the Lyapunov equation P (A − BK) + (A − BK)T P = −I. If

1 p11 p12
k<  2 , where P =
2 p12 + p222 p12 p22

then the solutions are globally ultimately bounded and the ultimate bound is pro-
portional to δ. If sin δ1 = 0, the foregoing bound on k ensures global exponential
stability of the origin. Satisfying the foregoing inequality is not a problem when the
uncertainty is sufficiently small, but for large uncertainty it constraints the choice of
K; it might even be impossible to choose K for which the inequality is satisfied. To
illustrate this point let us neglect friction by setting b = 0 and consider uncertainty
in the mass m. Then, ĉ = mo /m̂, where m̂ is an estimate of m. The constants k
and δ are given by

 
 m̂ − m 
k = Δm 1 + k12 + k22 , δ = Δm | sin δ1 |, where Δm =  
m 
 
Let K = σ 2 2σ to assign eigenvalues of A − BK at −σ and −σ for a positive
σ. By solving the Lyapunov
 equation (9.4) with Q = I, it can be verified that the
inequality k < 1/(2 p212 + p222 ) is equivalent to

2σ 3
Δm < √ 
[1 + σ σ 2 + 4] 4σ 2 + (σ 2 + 1)2

By plotting the right-hand side as a function of σ it can be seen that it increases

monotonically with σ until it reaches a maximum value of 0.3951, after which it
decreases monotonically towards zero. Hence, the inequality cannot be satisfied if
Δm > 0.3951 and for every value of Δm less than 0.3951, the inequality is satisfied
for a range of values of σ. For example, for Δm = 0.3, σ should be restricted to
1.37 < σ < 5.57. 

The foregoing Lyapunov analysis shows that the stabilizing component of the
feedback control u = γ −1 (x)[−ψ(x)−KT (x)] achieves a certain degree of robustness
to model uncertainty. We will see in Chapter 10 that the stabilizing component can
be designed to achieve a higher degree of robustness by exploiting the fact that the
perturbation term BΔ(z) in (9.5) belongs to the range space of the input matrix B.
Such perturbation is said to satisfy the matching condition. The techniques of the
Chapter 10 can guarantee robustness to arbitrarily large Δ(z) provided an upper
bound on Δ(z) is known.

The basic philosophy of feedback linearization is to cancel the nonlinear terms

of the system. Aside from the issues of whether or not we can cancel the nonlinear
terms, effect of uncertainties, implementation factors, and so on, we should examine
the philosophy itself: Is it a good idea to cancel nonlinear terms? Our motivation to
206 CHAPTER 9. STATE FEEDBACK STABILIZATION

do so has been mathematically driven. We wanted to linearize the system to make

it more tractable and use the relatively well-developed linear control theory. From
a performance viewpoint, however, a nonlinear term could be “good” or “bad” and
the decision whether we should use feedback to cancel a nonlinear term is, in reality,
problem dependent. Let us use a couple of examples to illustrate this point.

Example 9.5 Consider the scalar system4

ẋ = ax − bx3 + u

where a and b are positive constants. The feedback control

u = −(k + a)x + bx3 , k>0

cancels the nonlinear term −bx3 and results in the closed-loop system ẋ = −kx.
But, −bx3 provides “nonlinear damping” and even without any feedback control
it would guarantee boundedness of the solution despite the fact that the origin is
unstable. So, why should we cancel it? If we simply use the linear control

u = −(k + a)x, k>0

we will obtain the closed-loop system ẋ = −kx − bx3 whose origin is globally expo-
nentially stable and its trajectories approach the origin faster than the trajectories
of ẋ = −kx. Moreover, the linear control uses less control effort and is simpler to
implement. 

Example 9.6 Consider the two-dimensional system

ẋ1 = x2 , ẋ2 = −h(x1 ) + u

where h(0) = 0 and x1 h(x1 ) > 0 for all x1 = 0. The system is feedback linearizable
and a linearizing–stabilizing feedback control can be taken as

u = h(x1 ) − (k1 x1 + k2 x2 )

where k1 and k2 are chosen to assign the closed-loop eigenvalues at desired locations
in the left-half complex plane. On the other hand, with y = x2 as  x the output, the
system is passive with the positive definite storage function V = 0 1 h(z) dz + 12 x22 .
In particular, V̇ = yu. The feedback control

u = −σ(x2 )

where σ is any locally Lipschitz function that satisfies σ(0) = 0 and yσ(y) > 0 for
y = 0, results in
V̇ = −x2 σ(x2 ) ≤ 0
4 The example is taken from [42].
9.4. PARTIAL FEEDBACK LINEARIZATION 207

Because
x2 (t) ≡ 0 ⇒ ẋ2 (t) ≡ 0 ⇒ h(x1 (t)) ≡ 0 ⇒ x1 (t) ≡ 0
asymptotic stability of the origin follows from the invariance principle. The control
u = −σ(x2 ) has two advantages over the linearizing feedback control. First, it does
not use a model of the nonlinear function h. Hence, it is robust to uncertainty in
modeling h. Second, the flexibility in choosing the function σ can be used to reduce
the control effort. For example, we can meet any constraint of the form |u| ≤ k,
by choosing u = −k sat(x2 ). However, the control u = −σ(x2 ) cannot arbitrarily
assign the rate of decay of x(t). Linearization of the closed-loop system at the origin
yields the characteristic equation

s2 + σ  (0)s + h (0) = 0

One of thetwo roots of the foregoing equation cannot be moved to the left of
Re[s] = − h (0). Feedback controllers that exploit passivity will be discussed in
Section 9.6. 

These two examples make the point that there are situations where nonlinearities
are beneficial and cancelling them should not be an automatic choice. We should
try our best to understand the effect of the nonlinear terms and decide whether or
not cancellation is appropriate. Admittedly, this is not an easy task.

9.4 Partial Feedback Linearization

Consider the nonlinear system

ẋ = f (x) + G(x)u

where f (0) = 0, x ∈ Rn , and u ∈ Rm . Let D be a domain of Rn containing the

origin and suppose there is a diffeomorphism T (x) on D, with T (0) = 0, such that
the change of variables

η T1 (x)
z= = T (x) =
ξ T2 (x)

transforms the system into the form

η̇ = f0 (η, ξ), ξ˙ = Aξ + B[ψ(x) + γ(x)u] (9.7)

where ξ ∈ Rρ , η ∈ Rn−ρ , (A, B) is controllable, and γ(x) is a nonsingular matrix

for all x ∈ D. It follows from the preceding assumptions that f0 (0, 0) = 0 and
ψ(0) = 0. We assume that f0 , ψ, and γ are locally Lipschitz. Our goal is to design
state feedback control to stabilize the origin. The form (9.7) is motivated by the
normal form of Section 8.1, but the output equation is dropped since it plays no
208 CHAPTER 9. STATE FEEDBACK STABILIZATION

role in state feedback stabilization. We do not restrict our discussion in this section
to single-input systems or to a pair (A, B) in the controllable canonical form.

The state feedback control

u = γ −1 (x)[−ψ(x) + v]

reduces (9.7) to
η̇ = f0 (η, ξ), ξ˙ = Aξ + Bv (9.8)
Equation ξ˙ = Aξ + Bv can be stabilized by v = −Kξ, where (A − BK) is Hurwitz.
The closed-loop system

η̇ = f0 (η, ξ), ξ˙ = (A − BK)ξ (9.9)

is a cascade connection of the system η̇ = f0 (η, ξ) with the linear system ξ˙ =

(A − BK)ξ. It is clear that asymptotic stability of the origin of η̇ = f0 (η, 0) is a
necessary condition for asymptotic stability of the origin of the cascade connection
because with the initial state ξ(0) = 0 we have ξ(t) ≡ 0 and the trajectories of the
cascade connection coincide with those of η̇ = f0 (η, 0). The next lemma shows that
this condition is sufficient as well.

Lemma 9.2 The origin of the cascade connection (9.9) is asymptotically (respec-
tively, exponentially) stable if the origin of η̇ = f0 (η, 0) is asymptotically (respec-
tively, exponentially) stable. 3

The lemma is proved by constructing composite Lyapunov functions as in Sec-

tion C.1. In the case of asymptotic stability, the Lyapunov function is bV1 (η) +

ξ T P ξ, with sufficiently small b > 0, where V1 (η) is provided by (the converse
Lyapunov) Theorem 3.9 and P = P T > 0 is the solution of the Lyapunov equation
P (A−BK)+(A−BK)T P = −I. In the case of exponential stability, the Lyapunov
function is bV1 (η) + ξ T P ξ, with sufficiently small b > 0, where V1 (η) is provided by
(the converse Lyapunov) Theorem 3.8.

Lemma 9.2 shows that a minimum phase, input–output linearizable, single-

input–single-output system can be stabilized by the state feedback control

u = γ −1 (x)[−ψ(x) − KT2 (x)] (9.10)

which is independent of T1 (x). Therefore, it is independent of the function φ that

satisfies the partial differential equation (7.8).

Lemma 9.2 is a local result. The next two examples show that global asymptotic
stability of the origin of η̇ = f0 (η, ξ) is not sufficient for global asymptotic stability
of the origin of the cascade connection.
9.4. PARTIAL FEEDBACK LINEARIZATION 209

Example 9.7 Consider the two-dimensional system5

η̇ = −η + η 2 ξ, ξ˙ = v
The origin of η̇ = −η is globally exponentially stable. With v = −kξ, k > 0, we can
see by linearization that the origin (η, ξ) = (0, 0) of the full system is exponentially
stable. However, it is not globally asymptotically stable. This fact can be seen as
follows. Taking ν = ηξ and noting that
ν̇ = η ξ˙ + η̇ξ = −kηξ − ηξ + η 2 ξ 2 = −(1 + k)ν + ν 2
we see that the set {ηξ < 1 + k} is positively invariant. The boundary ηξ = 1 + k,
is a trajectory of the system where η(t) = (1 + k)ekt and ξ(t) = e−kt . Inside
{ηξ < 1 + k}, ν(t) will be strictly decreasing and after a finite time T , ν(t) ≤ 12
for all t ≥ T . Then, η η̇ ≤ − 12 η 2 , for all t ≥ T , which shows that the trajectory
approaches the origin as t tends to infinity. Hence, {ηξ < 1 + k} is the region of
attraction. While the origin is not globally asymptotically stable, we can include
any compact set in the region of attraction by choosing k large enough. Thus,
v = −kξ can achieve semiglobal stabilization. 
If the origin of η̇ = f0 (η, 0) is globally asymptotically stable, one might think
that the system (9.8) can be globally stabilized, or at least semiglobally stabilized,
by designing the linear feedback control v = −Kξ to assign the eigenvalues of
(A − BK) far to the left in the complex plane so that the solution of ξ˙ = (A − BK)ξ
decays to zero arbitrarily fast. Then, the solution of ξ˙ = f0 (η, ξ) will quickly
approach the solution of η̇ = f0 (η, 0), which is well behaved because its origin is
globally asymptotically stable. It may even appear that this strategy was used in
the preceding example. The next example shows why such strategy may fail.
Example 9.8 Consider the three-dimensional system6
η̇ = − 12 (1 + ξ2 )η 3 , ξ˙1 = ξ2 , ξ˙2 = v
The linear feedback control
def
v = −k 2 ξ1 − 2kξ2 = −Kξ
assigns the eigenvalues of

0 1
A − BK =
−k 2 −2k
at −k and −k. The exponential matrix
⎡ ⎤
(1 + kt)e−kt te−kt
e(A−BK)t
= ⎣ ⎦
−k 2 te−kt (1 − kt)e −kt

5 The example is taken from [19].

6 The example is taken from [137].
210 CHAPTER 9. STATE FEEDBACK STABILIZATION

shows that as k → ∞, the solution ξ(t) will decay to zero arbitrarily fast. Notice,
however, that the coefficient of the (2,1) element of the exponential matrix is a
quadratic function of k. It can be shown that the absolute value of this element
reaches a maximum value k/e at t = 1/k. While this term can be made to decay to
zero arbitrarily fast by choosing k large, its transient behavior exhibits a peak of the
order of k. This phenomenon is known as the peaking phenomenon.7 The interaction
of peaking with nonlinear functions could destabilize the system. In particular, for
the initial states η(0) = η0 , ξ1 (0) = 1, and ξ2 (0) = 0, we have ξ2 (t) = −k 2 te−kt and
 
η̇ = − 12 1 − k 2 te−kt η 3
During the peaking period, the coefficient of η 3 is positive, causing |η(t)| to grow.
Eventually, the coefficient of η 3 will become negative, but that might not happen
soon enough to prevent a finite escape time. Indeed, the solution
η02
η 2 (t) =
1 + η02 [t + (1 + kt)e−kt − 1]
shows that if η02 > 1, the system will have a finite escape time if k is chosen large
enough. 
A sufficient condition for global asymptotic stability is given in the next lemma,
which follows from Lemma 4.6.
Lemma 9.3 The origin of the cascade connection (9.9) is globally asymptotically
stable if the system η̇ = f0 (η, ξ) is input-to-state stable 3
In the previous section we studied the effect of uncertainty on feedback lineariza-
tion. A similar study is now performed for partial feedback linearization. Assuming
that the implemented control is
u = γ̂ −1 (x)[−ψ̂(x) − K T̂2 (x)]

where γ̂, ψ̂, and T̂2 are approximations of γ, ψ, and T2 , respectively, the closed-loop
system is represented by
η̇ = f0 (η, ξ), ξ˙ = (A − BK)ξ + BΔ(z) (9.11)
where
! "
Δ(z) = ψ(x) + γ(x)γ̂ −1 (x)[−ψ̂(x) − K T̂2 (x)] + KT2 (x)
x=T −1 (z)

The system (9.11) is a cascade connection of η̇ = f0 (η, ξ), with ξ as input, and
ξ˙ = (A−BK)ξ+BΔ, with Δ as input. It follows from Lemma 4.6 that if η̇ = f0 (η, ξ)
is input-to-state stable, so is the system (9.11), with Δ as input. This observation
proves the following lemma.
7 Toread more about the peaking phenomenon, see [137]. For illustration of the peaking
phenomenon in high-gain observers, see Section 11.4.
9.5. BACKSTEPPING 211

Lemma 9.4 Consider the system (9.11) where η̇ = f0 (η, ξ) is input-to-state stable
and Δ(z) ≤ δ for all z, for some δ > 0. Then, z(t) is globally ultimately
bounded by a class K function of δ. 3
The bound Δ(z) ≤ δ could be conservative if Δ(z), or a component of it,
vanishes at z = 0. By viewing (9.11) as a perturbation of the nominal system (9.9)
and using the analysis of Sections 4.2 and 4.3, we arrive at the following lemma,
whose proof is given in Appendix D.
Lemma 9.5 Suppose the origin of η̇ = f0 (η, 0) is exponentially stable, the closed-
loop system (9.11) is well defined in a domain Dz ⊂ Rn containing the origin, and
Δ(z) ≤ kz + δ for all z ∈ Dz for some positive constants k and δ. Then, there
exist a neighborhood Nz of z = 0 and positive constants k ∗ , δ ∗ , and c such that for
k < k∗ , δ < δ ∗ , and z(0) ∈ Nz , z(t) will be ultimately bounded by cδ. If δ = 0,
the origin of (9.11) will be exponentially stable. 3

9.5 Backstepping
Consider the system

η̇ = fa (η) + ga (η)ξ (9.12)

ξ˙ = fb (η, ξ) + gb (η, ξ)u (9.13)

where η ∈ Rn and ξ ∈ R are the state variables and u ∈ R is the control input.
The functions fa , fb , ga , and gb are smooth8 in a domain that contains the origin
(η = 0, ξ = 0). Furthermore, fa (0) = 0. We want to design state feedback control
to stabilize the origin. This system can be viewed as a cascade connection of two
subsystems; the first is (9.12), with ξ as input, and the second (9.13), with u as
input. We start with the system (9.12) and view ξ as the control input. Suppose
the system (9.12) can be stabilized by a smooth state feedback control ξ = φ(η),
with φ(0) = 0; that is, the origin of

η̇ = fa (η) + ga (η)φ(η)

is asymptotically stable. Suppose further that we know a (smooth, positive definite)

Lyapunov function Va (η) that satisfies the inequality
∂Va
[fa (η) + ga (η)φ(η)] ≤ −W (η) (9.14)
∂η
where W (η) is positive definite. By adding and subtracting ga (η)φ(η) on the right-
hand side of (9.12), we obtain the equivalent representation

η̇ = fa (η) + ga (η)φ(η) + ga (η)[ξ − φ(η)]

8 We require smoothness of all functions for convenience. It will become clear, however, that in
a particular problem, we only need existence of derivatives up to a certain order.
212 CHAPTER 9. STATE FEEDBACK STABILIZATION

The change of variables

z = ξ − φ(η)
results in the system

η̇ = fa (η) + ga (η)φ(η) + ga (η)z

ż = F (η, ξ) + gb (η, ξ)u

where
∂φ
F (η, ξ) = fb (η, ξ) − [fa (η) + ga (η)φ(η) + ga (η)z]
∂η
which is similar to the system we started from, except that now the first subsystem
has an asymptotically stable origin when the input z is zero. This feature will be
exploited in the design of u to stabilize the overall system. Using

V (η, ξ) = Va (η) + 12 z 2 = Va (η) + 12 [ξ − φ(η)]2 (9.15)

as a Lyapunov function candidate, we obtain

∂Va ∂Va
V̇ = [fa (η) + ga (η)φ(η)] + ga (η)z + zF (η, ξ) + zgb (η, ξ)u
∂η ∂η

∂Va
≤ −W (η) + z ga (η) + F (η, ξ) + gb (η, ξ)u
∂η

The first term is negative definite in η. If we can choose u to make the second term
negative definite in z, V̇ will be negative definite in (η, z). This is possible because
gb = 0. In particular, choosing

1 ∂Va
u=− ga (η) + F (η, ξ) + kz (9.16)
gb (η, ξ) ∂η

for some k > 0, yields

V̇ ≤ −W (η) − kz 2
which shows that the origin (η = 0, z = 0) is asymptotically stable. Since φ(0) =
0, we conclude that the origin (η = 0, ξ = 0) is asymptotically stable. If all
the assumptions hold globally and Va (η) is radially unbounded, the origin will be
globally asymptotically stable.

Example 9.9 Consider the system

ẋ1 = x21 − x31 + x2 , ẋ2 = u

which takes the form (9.12)–(9.13) with η = x1 and ξ = x2 . We start with the
scalar system
ẋ1 = x21 − x31 + x2
9.5. BACKSTEPPING 213

with x2 viewed as the control input and proceed to design the feedback control
x2 = φ(x1 ) to stabilize the origin x1 = 0. With
def
x2 = −x21 − x1 = φ(x1 ) and Va (x1 ) = 12 x21

we obtain

ẋ1 = −x1 − x31 and V̇a = −x21 − x41 , ∀ x1 ∈ R

Hence, the origin of ẋ1 = −x1 − x31 is globally asymptotically stable. To backstep,
we use the change of variables

z2 = x2 − φ(x1 ) = x2 + x1 + x21

to transform the system into the form

ẋ1 = −x1 − x31 + z2 , ż2 = u + (1 + 2x1 )(−x1 − x31 + z2 )

Taking V (x) = 12 x21 + 12 z22 as a Lyapunov function candidate for the overall system,
we obtain

V̇ = x1 (−x1 − x31 + z2 ) + z2 [u + (1 + 2x1 )(−x1 − x31 + z2 )]

= −x21 − x41 + z2 [x1 + (1 + 2x1 )(−x1 − x31 + z2 ) + u]

The control
u = −x1 − (1 + 2x1 )(−x1 − x31 + z2 ) − z2
yields
V̇ = −x21 − x41 − z22
Hence, the origin is globally asymptotically stable. 

The application of backstepping in the preceding example is straightforward due

to the simplicity of scalar designs. For higher-dimensional systems, we may retain
this simplicity via recursive application of backstepping, as illustrated by the next
example.

Example 9.10 The three-dimensional system

ẋ1 = x21 − x31 + x2 , ẋ2 = x3 , ẋ3 = u

is composed of the two-dimensional system of the previous example with an ad-

ditional integrator at the input side. We proceed to apply backstepping as in the
previous example. After one step of backstepping, we know that the system

ẋ1 = x21 − x31 + x2 , ẋ2 = x3

214 CHAPTER 9. STATE FEEDBACK STABILIZATION

with x3 as input, can be globally stabilized by the control

def
x3 = −x1 − (1 + 2x1 )(x21 − x31 + x2 ) − (x2 + x1 + x21 ) = φ(x1 , x2 )
and
Va (x1 , x2 ) = 12 x21 + 12 (x2 + x1 + x21 )2
is the corresponding Lyapunov function. To backstep, we apply the change of
variables z3 = x3 − φ(x1 , x2 ) to obtain
ẋ1 = x21 − x31 + x2
ẋ2 = φ(x1 , x2 ) + z3
∂φ 2 ∂φ
ż3 = u− (x − x31 + x2 ) − (φ + z3 )
∂x1 1 ∂x2
Using V = Va + 12 z32 as a Lyapunov function candidate for the three-dimensional
system, we obtain
∂Va 2 ∂Va
V̇ = (x1 − x31 + x2 ) + (z3 + φ)
∂x1 ∂x2

∂φ 2 ∂φ
+ z3 u − (x − x1 + x2 ) −
3
(z3 + φ)
∂x1 1 ∂x2
= −x21 − x41 − (x2 + x1 + x21 )2

∂Va ∂φ 2 ∂φ
+ z3 − (x1 − x31 + x2 ) − (z3 + φ) + u
∂x2 ∂x1 ∂x2
Taking
∂Va ∂φ 2 ∂φ
u=− + (x − x31 + x2 ) + (z3 + φ) − z3
∂x2 ∂x1 1 ∂x2
yields
V̇ = −x21 − x41 − (x2 + x1 + x21 )2 − z32
Hence, the origin is globally asymptotically stable. 
9
By recursive application of backstepping, we can stabilize strict-feedback sys-
tems of the form10
ẋ = f0 (x) + g0 (x)z1
ż1 = f1 (x, z1 ) + g1 (x, z1 )z2
ż2 = f2 (x, z1 , z2 ) + g2 (x, z1 , z2 )z3
.. (9.17)
.
żk−1 = fk−1 (x, z1 , . . . , zk−1 ) + gk−1 (x, z1 , . . . , zk−1 )zk
żk = fk (x, z1 , . . . , zk ) + gk (x, z1 , . . . , zk )u
9 More results on backstepping and other recursive design techniques, including their application

to uncertain systems, are available in [43, 75, 81, 107, 121].

10 If you draw a schematic diagram of the system with the integrators stacked from right to left

as x, z1 , z2 , etc., you will see that the nonlinearities fi and gi in the żi -equation (i = 1, . . . , k)
depend only on x, z1 , . . . , zi ; that is, on the state variables that are “fed back.”
9.5. BACKSTEPPING 215

where x ∈ Rn , z1 to zk are scalars, f0 to fk vanish at the origin, and

gi (x, z1 , . . . , zi ) = 0 for 1 ≤ i ≤ k

over the domain of interest. The recursive procedure starts with the system

ẋ = f0 (x) + g0 (x)z1

where z1 is viewed as the control input. We assume that it is possible to find

stabilizing state feedback control z1 = φ0 (x), with φ0 (0) = 0, and a Lyapunov
function V0 (x) such that

∂V0
[f0 (x) + g0 (x)φ0 (x)] ≤ −W (x)
∂x
over the domain of interest for some positive definite function W (x). In many appli-
cations of backstepping, the variable x is scalar, which simplifies this stabilization
problem. With φ0 (x) and V0 (x) in hand, we proceed to apply backstepping in a
systematic way. First, we consider the system

ẋ = f0 (x) + g0 (x)z1
ż1 = f1 (x, z1 ) + g1 (x, z1 )z2

as a special case of (9.12)–(9.13) with

η = x, ξ = z1 , u = z2 , fa = f0 , ga = g0 , fb = f1 , gb = g1

We use (9.15) and (9.16) to obtain the stabilizing state feedback control and Lya-
punov function as

1 ∂φ0 ∂V0
φ1 (x, z1 ) = (f0 + g0 z1 ) − g0 − k1 (z1 − φ0 ) − f1
g1 ∂x ∂x

for some k1 > 0 and

V1 (x, z1 ) = V0 (x) + 12 [z1 − φ0 (x)]2

Next, we consider the system

ẋ = f0 (x) + g0 (x)z1
ż1 = f1 (x, z1 ) + g1 (x, z1 )z2
ż2 = f2 (x, z1 , z2 ) + g2 (x, z1 , z2 )z3

as a special case of (9.12)–(9.13) with

x f0 + g0 z1 0
η= , ξ = z2 , u = z3 , fa = , ga = , fb = f2 , gb = g2
z1 f1 g1
216 CHAPTER 9. STATE FEEDBACK STABILIZATION

Using (9.15) and (9.16), we obtain the stabilizing state feedback control and Lya-
punov function as

1 ∂φ1 ∂φ1 ∂V1

φ2 (x, z1 , z2 ) = (f0 + g0 z1 ) + (f1 + g1 z2 ) − g1 − k2 (z2 − φ1 ) − f2
g2 ∂x ∂z1 ∂z1
for some k2 > 0 and
V2 (x, z1 , z2 ) = V1 (x, z1 ) + 12 [z2 − φ2 (x, z1 )]2
This process is repeated k times to obtain the overall stabilizing state feedback
control u = φk (x, z1 , . . . , zk ) and Lyapunov function Vk (x, z1 , . . . , zk ).
Example 9.11 Consider a single-input–single output system in the normal form11
ẋ = f0 (x) + g0 (x)z1
żi = zi+1 , for 1 ≤ i ≤ ρ − 1
żρ = ψ(x, z) + γ(x, z)u
y = z1
where x ∈ Rn−ρ , z1 to zρ are scalars, and γ(x, z) = 0 for all (x, z). The system is in
the strict feedback form. The origin can be stabilized by recursive backstepping if
we can find a smooth function φ0 (x) and a smooth Lyapunov function V0 (x) such
that
∂V0
[f0 (x) + g0 (x)φ0 (x)] ≤ −W (x)
∂x
for some positive definite function W (x). If the system is minimum phase, then the
origin of ẋ = f0 (x) is asymptotically stable. In this case, if we know a Lyapunov
function V0 (x) that satisfies
∂V0
f0 (x) ≤ −W (x)
∂x
for some positive definite function W (x), we can simply take φ0 (x) = 0. 
Example 9.12 The system
ẋ = −x + x2 z, ż = u
was considered in Example 9.7, where it was shown that u = −kz, with sufficiently
large k > 0, can achieve semiglobal stabilization. In this example, we achieve global
stabilization via backstepping.12 Starting with the system ẋ = −x + x2 z, it is clear
that z = 0 and V0 (x) = 12 x2 result in
∂V0
(−x) = −x2 , ∀x∈R
∂x
11 This system is a special case of the normal form of Section 8.1 where the ẋ-equation takes the

form f0 (x) + g0 (x)z1 , instead of the more general form f0 (x, z).
12 The reader can verify that the system is not feedback linearizable near the origin.
9.6. PASSIVITY-BASED CONTROL 217

Using V = V0 + 12 z 2 = 12 (x2 + z 2 ), we obtain

V̇ = x(−x + x2 z) + zu = −x2 + z(x3 + u)

Thus, u = −x3 − kz, with k > 0, globally stabilizes the origin. 

Example 9.13 As a variation from the previous example, consider the system13

ẋ = x2 − xz, ż = u

This time, ẋ = x2 − xz cannot be stabilized by z = 0. It is easy, however, to see

that z = x + x2 and V0 (x) = 12 x2 result in

∂V0 2
[x − x(x + x2 )] = −x4 , ∀x∈R
∂x
Using V = V0 + 12 (z − x − x2 )2 , we obtain

V̇ = x(x2 − xz) + (z − x − x2 )[u − (1 + 2x)(x2 − xz)]

= −x4 + (z − x − x2 )[−x2 + u − (1 + 2x)(x2 − xz)]

The control

u = (1 + 2x)(x2 − xz) + x2 − k(z − x − x2 ), k>0

yields
V̇ = −x4 − k(z − x − x2 )2
Hence, the origin is globally asymptotically stable. 

9.6 Passivity-Based Control

In Chapters 5 and 7, we introduced passivity and studied its role in analyzing the
stability of feedback connections. The passivity-based control that we are going
to introduce here is a straightforward application of these chapters, but we do not
need their elaborate details.14 It is enough to recall the definitions of passivity and
zero-state observability.
We consider the m-input–m-output system

ẋ = f (x, u), y = h(x) (9.18)

where f is locally Lipschitz in (x, u) and h is continuous in x, for all x ∈ Rn and

u ∈ Rm . We assume that f (0, 0) = 0, so that the origin x = 0 is an open-loop
13 With the output y = z, the system is nonminimum phase, because the origin of the zero-

dynamics ẋ = x2 is unstable.
14 More results on passivity-based control are available in [121, 143] and extensive applications

to physical systems are given in [87, 101].

218 CHAPTER 9. STATE FEEDBACK STABILIZATION

equilibrium point, and h(0) = 0. We recall that the system (9.18) is passive if there
exists a continuously differentiable positive semidefinite storage function V (x) such
that
∂V
uT y ≥ V̇ = f (x, u), ∀ (x, u) ∈ Rn × Rm (9.19)
∂x
The system is zero-state observable if no solution of ẋ = f (x, 0) can stay identically
in the set {h(x) = 0} other than the zero solution x(t) ≡ 0. Throughout this
section we will require the storage function to be positive definite. The basic idea
of passivity-based control is illustrated in the next theorem.
Theorem 9.1 If the system (9.18) is
1. passive with a radially unbounded positive definite storage function, and
2. zero-state observable,
then the origin x = 0 can be globally stabilized by u = −φ(y), where φ is any locally
Lipschitz function such that φ(0) = 0 and y T φ(y) > 0 for all y = 0.
Proof: Use the storage function V (x) as a Lyapunov function candidate for the
closed-loop system
ẋ = f (x, −φ(y))
The derivative of V is given by
∂V
V̇ = f (x, −φ(y)) ≤ −y T φ(y) ≤ 0
∂x
Hence, V̇ is negative semidefinite and V̇ = 0 if and only if y = 0. By zero-state
observability,
y(t) ≡ 0 ⇒ u(t) ≡ 0 ⇒ x(t) ≡ 0
Therefore, by the invariance principle, the origin is globally asymptotically stable.
2
The intuition behind the theorem becomes clear when we think of the storage
function as the energy of the system. A passive system has a stable origin. All
that is needed to stabilize the origin is the injection of damping so that energy will
dissipate whenever x(t) is not identically zero. The required damping is injected by
the function φ. There is great freedom in the choice of φ. We can choose it to meet
any constraint on the magnitude of u. For example, if u is constrained to |ui | ≤ ki
for 1 ≤ i ≤ m, we can choose φi (y) = ki sat(yi ) or φi (y) = (2ki /π) tan−1 (yi ).
The utility of Theorem 9.1 can be increased by transforming nonpassive systems
into passive ones. Consider, for example, a special case of (9.18), where
ẋ = f (x) + G(x)u (9.20)
Suppose a radially unbounded, positive definite, continuously differentiable function
V (x) exists such that
∂V
f (x) ≤ 0, ∀ x
∂x
9.6. PASSIVITY-BASED CONTROL 219

Take
T
def ∂V
y = h(x) = G(x)
∂x
Then the system with input u and output y is passive. If it is also zero-state
observable, we can apply Theorem 9.1.

Example 9.14 Consider the system

ẋ1 = x2 , ẋ2 = −x31 + u

With V (x) = 14 x41 + 12 x22 ,

V̇ = x31 x2 − x2 x31 + x2 u = x2 u

Set y = x2 and note that, with u = 0, y(t) ≡ 0 implies that x(t) ≡ 0. Thus, all
the conditions of Theorem 9.1 are satisfied and a globally stabilizing state feedback
control can be taken as u = −kx2 or u = −(2k/π) tan−1 (x2 ) with any k > 0. 

Allowing ourselves the freedom to choose the output function is useful, but we
are still limited to state equations for which the origin is open-loop stable. We can
cover a wider class of systems if we use feedback to achieve passivity. Consider
again the system (9.20). If a feedback control

u = α(x) + β(x)v (9.21)

and an output function h(x) exist such that the system

ẋ = f (x) + G(x)α(x) + G(x)β(x)v, y = h(x) (9.22)

with input v and output y, satisfies the conditions of Theorem 9.1, we can globally
stabilize the origin by using v = −φ(y). The use of feedback to convert a nonpassive
system into a passive one is known as feedback passivation.15
Example 9.15 Note from Appendix A that the nonlinear dynamic equations for
an m-link robot take the form

M (q)q̈ + C(q, q̇)q̇ + Dq̇ + g(q) = u

where the m-dimensional vectors q and u represent generalized coordinates and con-
trol, respectively. The inertia matrix M (q) is symmetric and positive definite for all
q, Ṁ −2C is skew-symmetric for all q and q̇, where Ṁ is the total derivative of M (q)
with respect to t, D is symmetric and positive semidefinite, and the gravity term
15 For the system (9.20) with output y = h(x), it is shown in [20] that the system is

locally feedback equivalent to a passive system with a positive definite storage function if
rank{[∂h/∂x](0)G(0)} = m and the zero dynamics have a stable equilibrium point at the ori-
gin with a positive definite Lyapunov function.
220 CHAPTER 9. STATE FEEDBACK STABILIZATION

g(q) = [∂P (q)/∂q]T , where P (q) is the total potential energy. Consider the prob-
lem of designing state feedback control to asymptotically regulate q to a constant
reference qr . The regulation error e = q − qr satisfies the differential equation

M (q)ë + C(q, q̇)ė + Dė + g(q) = u

The regulation task is achieved by stabilizing the system at (e = 0, ė = 0), but this
point is not an open-loop equilibrium point. Taking

u = g(q) − Kp e + v

where Kp is a positive definite symmetric matrix and v an additional control com-

ponent to be chosen, results in the system

M (q)ë + C(q, q̇)ė + Dė + Kp e = v

which has an equilibrium point at (e = 0, ė = 0) when v = 0, and its potential

energy 12 eT Kp e has a unique minimum at e = 0. The sum of the kinetic energy and
the reshaped potential energy is the positive definite function

V = 12 ėT M (q)ė + 12 eT Kp e

Taking V as a storage function candidate, we have

V̇ = ėT M ë + 12 ėT Ṁ ė + eT Kp ė
= 1 T
2 ė (Ṁ − 2C)ė − ėT Dė − ėT Kp e + ėT v + eT Kp ė ≤ ėT v

Defining the output as y = ė, we see that the system with input v and output y is
passive with V as the storage function. With v = 0,

y(t) ≡ 0 ⇔ ė(t) ≡ 0 ⇒ ë(t) ≡ 0 ⇒ Kp e(t) ≡ 0 ⇒ e(t) ≡ 0

which shows that the system is zero-state observable. Hence, it can be globally
stabilized by the control v = −φ(ė) with any function φ such that φ(0) = 0 and
y T φ(y) > 0 for all y = 0. The choice v = −Kd ė with a positive definite symmetric
matrix Kd results in
u = g(q) − Kp (q − qr ) − Kd q̇
which is a PD (Proportional-Derivative) controller with gravity compensation. 
A class of systems that is amenable to feedback passivation is the cascade con-
nection of a passive system with a system whose unforced dynamics have a stable
equilibrium point at the origin. Consider the system

ẋ = fa (x) + F (x, y)y (9.23)

ż = f (z) + G(z)u (9.24)
y = h(z) (9.25)
9.6. PASSIVITY-BASED CONTROL 221

where fa (0) = 0, f (0) = 0, and h(0) = 0. The functions fa , F , f , and G are locally
Lipschitz and h is continuous. We view the system as a cascade connection of the
driving system (9.24)–(9.25) and the driven system (9.23).16 We assume that the
representation (9.23)–(9.25) is valid globally, the driving system is passive with a
radially unbounded positive definite storage function V (z), the origin of ẋ = fa (x) is
stable, and we know a radially unbounded Lyapunov function W (x) for ẋ = fa (x),
which satisfies
∂W
fa (x) ≤ 0, ∀ x
∂x
Using U (x, z) = W (x) + V (z) as a storage function candidate for the full system
(9.23)–(9.25), we obtain
∂W ∂W ∂V
U̇ = fa (x) + F (x, y)y + [f (z) + G(z)u]
∂x ∂x ' ∂z  T (
∂W ∂W
≤ F (x, y)y + y T u = y T u + F (x, y)
∂x ∂x

The feedback control  T

∂W
u=− F (x, y) +v
∂x
results in
U̇ ≤ y T v
Hence, the system
ẋ = fa (x) + F (x, y)y (9.26)
 T
∂W
ż = f (z) − G(z) F (x, y) + G(z)v (9.27)
∂x
y = h(z) (9.28)
with input v and output y is passive with U as the storage function. If the system
(9.26)–(9.28) is zero-state observable, we can apply Theorem 9.1 to globally stabilize
the origin. Checking zero-state observability of the full system can be reduced to
checking zero-state observability of the driving system (9.24)–(9.25) if we strengthen
the assumption on W (x) to
∂W
fa (x) < 0, ∀ x = 0 (9.29)
∂x
which implies that the origin of ẋ = fa (x) is globally asymptotically stable. Taking
 T
∂W
u=− F (x, y) − φ(y) (9.30)
∂x
16 A driven system of the form ẋ = f0 (x, y) with sufficiently smooth f0 can be represented in

the form (9.23) by taking fa (x) = f0 (x, 0) and F (x, y) = 01 [∂f0 /∂y](x, sy) ds.
222 CHAPTER 9. STATE FEEDBACK STABILIZATION

where φ is any locally Lipschitz function such that φ(0) = 0 and y T φ(y) > 0 for all
y = 0, and using U as a Lyapunov function candidate for the closed-loop system,
we obtain
∂W
U̇ ≤ fa (x) − y T φ(y) ≤ 0
∂x
U̇ = 0 ⇒ x = 0 and y = 0 ⇒ u=0
If the driving system (9.24)–(9.25) is zero-state observable, the conditions u(t) ≡ 0
and y(t) ≡ 0 imply that z(t) ≡ 0. Hence, by the invariance principle, the origin
(x = 0, z = 0) is globally asymptotically stable. We summarize this conclusion in
the next theorem.

Theorem 9.2 Suppose the system (9.24)–(9.25) is zero-state observable and passive
with a radially unbounded, positive definite storage function. Suppose the origin of
ẋ = fa (x) is globally asymptotically stable and let W (x) be a radially unbounded,
positive definite Lyapunov function that satisfies (9.29). Then, the control (9.30)
globally stabilizes the origin (x = 0, z = 0). 3

Example 9.16 Consider the system

ẋ = −x + x2 z, ż = u

which was studied in Examples 9.7 and 9.12. With y = z as the output, the system
takes the form (9.23)–(9.25). The system ż = u, y = z is passive with the storage
function V (z) = 12 z 2 . It is zero-state observable because y = z. The origin of
ẋ = −x is globally exponentially stable with the Lyapunov function W (x) = 12 x2 .
Thus, all the conditions of Theorem 9.2 are satisfied and a globally stabilizing state
feedback control is given by

u = −x3 − kz, k>0

which is the same control we derived by using backstepping 

9.7 Control Lyapunov Functions

Consider the stabilization problem for the system

ẋ = f (x) + g(x)u (9.31)

where x ∈ Rn , u ∈ R, f (x) and g(x) are locally Lipschitz, and f (0) = 0. Suppose
there is a locally Lipschitz stabilizing state feedback control u = χ(x) such that the
origin of
ẋ = f (x) + g(x)χ(x) (9.32)
9.7. CONTROL LYAPUNOV FUNCTIONS 223

is asymptotically stable. Then, by (the converse Lyapunov) Theorem 3.9, there is

a smooth Lyapunov function V (x) such that
∂V
[f (x) + g(x)χ(x)] < 0, ∀ x = 0 (9.33)
∂x
in some neighborhood of the origin. If the origin of (9.32) is globally asymptotically
stable, then V (x) is radially unbounded and inequality (9.33) holds globally. A
function V satisfying (9.33) must have the property:
∂V ∂V
g(x) = 0 and x = 0 ⇒ f (x) < 0 (9.34)
∂x ∂x
Thus, the existence of a function V satisfying (9.34) is a necessary condition for the
existence of a stabilizing state feedback control. A positive definite function V (x)
of class C 2 that satisfies (9.34) in some neighborhood of the origin is called control
Lyapunov function. If it is radially unbounded and (9.34) is satisfied globally, it is
called global control Lyapunov function.17 It turns out that the existence of a control
Lyapunov function is also a sufficient condition for the existence of a stabilizing state
feedback control. In particular, given a control Lyapunov function V , a stabilizing
state feedback control is given by u = φ(x), where18
⎧

⎨  ∂V   ∂V 2  ∂V 4 * ∂V   
− ∂x f + ∂x f + ∂x g ∂x g if ∂V
∂x g = 0
φ(x) =   (9.35)
⎩
0 if ∂V
∂x g = 0
This can be seen by using V as a Lyapunov function candidate for the closed-loop
system
ẋ = f (x) + g(x)φ(x)
For x = 0, if [∂V /∂x]g = 0, then
∂V
V̇ = f <0
∂x
and if [∂V /∂x]g = 0, then
⎡ ⎤
 2  4
∂V ∂V ∂V ∂V
V̇ = f −⎣ f+ f + g ⎦
∂x ∂x ∂x ∂x
 2  4
∂V ∂V
= − f + g <0
∂x ∂x

Thus, for all x = 0 in a neighborhood of the origin, V̇ (x) < 0, which shows that
the origin is asymptotically stable. If V is a global control Lyapunov function, the
foregoing calculations show that the origin is globally asymptotically stable.
17 See [43] for the definition of robust control Lyapunov functions.
18 Equation (9.35) is known as Sontag’s formula.
224 CHAPTER 9. STATE FEEDBACK STABILIZATION

The function φ(x) of (9.35) must be sufficiently smooth to be a valid state

feedback control law. This is established in the following lemma.19

Lemma 9.6 If f (x), g(x) and V (x) are smooth then φ(x), defined by (9.35), will
be smooth for x = 0. If they are of class C +1 for  ≥ 1, then φ(x) will be of class
C  . Continuity at x = 0 follows from one of the following two cases.

• φ(x) is continuous at x = 0 if V (x) has the small control property; namely,

given any ε > 0 there δ > 0 such that if x = 0 and x < δ, then there is u
with u < ε such that
∂V
[f (x) + g(x)u] < 0
∂x
• φ(x) is locally Lipschitz at x = 0 if there is a locally Lipschitz function χ(x),
with χ(0) = 0, such that

∂V
[f (x) + g(x)χ(x)] < 0, for x = 0
∂x

To use the formula (9.35), the control Lyapunov function should satisfy one of the
two bullets of Lemma 9.6, preferably the second one since we usually require the
feedback control to be locally Lipschitz.
The concept of control Lyapunov functions gives a necessary and sufficient con-
dition for the stabilizability of a nonlinear system. However, it may not be a useful
tool for stabilization because to use (9.35) we need to find a control Lyapunov func-
tion, which may not be easy. One case where a control Lyapunov function would
be readily available for the system (9.31) is when a stabilizing control u = χ(x) is
available together with a Lyapunov function V (x) that satisfies (9.33). The use of
V in (9.35) results in another stabilizing state feedback control u = φ(x), which
may have different transient response or robustness properties compared with the
original control u = χ(x). We note that in this case the function V satisfies the
second bullet of Lemma 9.6; hence φ(x) will be locally Lipschitz. Examples of stabi-
lization techniques where a Lyapunov function satisfying (9.33) is always available
are the feedback linearization method of Section 9.3 and the backstepping method
of Section 9.5. In backstepping, the recursive procedure systematically produces
a Lyapunov function that satisfies (9.33). In feedback linearization, a quadratic
Lyapunov function can be obtained by solving a Lyapunov equation for the closed-
loop linear system. In particular, if there is a change of variables z = T (x) that
transforms the system (9.31) into the controller form

ż = Az + B[ψ(x) + γ(x)u]
19 The proof of the lemma, except for the second bullet, is given in [132, Section 5.9]. The proof
of the second bullet is in [121, Proposition 3.43].
9.7. CONTROL LYAPUNOV FUNCTIONS 225

then the system can be stabilized by the control

u = γ −1 (x)[−ψ(x) − Kz]

where (A − BK) is Hurwitz. A Lyapunov function in the x-coordinates is given

by V (x) = T T (x)P T (x), where P is the positive definite solution of the Lyapunov
equation P (A − BK) + (A − BK)T P = −I.
Example 9.17 The scalar system

ẋ = x − x3 + u

can be globally stabilized by the feedback control u = χ(x) = −x + x3 − αx, for

α > 0, which is designed using feedback linearization. The closed-loop system
ẋ = −αx has the Lyapunov function V (x) = 12 x2 . Using V as a global control
Lyapunov function, the formula (9.35) results in the control

u = φ(x) = −x + x3 − x (1 − x2 )2 + 1

and the closed-loop system is


ẋ = −x (1 − x2 )2 + 1

To compare the two feedback control laws, Figure √ 9.1 plots u versus x and the
closed-loop ẋ versus x for each control, with α = 2. The plots reveal that φ has
an advantage over χ when operating far from the origin. The magnitude of φ is
much smaller for large |x|, while the closed-loop system under under φ has a much
faster decay towards zero. In a way, the control φ recognizes the beneficial effect of
the nonlinearity −x3 , which is ignored by feedback linearization. 
The advantage of control-Lyapunov-function stabilization over feedback lin-
earization that we saw in the previous example cannot be proved, and indeed may
not be true in general. What is true, however, is that stabilization using (9.35) has
a robustness property that is not enjoyed by most other stabilization techniques.
Suppose the system (9.31) is perturbed in such a way that the control u is multiplied
by a positive gain k before it is applied to the system; that is, the perturbed system
takes the form ẋ = f (x) + g(x)ku. If the control u stabilizes the origin x = 0 for
all k ∈ [α, β], we say that the closed-loop system has a gain margin [α, β]. When
k ∈ [1, gm ) the system has the classical gain margin of gm . The following lemma
shows that control-Lyapunov-function stabilization using (9.35) has a gain margin
[ 12 , ∞).20
20 The gain margin of Lemma 9.7 is a characteristic feature of optimal stabilizing controls that

minimize a cost functional of the form 0∞ [q(x) + r(x)u2 ] dt, where r(x) > 0 and q(x) is positive
semidefinite. The gain margin extends to the control (9.35) because it has an inverse optimality
property; namely, for any control given by (9.35) there are q(x) and r(x) such that the control
(9.35) is the optimal control for 0∞ [q(x) + r(x)u2 ] dt; see [121] for further discussion of optimal
and inverse optimal control.
226 CHAPTER 9. STATE FEEDBACK STABILIZATION

20
20

10
10

0 0
u

f
−10
−10 FL
CLF
−20
−20
−3 −2 −1 0 1 2 3 −3 −2 −1 0 1 2 3
x x

Figure 9.1: Comparison of the control u and the closed-loop ẋ = f between feedback
linearization (FL) and control Lyapunov function (CLF) in Example 9.17.

Lemma 9.7 Suppose f , g, and V satisfy the conditions of Lemma 9.6, V is a

control Lyapunov function, and φ is given by (9.35). Then, the origin of ẋ =
f (x) + g(x)kφ(x) is asymptotically stable for all k ≥ 12 . If V is a global control
Lyapunov function, then the origin is globally asymptotically stable. 3

Proof: Let ⎡ ⎤
 2  4
∂V ∂V ∂V
q(x) = 12 ⎣− f+ f + g ⎦
∂x ∂x ∂x

Because V (x) is positive definite, it has a minimum at x = 0, which implies that

[∂V /∂x](0) = 0. Hence, q(0) = 0. For x = 0, if [∂V /∂x]g = 0, q(x) > 0 and if
[∂V /∂x]g = 0, q = −[∂V /∂x]f > 0. Therefore, q(x) is positive definite. Using V as
a Lyapunov function candidate for the system ẋ = f (x) + g(x)kφ(x), we have

∂V ∂V
V̇ = f+ gkφ
∂x ∂x

For x = 0, if [∂V /∂x]g = 0, V̇ = [∂V /∂x]f < 0. If [∂V /∂x]g = 0, we have

∂V ∂V
V̇ = −q + q + f+ gkφ
∂x ∂x
Noting that
⎡ ⎤
 2  4
∂V ∂V   ∂V ∂V ∂V
q+ f+ gkφ = − k − 12 ⎣ f+ f + g ⎦≤0
∂x ∂x ∂x ∂x ∂x

we see that V̇ (x) < 0 for x = 0, which completes the proof. 2

9.8. EXERCISES 227

Example 9.18 In Example 9.17 we compared the feedback-linearization control

u = χ(x) =−x + x3 − αx with the control-Lyapunov-function formula u = φ(x) =
−x+x3 −x (1 − x2 )2 + 1 for the system ẋ = x−x3 +u. According to Lemma 9.7 the
origin of the closed-loop system ẋ = x − x3 + kφ(x) is globally asymptotically stable
for all k ≥ 12 . Let us study the gain margin of the system ẋ = x − x3 + kχ(x). As
far as local stabilization is concerned, the system has a gain margin (1/(1 + α), ∞)
because linearization as the origin yields ẋ = −[k(1 + α) − 1]x. But, the origin
will not be globally asymptotically stable  for any k > 1 because for k > 1 the
system
 has equilibrium pointsat 0 and ± 1 + kα/(k − 1). The region of attraction
√
{−
√ 1 + kα/(k − 1) < x < 1 + kα/(k − 1))} shrinks toward {− 1 + α < x <
1 + α} as k → ∞. 

9.8 Exercises
9.1 For each of the following systems, design a globally stabilizing state feedback
controller.
(1) ẋ1 = x21 + x2 , ẋ2 = u
(2) ẋ1 = x2 , ẋ2 = −x31 + sat(u)
(3) ẋ1 = x1 x2 , ẋ2 = u
(4) ẋ1 = −x1 + tanh(x2 ), ẋ2 = x2 + x3 , ẋ3 = u
(5) ẋ1 = −x1 + x1 x3 , ẋ2 = x3 , ẋ3 = x21 + u
(6) ẋ1 = x31 + x2 , ẋ2 = x1 x3 + u, ẋ3 = x1 x2 − x3

9.2 For each of the following systems, design a linear globally stabilizing state
feedback controller.
(1) ẋ1 = x2 , ẋ2 = x1 + x2 − x31 + u
(2) ẋ1 = sin x1 + x2 , ẋ2 = u
(3) ẋ1 = −x31 + x22 , ẋ2 = u
(4) ẋ1 = x2 , ẋ2 = −x2 |x2 | + u
(5) ẋ1 = x2 , ẋ2 = sin x1 + u

9.3 For each of the following systems, it is required to design a state feedback
controller to globally stabilize the origin.
(a) Design the controller using feedback linearization.
(b) Design the controller using backstepping.
228 CHAPTER 9. STATE FEEDBACK STABILIZATION

(c) Compare the performance of the two controllers using simulation.

(1) ẋ1 = x1 + x2 /(1 + x21 ), ẋ2 = −x2 + u

(2) ẋ1 = sin x1 + x2 , ẋ2 = u
(3) ẋ1 = x2 + 1 + (1 − x1 )3 , ẋ2 = −x1 + u

9.4 Consider the inverted pendulum on a cart (A.41)–(A.44), where it is required

to stabilize the pendulum at θ = 0 and the cart at y = 0.
(a) Show that the task cannot be achieved by open-loop control.
(b) Using linearization, design a stabilizing state feedback controller.
(c) Using simulation with the (A.45) data, find
(i) the range of x1 (0), when x(0) = col(x1 (0), 0, 0, 0), for which the pendulum
is stabilized;
(ii) the effect of ±30% perturbation in m and J.

9.5 Consider the inverted pendulum (A.47), where it is required to stabilize the
pendulum at θ = 0 using state feedback.
(a) Design a stabilizing controller using linearization.
(b) Design a stabilizing controller using feedback linearization and determine its
domain of validity.
(c) For each of the two controllers, use simulation with a = 1 to find the range of
x1 (0), when x2 (0) = 0, for which the pendulum is stabilized and |x1 (t)| ≤ π/2
for all t.
(d) Compare the performance of the two controllers.

9.6 Consider the controlled van der Pol equation (A.13). Design a globally stabi-
lizing state feedback controller using
(a) feedback linearization
(b) backstepping
(c) feedback passivation and passivity based control
(d) control Lyapunov function, starting from the Lyapunov function of part (b).

9.7 Consider the biochemical reactor (A.19) with known ν(x2 ) given by (A.20).
(a) Using partial feedback linearization, design a state feedback controller such
that ẋ2 = −λ(x2 − 1) with λ > 0. Determine the region of validity of the
controller.
9.8. EXERCISES 229

(b) Show that the controller of part (a) stabilizes the equilibrium point (x1 =
1, x2 = 1) and estimate the region of attraction.

9.8 Consider the magnetic levitation system (A.30)–(A.32) with b = 0, c = 1,

α = 1.5, and β = 2. It is required to stabilize the ball at x1 = 1.

(a) Find the steady-state input uss that maintains equilibrium at x1 = 1, and show
that the equilibrium point is unstable under the open-loop control u = uss .

In the rest of this exercise we design three different state feedback controllers. Using
simulation, choose the controller parameters in each case to have a settling time less
than 20 time units, following a small perturbation from the equilibrium point, and
find the range of x1 (0), when the other states start from their equilibrium values,
for which the ball will be stabilized at x1 = 1.

(b) Design a controller using linearization.

(c) Transform the state equation into the controller form and determine the domain
of validity of the transformation.

(d) Design a controller using feedback linearization.

(e) Starting from the controller form, design a controller using backstepping.

(f ) Compare the performance of the three controllers when x(0) = col(2, 0, 1.5).

9.9 Consider the electrostatic microactuator (A.33). It is required to stabilize the

plate at x1 = r < 1.

(a) By studying the equilibrium points for a constant input u = ū < 1, determine
the restrictions on r and the region of attraction if the system is to be operated
without feedback control.

In the rest of this exercise let r = 0.5, ζ = 0.1, and T = 0.2. We design four differ-
ent state feedback controllers. Using simulation, choose √ the controller parameters
in each case so that |u(t)| ≤ 2√when x(0) = col(0.1, 0. 0.3). For this choice of
parameters, let x(0) = col(α, 0, 3α) and find the range of α for which x(0) belongs
to the region of attraction if the control was not constrained.

(b) Design a controller using linearization.

(c) Transform the state equation into the controller form and determine the domain
of validity of the transformation.

(d) Design a controller using feedback linearization.

(e) Starting from the controller form, design a controller using backstepping.
230 CHAPTER 9. STATE FEEDBACK STABILIZATION

(f ) Using partial feedback linearization to convert the ẋ3 -equation into ẋ3 = v,
and then shifting the desired equilibrium point to the origin, show that the
system can be represented in the form (9.23)–(9.25) and the conditions of
Theorem 9.2 are satisfied.
(g) Design a controller using passivity-based control.
(h) Compare the performance of the four controllers.
9.10 Consider the boost converter (A.16).
(a) Show that, with u = 0, the open-loop system is linear and has an asymptotically
stable origin.
(b) Using passivity-based control, design a state feedback controller such that the
origin is globally asymptotically stable, 0 < μ = u + 1 − 1/k < 1, and the
closed-loop system has more damping than the open-loop one.
(c) Simulate the open-loop and closed-loop systems with α = 0.1, k = 2, x1 (0) =
−0.7/4, and x2 (0) = − 12 .
9.11 Consider the DC motor (A.25) with δ = 0 and f (x3 ) = bx3 where b ≥ 0. It
is desired to design a state feedback controller to stabilize the speed at ω ∗ .
(a) Transform equation (A.25) into the normal form, as described in Example 8.6.
(b) Using partial feedback linearization bring the external part of the normal form
into ξ˙1 = ξ2 , ξ˙2 = v and determine its domain of validity.
(c) Find the equilibrium point at which ω = ω ∗ .
(d) Design a state feedback controller for v to stabilize the equilibrium point and
find under what conditions is the controller valid.
9.12 Consider the two-link manipulator defined by (A.36) and (A.37) with the
(A.38) data. It is desired to regulate the manipulator to qr = (π/2, π/2) under the
control constraints |u1 | ≤ 6000 and |u2 | ≤ 5000.
(a) Apply the passivity-based controller of Example 9.15. Use simulation to choose
the controller parameters to meet the control constraints.
(b) Design a feedback-linearization controller based on Example 8.7. Use simula-
tion to choose the controller parameters to meet the control constraints.
(c) Compare the performance of the two controllers under the nominal parameters
of (A.38) and the perturbed parameters of (A.39).
9.13 Consider the TORA system of (A.49)–(A.52). Show that the origin can be
globally stabilized by u = −φ1 (x1 ) − φ2 (x2 ) where φ1 and φ2 are locally Lips-
chitz functions that satisfy φi (0) = 0 and yφi (y) > 0 for all y = 0. Hint: See
Exercise 5.10.
Chapter 10

Robust State Feedback Stabilization

The stabilization techniques of Chapter 9, except possibly for passivity-based con-

trol, require a good model of the system. Perturbation analysis of the closed-loop
system determines the level of model uncertainty that can be tolerated while main-
taining asymptotic stability. However, the level of model uncertainty cannot be
prescribed. The robust stabilization techniques of this chapter can ensure stabiliza-
tion under “large” matched uncertainties provided upper bounds on the uncertain
terms are known. The uncertainty is matched when the uncertain terms enter the
state equation at the same point as the control input, which allows the control to
dominate their effect.
Three robust stabilization techniques are presented in this chapter. In the sliding
mode control of Section 10.1, a sliding manifold, independent of model uncertainty,
is designed such that trajectories on the manifold converge to the equilibrium point.
Then, the control is designed to force all trajectories to reach the manifold in finite
time and stay on it for all future time. The control uses only upper bounds on
the uncertain terms. In the Lyapunov redesign of Section 10.2, a stabilizing con-
trol is designed first for a nominal model with no uncertainty. Then, a Lyapunov
function of the nominal closed-loop system is used to design an additional control
component that makes the design robust to matched uncertainties. Both sliding
mode control and Lyapunov redesign produce discontinuous feedback controllers,
which poses theoretical as well as practical challenges. Theoretically, the right-hand
side function of the closed-loop system will not be locally Lipschitz, which is the
underlying assumption for the theory presented in this book. Practically, the im-
plementation of such discontinuous controllers is characterized by the phenomenon
of chattering. Therefore, in Section 10.1 and 10.2 we develop continuous approxi-
mations of the discontinuous controllers. In Section 10.3, we show how to achieve
robust stabilization by high-gain feedback.
When the uncertain terms do not vanish at the equilibrium point, locally Lips-
chitz feedback controllers cannot stabilize the equilibrium point but can bring the
trajectories arbitrarily close to it. This is known as practical stabilization, which is

231
232 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

defined next for the system

ẋ = f (x, u) + δ(t, x, u)
where f (0, 0) = 0 and δ(t, 0, 0) = 0. For a locally Lipschitz feedback control u =
φ(x), with φ(0) = 0, the origin x = 0 is not an equilibrium point of the closed-loop
system ẋ = f (x, φ(x)) + δ(t, x, φ(x)).
Definition 10.1 The system ẋ = f (x, u) + δ(t, x, u) is said to be practically stabi-
lizable by the feedback control u = φ(x) if for every b > 0, the control can be designed
such that the solutions are ultimately bounded by b; that is
x(t) ≤ b, ∀t≥T
for some T > 0.
Typically, the control u = φ(x) depends on a design parameter that can be chosen to
achieve the ultimate bound b. In Section 9.1 we defined four notions of stabilization
depending on the region of initial states. With practical stabilization, we can use the
same notions and talk of local, regional, global or semiglobal practical stabilization.
For example, the system is semiglobally practically stabilizable by the feedback
control u = φ(x) if given any compact set G and any b > 0, φ(x) can be designed
to ensure that for all initial states in G, the solutions are ultimately bounded by b.

10.1 Sliding Mode Control

We start with a motivating example, then present the theory for n-dimensional
multi-input systems.
Example 10.1 Consider the two-dimensional, single-input system
ẋ1 = x2 , ẋ2 = h(x) + g(x)u
where h and g are unknown locally Lipschitz functions and g(x) ≥ g0 > 0 for all x.
We want to design a state feedback controller to stabilize the origin. Suppose we
can design a controller that constrains the trajectory to the manifold (or surface)
s = ax1 + x2 = 0, where the motion is governed by ẋ1 = −ax1 . Choosing a > 0
guarantees that x(t) tends to zero as t tends to infinity and the rate of convergence
can be controlled by choice of a. The motion on the manifold s = 0 is independent
of h and g. How can we bring the trajectory to this manifold and maintain it there?
The variable s satisfies the equation
ṡ = aẋ1 + ẋ2 = ax2 + h(x) + g(x)u
Suppose h and g satisfy the inequality
 
 ax2 + h(x) 
  ≤ (x), ∀ x ∈ R2
 g(x) 
10.1. SLIDING MODE CONTROL 233

1 2
for some known function (x). With V = 2s as a Lyapunov function candidate
for ṡ = ax2 + h(x) + g(x)u, we have

V̇ = sṡ = s[ax2 + h(x)] + g(x)su ≤ g(x)|s|(x) + g(x)su

We want to choose u such that when |s| = 0 the term g(x)su is negative and
dominates the positive term g(x)|s|(x). Moreover, we want the net negative term
to force |s| to reach zero in finite time. This can be achieved by taking

u = −β(x) sgn(s)

where the signum function sgn(·) is defined by


1, s>0
sgn(s) =
−1, s<0

and β(x) ≥ (x) + β0 , β0 > 0. Whenever |s| = 0, we have

V̇ ≤ g(x)|s|(x) − g(x)[(x) + β0 ]s sgn(s) = −g(x)β0 |s| ≤ −g0 β0 |s|

√
Thus, W = 2V = |s| satisfies the differential inequality Ẇ ≤ −g0 β0 . By integra-
tion, we see that
|s(t)| ≤ |s(0)| − g0 β0 t
Therefore, the trajectory reaches the manifold s = 0 in finite time and, once on the
manifold, it cannot leave, as seen from the inequality V̇ ≤ −g0 β0 |s|. In summary,
the motion consists of a reaching phase during which trajectories starting off the
manifold s = 0 move toward it and reach it in finite time, followed by a sliding
phase during which the motion is confined to the manifold s = 0 and the dynamics
of the system are represented by the reduced-order model ẋ1 = −ax1 . A sketch of
the phase portrait is shown in Figure 10.1. The manifold s = 0 is called the sliding
manifold and u = −β(x) sgn(s) is the sliding mode control. The striking feature
of sliding mode control is its robustness with respect to h and g. We only need to
know the upper bound (x). During the sliding phase, the motion is completely
independent of h and g.
Ideally, maintaining the motion on the surface s = 0 takes places because the
control switches with infinite frequency. The average of this infinite-frequency os-
cillation produces the control that is needed to maintain the condition ṡ = 0. To
get a feel for this behavior, we simulate a sliding mode controller for the pendulum
equation (A.1):
θ̈ + sin θ + bθ̇ = cu
Suppose the uncertain parameters b and c satisfy the inequalities 0 ≤ b ≤ 0.2,
0.5 ≤ c ≤ 2, and we want to stabilize the pendulum at θ = π/2. Taking x1 = θ−π/2
and x2 = θ̇, we obtain the state model

ẋ1 = x2 , ẋ2 = − cos x1 − bx2 + cu

234 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

s= 0

Figure 10.1: A typical phase portrait under sliding mode control.

With s = x1 + x2 , we have

ṡ = x2 − cos x1 − bx2 + cu

From the inequality

   
 x2 − cos x1 − bx2   (1 − b)x2 − cos x1 
 =  ≤ 2(|x2 | + 1)
 c   c 

we take the control as

u = −(2.5 + 2|x2 |) sgn(s)
Simulation of the closed-loop system when b = 0.01, c = 0.5, and θ(0) = θ̇(0) = 0
is shown in Figure 10.2. The trajectory reaches the surface s = 0 at about t = 0.7
and from that point on the control oscillates with very high frequency.1 Passing u
through the low-pass filter 1/(0.1s+1) produces the filtered u signal, which balances
the right-hand side of the ṡ-equation to maintain the condition s ≡ 0.
The sliding mode controller simplifies if, in some domain of interest, h and g
satisfy the inequality  
 ax2 + h(x) 
  ≤ k1
 g(x) 

for some known constant k1 . In this case, we can take

u = −k sgn(s), k > k1

which takes the form of a simple relay. This form, however, usually leads to a finite
region of attraction, which can be estimated as follows: The condition sṡ ≤ 0 in the
1 Naturally, infinite frequency oscillation cannot be realized in numerical simulation.
10.1. SLIDING MODE CONTROL 235

2
0
1.5
−0.5

1
−1

s
θ

0.5 −1.5

0 −2
0 2 4 6 8 10 0 0.2 0.4 0.6 0.8 1
Time Time

5 5

3
Filtered u

0 2
u

−5 −1
0 0.2 0.4 0.6 0.8 1 0 2 4 6 8 10
Time Time

Figure 10.2: Simulation of the pendulum equation under sliding mode control.

set {|s| ≤ c} makes it positively invariant. From the equation

ẋ1 = x2 = −ax1 + s

and the function V0 = 12 x21 , we have

c
V̇0 = x1 ẋ1 = −ax21 + x1 s ≤ −ax21 + |x1 |c ≤ 0, ∀ |x1 | ≥
a
Thus, the set Ω = {|x1 | ≤ c/a, |s| ≤ c}, sketched in Figure 10.3, is positively in-
variant if |ax2 + h(x)|/g(x) ≤ k1 for all x ∈ Ω. Moreover, every trajectory starting
in Ω approaches the origin as t tends to infinity. By choosing c large enough, any
compact set in the plane can be made a subset of Ω. Therefore, if k can be chosen
arbitrarily large, the foregoing controller can achieve semiglobal stabilization.
In its ideal setup, sliding mode control requires the control input to oscillate
with very high (ideally infinite) frequency. This clearly cannot be achieved for
many physical inputs. For example, if the input is force or torque, it has to be
actuated by an electric motor and the maximum frequency of oscillation is limited
236 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

x2 Sliding manifold

s= 0
c

c/a
x1 s<0 s>0

Figure 10.3: Estimate of the Figure 10.4: Chattering due to de-

region of attraction. lay in control switching.

by the actuator bandwidth. Another deviation from the ideal setup arises because of
imperfections in switching devices, delays, and unmodeled high-frequency dynamics.
The trajectory does not stay identically on the sliding surface. Instead, it oscillates
around the surface in what is called chattering. Figure 10.4 shows how delays can
cause chattering. It depicts a trajectory in the region s > 0 heading toward the
sliding manifold s = 0. It first hits the manifold at point b. In ideal sliding mode
control, the trajectory should start sliding on the manifold from point b. In reality,
there will be a delay between the time the sign of s changes and the time the
control switches. During this delay period, the trajectory crosses the manifold into
the region s < 0. After a delay period, the control switches and the trajectory
reverses direction heading again towards the manifold. Once again it crosses the
manifold, and repetition of this process creates the “zig-zag” motion (oscillation)
shown in Figure 10.4. Chattering results in low control accuracy, high heat losses
in electrical power circuits, and high wear of moving mechanical parts. It may also
excite unmodeled high-frequency dynamics, which degrades the performance of the
system. The dashed lines in Figure 10.5 show the behavior of the sliding mode
control of the pendulum, which we designed earlier, when switching is delayed by
unmodeled actuator dynamics having the transfer function 1/(0.01s + 1)2 .
Chattering can be reduced by dividing the control into continuous and switching
components so as to reduce the amplitude of the switching one. Let ĥ(x) and ĝ(x)
be nominal models of h(x) and g(x), respectively. Taking

ax2 + ĥ(x)
u=− +v
ĝ(x)
results in
def
ṡ = a [1 − g(x)/ĝ(x)] x2 + h(x) − [g(x)/ĝ(x)] ĥ(x) + g(x)v = δ(x) + g(x)v
10.1. SLIDING MODE CONTROL 237

2
0
1.5
−0.5

1
θ

s
−1

0.5 −1.5

0 −2
0 2 4 6 8 10 0 1 2 3
Time Time
x 10−3
1.575 5

π/2 0
1.57

−5
θ

1.565
−10

1.56 −15
9 9.2 9.4 9.6 9.8 10 9.5 9.6 9.7 9.8 9.9 10
Time Time

Figure 10.5: Simulation of the pendulum equation under sliding mode control in the
presence of unmodeled actuator dynamics. The dashed lines are obtained when the
amplitude of the signum function is (2.5 + 2|x2 |) while the solid lines when it is (1 +
0.8|x2 |).

If the perturbation term δ(x) satisfies the inequality |δ(x)/g(x)| ≤ (x), we can take

v = −β(x) sgn(s)

where β(x) ≥ (x)+β0 . Because  is an upper bound on the perturbation term, it is

likely to be smaller than an upper bound on the whole function. Consequently, the
amplitude of the switching component would be smaller. For example, returning
to the pendulum equation and letting b̂ = 0 and ĉ be nominal values of b and c,
respectively, we can take the control as
−x2 + cos x1
u= +v
ĉ
which results in
ṡ = δ + cv
238 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

where    
δ 1−b 1 1 1
= − x2 − − cos x1
c c ĉ c ĉ
To minimize |(1 − b)/c − 1/ĉ|, we take ĉ = 1/1.2. Then,
 
δ 
  ≤ 0.8|x2 | + 0.8
c

and the control is taken as

u = 1.2 cos x1 − 1.2x2 − (1 + 0.8|x2 |) sgn(s)

The solid lines in Figure 10.5 show the behavior of this controller in the presence
of the unmodeled actuator dynamics 1/(0.01s + 1)2 . Comparison of the solid and
dashed lines shows that reducing the amplitude of the switching component has
reduced the amplitude of the s oscillation by a factor of 2 and reduced the error
|θ − π/2| by a similar factor. On the other hand, it did increase the reaching time
to the sliding surface.
One idea to avoid infinite-frequency oscillation of the control is to replace the
signum function by a high-slope saturation function; that is, the control is taken as
 
s
u = −β(x) sat
μ

where sat(·) is the saturation function defined by

⎧
⎨ y, if |y| ≤ 1
sat(y) =
⎩
sgn(y), if |y| > 1

and μ is a positive constant. The signum and saturation functions are shown in
Figure 10.6. The slope of the linear portion of sat(s/μ) is 1/μ. Good approximation
requires the use of small μ. In the limit, as μ → 0, sat(s/μ) approaches sgn(s). To
analyze the performance of the continuously implemented sliding mode controller,
we examine the reaching phase by using the function V = 12 s2 whose derivative
satisfies the inequality V̇ ≤ −g0 β0 |s| when |s| ≥ μ; that is, outside the boundary
layer {|s| ≤ μ}. Therefore, whenever |s(0)| > μ, |s(t)| will be strictly decreasing
until it reaches the set {|s| ≤ μ} in finite time and remains inside thereafter. In the
boundary layer, we have
ẋ1 = −ax1 + s
and |s| ≤ μ. The derivative of V0 = 12 x21 satisfies
μ
V̇0 = −ax21 + x1 s ≤ −ax21 + |x1 |μ ≤ −(1 − θ1 )ax21 , ∀ |x1 | ≥
aθ1
10.1. SLIDING MODE CONTROL 239

sgn (y) sat (y/μ)

1 1

y μ y

−1 −1

Figure 10.6: The signum nonlinearity and its saturation function approximation.

where 0 < θ1 < 1. Thus, the trajectory reaches the set Ωμ = {|x1 | ≤ μ/(aθ1 ), |s| ≤
μ} in finite time. In general, the controller does not stabilize the origin, but achieves
practical stabilization because the ultimate bound can be reduced by decreasing μ.
For better accuracy, we need to choose μ as small as possible, but a too small
value of μ will induce chattering in the presence of time delays or unmodeled fast
dynamics. Figure 10.7 shows the performance of the pendulum equation under the
continuously implemented sliding mode controller

u = −(2.5 + 2|x2 |) sat(s/μ)

for two different values of μ, with and without the unmodeled actuator 1/(0.01s +
1)2 . In the absence of actuator dynamics, reducing μ does indeed bring s close to
zero and reduces the error |θ −π/2|. However, in the presence of actuator dynamics,
the smaller value of μ induces chattering of s.
One special case where we can stabilize the origin without pushing μ too small
arises when h(0) = 0. In this case, the system, represented inside the boundary
layer by
ẋ1 = x2 , ẋ2 = h(x) − [g(x)β(x)/μ] (ax1 + x2 )

has an equilibrium point at the origin. We need to choose μ small enough to stabilize
the origin and make Ωμ a subset of its region of attraction. To illustrate this point,
consider the stabilization of the pendulum at θ = π. With x1 = θ − π, x2 = θ̇, and
s = x1 + x2 , we obtain
ṡ = x2 + sin x1 − bx2 + cu

Once again, assuming 0 ≤ b ≤ 0.2 and 0.5 ≤ c ≤ 2, it can be shown that

 
 (1 − b)x2 + sin x1 
  ≤ 2(|x1 | + |x2 |)
 c 
240 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

(a) (b)
2 0.1

1.5 0.05

1 0
θ

s
μ = 0.1
0.5 μ = 0.001 −0.05

0 −0.1
0 2 4 6 8 10 9.5 9.6 9.7 9.8 9.9 10
Time Time
(c) (d)
2 0.1

1.5 0.05

1 0
θ

0.5 −0.05

0 −0.1
0 2 4 6 8 10 9.5 9.6 9.7 9.8 9.9 10
Time Time

Figure 10.7: Simulation of the pendulum equation under continuously-implemented

sliding mode control without ((a) and (b)) and with ((c) and (d)) unmodeled actuator
dynamics.

Therefore, the control is taken as

u = −2(|x1 | + |x2 | + 1) sat(s/μ)

Inside the boundary layer {|s| ≤ μ}, the system is given by

ẋ1 = −x1 + s, ṡ = (1 − b)x2 + sin x1 − 2c(|x1 | + |x2 | + 1)s/μ

and has an equilibrium point at the origin. The derivative of the Lyapunov function
candidate V1 = 12 x21 + 12 s2 satisfies

V̇1 = −x21 + x1 s + (1 − b)(s − x1 )s + s sin x1 − 2c(|x1 | + |x2 | + 1)s2 /μ

T

|x1 | 1 −3/2 |x1 |

≤ −x1 + 3|x1 | |s| − (1/μ − 1)s = −
2 2
|s| −3/2 (1/μ − 1) |s|
10.1. SLIDING MODE CONTROL 241

Choosing μ < 4/13 = 0.308 makes V̇1 negative definite; hence, the origin is asymp-
totically stable. 

Consider the system

ẋ = f (x) + B(x)[G(x)u + δ(t, x, u)] (10.1)

where x ∈ Rn is the state, u ∈ Rm is the control input, f , B, and G are suf-

ficiently smooth functions in a domain D ⊂ Rn that contains the origin, and
G(x) is a positive definite symmetric matrix whose minimum eigenvalue satisfies
λmin (G(x)) ≥ λ0 > 0, for all x ∈ D. The function δ is piecewise continuous in t
and sufficiently smooth in (x, u) for (t, x, u) ∈ [0, ∞) × D × Rm . We assume that f
and B are known, while G and δ could be uncertain. Suppose f (0) = 0 so that, in
the absence of δ, the origin is an open-loop equilibrium point. Our goal is to design
a state feedback controller to stabilize the origin for all uncertainties in G and δ.
Let T : D → Rn be a diffeomorphism such that

∂T 0
B(x) = (10.2)
∂x I

where I is the m × m identity matrix.2 The change of variables

η
= T (x), η ∈ Rn−m , ξ ∈ Rm (10.3)
ξ

transforms the system into the form

η̇ = fa (η, ξ), ξ˙ = fb (η, ξ) + G(x)u + δ(t, x, u) (10.4)

The form (10.4) is usually referred to as the regular form. To design the sliding
mode controller, we start by designing the sliding manifold s = ξ − φ(η) = 0 such
that, when motion is restricted to the manifold, the reduced-order model

η̇ = fa (η, φ(η)) (10.5)

has an asymptotically stable equilibrium point at the origin. The design of φ(η)
amounts to solving a stabilization problem for the system η̇ = fa (η, ξ) with ξ viewed
as the control input. We assume that we can find a stabilizing, sufficiently smooth
function φ(η) with φ(0) = 0. Next, we design u to bring s to zero in finite time and
maintain it there for future time. Toward that end, let us write the ṡ-equation:
∂φ
ṡ = fb (η, ξ) − fa (η, ξ) + G(x)u + δ(t, x, u) (10.6)
∂η
2 If B(x) has full rank and the span of its columns is involutive, then Frobenius theorem [66]

ensures local existence of T (x) in the neighborhood of each point in D. If B is independent of x

and has full rank, then T (x) exists globally and takes the form T (x) = Lx where LB = col(0, I).
242 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

As we saw in Example 10.1, we can design u as a pure switching component or it may

contain an additional continuous component. If Ĝ(x) is a nominal model of G(x),
the continuous component of u could be −Ĝ−1 [fb − (∂φ/∂η)fa ]. In the absence of
uncertainty; that is, when δ = 0 and G is known, taking u = −G−1 [fb − (∂φ/∂η)fa ]
results in ṡ = 0, which ensures that s = 0 is an invariant manifold. We take the
control u as
u = ψ(η, ξ) + v (10.7)
where typical choices of ψ are ψ = 0 and ψ = −Ĝ−1 [fb − (∂φ/∂η)fa ]. Other choices
of ψ would arise if, for example, we decide to cancel part of the known term on
the right-hand side of (10.6) rather than the whole term. Substituting (10.7) into
(10.6) yields
ṡ = G(x)v + Δ(t, x, v) (10.8)
where
∂φ
Δ(t, x, v) = fb (η, ξ) − fa (η, ξ) + G(x)ψ(η, ξ) + δ(t, x, ψ(η, ξ) + v)
∂η
Assume we know a locally Lipschitz function (x) ≥ 0 and a constant κ0 , 0 ≤ κ0 < 1,
such that

Δ(t, x, v)

λmin (G(x)) ≤ (x) + κ0 v, ∀ (t, x, v) ∈ [0, ∞) × D × Rm (10.9)

Utilizing V = 12 sT s as a Lyapunov function candidate for (10.8), we obtain

V̇ = sT ṡ = sT G(x)v + sT Δ(t, x, v)
Taking
s
v = −β(x) (10.10)
s
where β is a locally Lipschitz function that satisfies
(x)
β(x) ≥ + β0 , ∀x∈D (10.11)
1 − κ0
for some β0 > 0, yields
V̇ = −β(x)sT G(x)s/s + sT Δ(t, x, v)
≤ λmin (G(x))[−β(x) + (x) + κ0 β(x)] s
= λmin (G(x))[−(1 − κ0 )β(x) + (x)] s ≤ −λmin (G(x))β0 (1 − κ0 )s
≤ −λ0 β0 (1 − κ0 )s
√
The inequality V̇ ≤ −λ0 β0 (1 − κ0 )s = −λ0 β0 (1 − κ0 ) 2V ensures that all tra-
jectories starting off the manifold s = 0 reach it in finite time and those on the
manifold cannot leave it.
The procedure for designing a sliding mode stabilizing controller can be sum-
marized by the following steps:
10.1. SLIDING MODE CONTROL 243

• Design the sliding manifold ξ = φ(η) to stabilize the reduced-order system

(10.5).

• Take the control u = ψ(η, ξ) + v, where typical choices of ψ are ψ = 0 and

ψ = −Ĝ−1 [fb − (∂φ/∂η)fa ].

• Estimate (x) and κ0 in (10.9), where Δ depends on the choice of ψ in the

previous step.

• Choose β(x) that satisfies (10.11) and take the switching (discontinuous) con-
trol v as given by (10.10).

This procedure exhibits model-order reduction because the main design task is per-
formed on the reduced-order system (10.5). The key feature of sliding mode control
is its robustness to matched uncertainties. During the reaching phase, the task
of forcing the trajectories toward the sliding manifold and maintaining them there
is achieved by the switching control (10.10), provided β(x) satisfies the inequality
(10.11). From (10.9), we see that (x) is a measure of the size of the uncertainty.
Since we do not require (x) to be small, the switching controller can handle fairly
large uncertainties, limited only by practical constraints on the amplitude of the
control signals. During the sliding phase, the motion of the system, as determined
by (10.5), is independent of the matched uncertain terms G and δ.
The sliding mode controller contains the discontinuous function s/s, which
reduces to the signum function sgn(s) when s is scalar. The discontinuity of the
controller raises theoretical as well as practical issues. Theoretical issues like ex-
istence and uniqueness of solutions and validity of Lyapunov analysis will have to
be examined in a framework that does not require the state equation to have lo-
cally Lipschitz right-hand-side functions.3 There are also practical issues like the
trajectory chattering around the sliding surface due to imperfections of switching
devices, delays, and unmodeled high-frequency dynamics, as illustrated in Exam-
ple 10.1, and the inability of the control to oscillate with very high frequency due
to limited actuator bandwidth. To avoid these issues we use a continuous approxi-
mation of s/s.4 Define the vector saturation function Sat(y) by
⎧
⎨ y, if y ≤ 1
Sat(y) =
⎩
y/y, if y > 1

3 To read about differential equations with discontinuous right-hand side, consult [38, 102, 122,

151].
4 See [106, 124, 142, 153] for a deeper study of sliding mode control and different approaches

to deal with these practical issues. While we do not pursue rigorous analysis of the discontinuous
sliding mode controller, the reader is encouraged to use simulation to examine the performance of
both the discontinuous controller and its continuous implementation.
244 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

It can be verified that Sat(y) is locally Lipschitz. We approximate s/s by Sat(s/μ)

with small positive constant μ;5 that is,
 
s
v = −β(x) Sat (10.12)
μ
where β(x) satisfies (10.11). In the region s ≥ μ, Sat(s/μ) = s/s. Therefore,

sT ṡ ≤ −λ0 β0 (1 − κ0 )s

which shows that whenever s(0) > μ, s(t) will decrease until it reaches the
set {s ≤ μ} in finite time and remains inside thereafter. The set {s ≤ μ} is
called the boundary layer. To study the behavior of η, we assume that there is a
(continuously differentiable) Lyapunov function V0 (η) that satisfies the inequalities

α1 (η) ≤ V0 (η) ≤ α2 (η) (10.13)

∂V0
fa (η, φ(η) + s) ≤ −α3 (η), ∀ η ≥ α4 (s) (10.14)
∂η
for all (η, ξ) ∈ T (D), where α1 to α4 are class K functions.6 Noting that

s ≤ c ⇒ V̇0 ≤ −α3 (η), for η ≥ α4 (c)

we define a class K function α by α = α2 ◦ α4 . Then,

V0 (η) ≥ α(c) ⇔ V0 (η) ≥ α2 (α4 (c)) ⇒ α2 (η) ≥ α2 (α4 (c))

⇒ η ≥ α4 (c) ⇒ V̇0 ≤ −α3 (η) ≤ −α3 (α4 (c))

which shows that the set {V (η) ≤ c0 } with c0 ≥ α(c) is positively invariant because
V̇ is negative on the boundary V (η) = c0 . (See Figure 10.8.) It follows that the set

Ω = {V0 (η) ≤ c0 } × {s ≤ c}, with c0 ≥ α(c) (10.15)

is positively invariant whenever c > μ and Ω ⊂ T (D). Choose μ, c, and c0 such that
c > μ, c0 ≥ α(c), and Ω is compact and contained in T (D). The set Ω serves as our
estimate of the “region of attraction.” For all initial states in Ω, the trajectories
will be bounded for all t ≥ 0. After some finite time, we have s(t) ≤ μ. It follows
from the foregoing analysis that V̇0 ≤ −α3 (α4 (μ)) for all V0 (η) ≥ α(μ). Hence, the
trajectories will reach the positively invariant set

Ωμ = {V0 (η) ≤ α(μ)} × {s ≤ μ} (10.16)

in finite time. The set Ωμ can be made arbitrarily small by choosing μ small
enough, which shows that the continuously implemented sliding mode controller
5 Smooth approximations are discussed in Exercise 14.11 of [74].
6 By Theorem 4.7, inequality (10.14) implies regional input-to-state stability of the system
η̇ = fa (η, φ(η) + s) when s is viewed as the input.
10.1. SLIDING MODE CONTROL 245

V0

c0
α (.)
α (c)

α (μ)
μ c
IsI

Figure 10.8: Representation of the sets Ω and Ωμ for a scalar s. V̇0 < 0 above the
α(·)-curve.

achieves practical stabilization. It achieves global practical stabilization if all the

assumptions hold globally and V0 (η) is radially unbounded because Ω can be chosen
arbitrarily large to include any initial state. We summarize our conclusions in the
next theorem.

Theorem 10.1 Consider the system (10.4). Suppose there exist φ(η), V0 (η), (x),
and κ0 , which satisfy (10.9), (10.13), and (10.14). Let u and v be given by (10.7)
and (10.12), respectively. Suppose μ, c > μ, and c0 ≥ α(c) are chosen such that
the set Ω, defined by (10.15), is compact and contained in T (D). Then, for all
(η(0), ξ(0)) ∈ Ω, the trajectory (η(t), ξ(t)) is bounded for all t ≥ 0 and reaches
the positively invariant set Ωμ , defined by (10.16), in finite time. Moreover, if the
assumptions hold globally and V0 (η) is radially unbounded, the foregoing conclusion
holds for any initial state. 3

The sets Ω and Ωμ of Theorem 10.1 estimate the region of attraction and the
ultimate bound, respectively. These estimates are typically conservative and further
analysis in a specific problem might give less conservative estimates. This point will
be illustrated in the next example. Part of the conservatism of Ω stems from the
fact that it is not simply an estimate of the region of attraction; it is an estimate
of the region where the condition sT ṡ < 0 holds. Trajectories outside this region
might approach the origin, even though s might be increasing for a while before
entering the region where sT ṡ < 0.

Example 10.2 To stabilize the magnetic levitation system (A.29) at x1 = 1, the

nominal steady-state control is uss = −1. Shifting the equilibrium point to the
origin and neglecting friction, we arrive at the model

m − mo mo
ẋ1 = x2 , ẋ2 = + u
m m
246 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

under the constraints x1 ≥ −1 and |u| ≤ 1. Suppose that |(m − mo )/mo | ≤ 1

3.
Taking s = x1 + x2 yields ẋ1 = −x1 + s. The derivative of V0 = 12 x21 satisfies

V̇0 = −x21 + x1 s ≤ −(1 − θ)x21 , ∀ |x1 | ≥ |s|/θ

where 0 < θ < 1. Thus, V0 satisfies (10.13) and (10.14) with α1 (r) = α2 (r) = 12 r2 ,
α3 (r) = (1 − θ)r2 , and α4 (r) = r/θ. Hence, α(r) = 12 (r/θ)2 . With c0 = α(c), the
sets Ω and Ωμ of (10.15) and (10.16) are given by

Ω = {|x1 | ≤ c/θ} × {|s| ≤ c}, Ωμ = {|x1 | ≤ μ/θ} × {|s| ≤ μ}

To meet the constraint x1 ≥ −1, c is limited to c ≤ θ. From

m − mo mo
ṡ = x2 + + u
m m
we see that (10.9) takes the form
   
 x2 + (m − mo )/m   m m − mo  1
 = x + ≤ (4|x2 | + 1)
 mo /m   mo 2 mo  3

We limit our analysis to the set Ω where |x2 | ≤ |x1 | + |s| ≤ c(1 + 1/θ). With
1/θ = 1.1, the foregoing inequality takes the form
 
 x2 + (m − mo )/m  8.4c + 1
 ≤
 mo /m  3

To meet the constraint |u| ≤ 1, we take u = −sat(s/μ). Then, inequality (10.11)

limits c to
8.4c + 1
< 1 ⇔ c < 0.238
3
With c = 0.23, Theorem 10.1 ensures that all trajectories starting in Ω stay in Ω
and enter Ωμ in finite time. Inside Ωμ , x1 satisfies |x1 | ≤ μ/θ = 1.1μ. The constant
μ can be chosen small enough to meet any specified tolerance. For example, if it is
required that the ultimate bound on |x1 | be 0.01, μ can be taken as 0.01/1.1 ≈ 0.009.
With further analysis inside Ωμ we can derive a less conservative estimate of the
ultimate bound of |x1 |. In Ωμ , the closed-loop system is represented by the linear
equation
m − mo mo (x1 + x2 )
ẋ1 = x2 , ẋ2 = −
m mμ
which has a unique equilibrium point at (x1 = μ(m − mo )/mo , x2 = 0) and its
matrix is Hurwitz. Thus, all trajectories inside Ωμ converge to the equilibrium
point as time tends to infinity. Since |(m − mo )/mo | ≤ 13 , an ultimate bound on
|x1 | can be taken as 0.34μ. For an ultimate bound of 0.01 we take μ = 0.029,
which is about three times the choice of μ we obtained by using Theorem 10.1.
10.1. SLIDING MODE CONTROL 247

We can also obtain a less conservative estimate of the region of attraction. Taking
V1 = 12 (x21 + s2 ), it can be shown that
 

mo  m − mo 
V̇1 ≤ −x21 +s −2 
1−  |s| ≤ −x21 + s2 − 1 |s|
m mo  2

for |s| ≥ μ, and

 
 m − mo  2 2
V̇1 ≤ −x21 2
+s +  |s| − mo s ≤ −x21 + s2 + 1 |s| − 3s
mo  m μ 2 4μ

for |s| ≤ μ. With μ = 0.029, it can be verified that V̇1 is less than a negative
number in the set {0.0012 ≤ V1 ≤ 0.12}. Therefore, all trajectories starting in
Ω1 = {V1 ≤ 0.12} enter Ω2 = {V1 ≤ 0.0016} in finite time. Since Ω2 ⊂ Ω, our
earlier analysis holds and the ultimate bound of |x1 | is 0.01. The new estimate of
the region of attraction, Ω1 , is larger than Ω. 

Theorem 10.1 shows that the continuously implemented sliding mode controller
achieves practical stabilization, which is the best we can expect, in general, because
the uncertainty δ could be nonvanishing at x = 0. If, however, δ vanishes at the
origin, then we may be able to show asymptotic stability of the origin, as we do in
the next theorem.

Theorem 10.2 Suppose all the assumptions of Theorem 10.1 are satisfied with
(0) = 0. Suppose further that the origin of η̇ = fa (η, φ(η)) is exponentially stable.
Then, there exists μ∗ > 0 such that for all 0 < μ < μ∗ , the origin of the closed-loop
system is exponentially stable and Ω is a subset of its region of attraction. Moreover,
if the assumptions hold globally and V0 (η) is radially unbounded, the origin will be
globally uniformly asymptotically stable. 3

Proof: Theorem 10.1 confirms that all trajectories starting in Ω enter Ωμ in finite
time. Inside Ωμ , the closed-loop system is given by

η̇ = fa (η, φ(η) + s), μṡ = β(x)G(x)s + μΔ(t, x, v) (10.17)

By (the converse Lyapunov) Theorem 3.8, there exists a Lyapunov function V1 (η)
that satisfies

∂V1 ∂V1
c1 η2 ≤ V1 (η) ≤ c2 η2 , fa (η, φ(η)) ≤ −c3 η2 ,
∂η ∂η ≤ c4 η

in some neighborhood Nη of η = 0. By the smoothness of fa we have

fa (η, φ(η) + s) − fa (η, φ(η)) ≤ k1 s

248 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

in some neighborhood N of (η, ξ) = (0, 0). Choose μ small enough that Ωμ ⊂ Nη ∩N .

Inside Ωμ , we have

β(x) T
sT ṡ = − s G(x)s + sT Δ(t, x, v)
μ

β(x)λmin (G(x)) κ0 β(x)s

≤ − s + λmin (G(x)) (x) +
2
s
μ μ
λ0 β0 (1 − κ0 )
≤ − s2 + λmin (G(x))(x)s
μ

Since G is continuous and  is locally Lipschitz with (0) = 0, we arrive at

λ0 β0 (1 − κ0 )
sT ṡ ≤ − s2 + k2 ηs + k3 s2
μ

Using the Lyapunov function candidate W = V1 (η) + 12 sT s,7 it can be shown that

λ0 β0 (1 − κ0 )
Ẇ ≤ −c3 η2 + c4 k1 ηs + k2 ηs + k3 s2 − s2
μ

T

η c3 −(c4 k1 + k2 )/2 η

= −
s −(c4 k1 + k2 )/2 λ0 β0 (1 − κ0 )/μ − k3 s

The right-hand side is negative definite for

4c3 λ0 β0 (1 − κ0 )
μ<
4c3 k3 + (c4 k1 + k2 )2

The rest of the proof is straightforward. 2

The basic idea of the foregoing proof is that, inside the boundary layer, the
control v = −β(x)s/μ acts as high-gain feedback for small μ. By choosing μ small
enough, the high-gain feedback stabilizes the origin.
We have emphasized the robustness of sliding mode control with respect to
matched uncertainties. What about unmatched uncertainties? Suppose equation
(10.1) is modified to

ẋ = f (x) + B(x)[G(x)u + δ(t, x, u)] + δ1 (x)

The change of variables (10.3) transforms the system into

η̇ = fa (η, ξ) + δa (η, ξ), ξ˙ = fb (η, ξ) + G(x)u + δ(t, x, u) + δb (η, ξ)

where [∂T /∂x]δ1 is partitioned into δa and δb . The term δb is added to the matched
uncertainty δ. Its only effect is to change the upper bound on Δ/λmin (G). The
7 System (10.17) is singularly perturbed and the composite Lyapunov function W is constructed
as in Section C.3.
10.1. SLIDING MODE CONTROL 249

term δa , on the other hand, is unmatched. It changes the reduced-order model on

the sliding manifold to

η̇ = fa (η, φ(η)) + δa (η, φ(η))

The design of φ will have to guarantee asymptotic stability of the origin η = 0 in

the presence of the uncertainty δa . This is a robust stabilization problem that may
be approached by other robust stabilization techniques such as high-gain feedback.
The difference between matched and unmatched uncertainties is that sliding mode
control guarantees robustness for any matched uncertainty provided an upper bound
is known and the needed control effort can be provided. There is no such guarantee
for unmatched uncertainties. We may have to restrict its size to robustly stabilize
the system on the sliding manifold. The next two examples illustrate these points.

Example 10.3 Consider the system

ẋ1 = x2 + θ1 x1 sin x2 , ẋ2 = θ2 x22 + x1 + u

where θ1 and θ2 are unknown parameters that satisfy |θ1 | ≤ a and |θ2 | ≤ b for some
known bounds a and b. The system is already in the regular form with η = x1 and
ξ = x2 . Uncertainty due to θ2 is matched, while uncertainty due to θ1 is unmatched.
We start with the system
ẋ1 = x2 + θ1 x1 sin x2
and design x2 to robustly stabilize the origin x1 = 0. This can be achieved with
x2 = −kx1 , k > a, because

x1 ẋ1 = −kx21 + θ1 x21 sin(−kx1 ) ≤ −(k − a)x21

With s = x2 + kx1 ,

ṡ = θ2 x22 + x1 + u + k(x2 + θ1 x1 sin x2 )

To cancel the known term on the right-hand side, we take u = −x1 −kx2 +v to obtain
ṡ = v + Δ(x), where Δ(x) = θ2 x22 + kθ1 x1 sin x2 . Because |Δ(x)| ≤ ak|x1 | + bx22 ,
we take β(x) = ak|x1 | + bx22 + β0 , with β0 > 0, and

u = −x1 − kx2 − β(x) sat(s/μ)

By Theorem 10.2, this controller with sufficiently small μ, globally stabilizes the
origin. 

In the foregoing example, we were able to use high-gain feedback to robustly stabilize
the reduced-order model for unmatched uncertainties that satisfy |θ1 | ≤ a, without
having to restrict a. In general, this may not be possible, as illustrated by the next
example.
250 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

Example 10.4 Consider the system

ẋ1 = x1 + (1 − θ1 )x2 , ẋ2 = θ2 x22 + x1 + u

where θ1 and θ2 are unknown parameters that satisfy |θ1 | ≤ a and |θ2 | ≤ b. We
start with the system
ẋ1 = x1 + (1 − θ1 )x2
and design x2 to robustly stabilize the origin x1 = 0. We note that the system
is not stabilizable at θ1 = 1. Hence, we must limit a to be less than one. Using
x2 = −kx1 , we obtain

x1 ẋ1 = x21 − k(1 − θ1 )x21 ≤ −[k(1 − a) − 1]x21

The origin x1 = 0 can be stabilized by taking k > 1/(1 − a). Taking s = x2 + kx1
and proceeding as in the previous example, we end up with

u = −(1 + k)x1 − kx2 − β(x) sat(s/μ)

where β(x) = bx22 + ak|x2 | + β0 with β0 > 0. 

We conclude this section by showing a different form of the sliding mode con-
troller when m > 1. If G(x) is diagonal with positive elements, we can design each
component vi of v as a scalar saturation function of si /μ. Equation (10.8) can be
written as
ṡi = gi (x)vi + Δi (t, x, v), 1≤i≤m
where si , vi , and Δi are the ith components of s, v, and Δ, respectively, and gi is
the ith diagonal element of G. Assume that
 
 Δi (t, x, v) 
gi (x) ≥ g0 > 0 and    ≤ (x) + κ0 max |vi |
gi (x)  1≤i≤m

for all (t, x, v) ∈ [0, ∞) × D × Rm and all 1 ≤ i ≤ m, where (x) ≥ 0 (a locally

Lipschitz function) and κ0 ∈ [0, 1) are known. Take

vi = −β(x) sat(si /μ), 1≤i≤m (10.18)

where β is a locally Lipschitz function that satisfies (10.11). When |si | ≥ μ, the
derivative of Vi = 12 s2i satisfies

V̇i = si gi (x)vi + si Δi (t, x, v) ≤ gi (x){si vi + |si |[(x) + κ0 max |vi |]}

1≤i≤m
≤ gi (x)[−β(x) + (x) + κ0 β(x)]|si | = gi (x)[−(1 − κ0 )β(x) + (x)]|si |
≤ gi (x)[−(x) − (1 − κ0 )β0 + (x)]|si | ≤ −g0 β0 (1 − κ0 )|si |
10.2. LYAPUNOV REDESIGN 251

The inequality V̇i ≤ −g0 β0 (1 − κ0 )|si | ensures that all trajectories reach the bound-
ary layer {|si | ≤ μ, 1 ≤ i ≤ m} in finite time. The sets Ω and Ωμ of (10.15) and
(10.16) are modified to

Ω = {V0 (η) ≤ c0 } × {|si | ≤ c, 1 ≤ i ≤ m}, with c0 ≥ α(c)

and
Ωμ = {V0 (η) ≤ α(μ)} × {|si | ≤ μ, 1 ≤ i ≤ m}
√
where α(r) = α2 (α4 (r m)). Results similar to Theorems 10.1 and 10.2 can be
proved for the control (10.18).8

10.2 Lyapunov Redesign

Consider the system
ẋ = f (x) + G(x)[u + δ(t, x, u)] (10.19)
where x ∈ Rn is the state and u ∈ Rm is the control input. The functions f , G,
and δ are defined for (t, x, u) ∈ [0, ∞) × D × Rm , where D ⊂ Rn is a domain that
contains the origin. We assume that f , G are locally Lipschitz and δ is piecewise
continuous in t and locally Lipschitz in x and u. The functions f and G are known,
while δ is unknown. A nominal model of the system is

ẋ = f (x) + G(x)u (10.20)

We proceed to design a state feedback controller u = φ(x) to stabilize the origin of

the nominal closed-loop system

ẋ = f (x) + G(x)φ(x) (10.21)

Suppose we know a Lyapunov function for (10.21); that is, we have a continuously
differentiable positive definite function V (x) such that

∂V
[f (x) + G(x)φ(x)] ≤ −W (x) (10.22)
∂x
for all x ∈ D, where W (x) is positive definite. Suppose further that, with u =
φ(x) + v, the uncertain term δ satisfies the inequality

δ(t, x, φ(x) + v) ≤ (x) + κ0 v, 0 ≤ κ0 < 1 (10.23)

where (x) ≥ 0 is locally Lipschitz. The bound (10.23), with known  and κ0 , is the
only information we need to know about δ. The function  is a measure of the size
of the uncertainty. Our goal is to design an additional feedback control v such that
8 See Theorems 14.1 and 14.2 of [74].
252 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

the overall control u = φ(x) + v stabilizes the origin of the actual system (10.19).
The design of v is called Lyapunov redesign.9
Under the control u = φ(x) + v, the closed-loop system

ẋ = f (x) + G(x)φ(x) + G(x)[v + δ(t, x, φ(x) + v)] (10.24)

is a perturbation of the nominal system (10.21). The derivative of V (x) along the
trajectories of (10.24) is given by
∂V ∂V ∂V
V̇ = (f + Gφ) + G(v + δ) ≤ −W (x) + G(v + δ)
∂x ∂x ∂x
where, for convenience, we did not write the arguments of the various functions.
Set wT = [∂V /∂x]G and rewrite the last inequality as

V̇ ≤ −W (x) + wT v + wT δ

The first term on the right-hand side is due to the nominal closed-loop system.
The second and third terms represent, respectively, the effect of the control v and
the uncertain term δ on V̇ . Due to the matching condition, the uncertain term δ
appears at the same point where v appears. Consequently, it is possible to choose
v to cancel the (possibly destabilizing) effect of δ on V̇ . We have

wT v + wT δ ≤ wT v + w δ ≤ wT v + w[(x) + κ0 v]

Taking
w
v = −β(x) · (10.25)
w
with a nonnegative locally Lipschitz function β, we obtain

wT v + wT δ ≤ −βw + w + κ0 βw = −β(1 − κ0 )w + w

Choosing β(x) ≥ (x)/(1 − κ0 ) for all x ∈ D yields

wT v + wT δ ≤ −w + w = 0

Hence, with the control (10.25), the derivative of V (x) along the trajectories of the
closed-loop system (10.24) is negative definite.
The controller (10.25) is a discontinuous function of the state x. It takes the
form of the sliding mode controller (10.11). Similar to sliding mode control, we
implement a continuous approximation of (10.25), given by
⎧

β(x)w
 ⎪ ⎨ −β(x)(w/w), if β(x)w > μ
v = −β(x) Sat = (10.26)
μ ⎪
⎩
−β (x)(w/μ),
2
if β(x)w ≤ μ
9 It is also known as min-max control [29]. More results on Lyapunov redesign are available
in [13, 28, 107].
10.2. LYAPUNOV REDESIGN 253

which is a locally Lipschitz function of x. With (10.26), the derivative of V along

the trajectories of the closed-loop system (10.24) will be negative definite whenever
β(x)w > μ. We only need to check V̇ when β(x)w ≤ μ. In this case,
 
β2 κ0 β 2 β2
wT v + wT δ ≤ − w2 + w + w2 ≤ (1 − κ0 ) − w2 + βw
μ μ μ
The term −(β 2 /μ)w2 + βw attains the maximum value μ/4 at βw = μ/2.
Therefore,
μ(1 − κ0 )
V̇ ≤ −W (x) + (10.27)
4
whenever β(x)w ≤ μ. On the other hand, when β(x)w > μ, V̇ satisfies
μ(1 − κ0 )
V̇ ≤ −W (x) ≤ −W (x) +
4
Thus, the inequality (10.27) is satisfied irrespective of the value of β(x)w. Let α1
to α3 be class K functions such that V (x) and W (x) satisfy the inequalities
α1 (x) ≤ V (x) ≤ α2 (x), W (x) ≥ α3 (x) (10.28)
for all x ∈ D. Then, for 0 < θ < 1,
V̇ ≤ −(1 − θ)α3 (x) − θα3 (x) + μ(1 − κ0 )/4
 
−1 μ(1 − κ0 ) def
≤ −(1 − θ)α3 (x), ∀ x ≥ α3 = μ0
4θ

Take r > 0 such that Br ⊂ D and choose μ < 4θα3 (α2−1 (α1 (r)))/(1 − κ0 ) so that
μ0 < α2−1 (α1 (r)). Application of Theorem 4.4 results in the following theorem,
which shows that the solutions of the closed-loop system are uniformly ultimately
bounded by a class K function of μ.
Theorem 10.3 Consider the system (10.19). Let D ⊂ Rn be a domain that con-
tains the origin and Br = {x ≤ r} ⊂ D. Let φ(x) be a stabilizing feedback control
for the nominal system (10.20) with a Lyapunov function V (x) that satisfies (10.22)
and (10.28) for x ∈ D, with some class K functions α1 , α2 , and α3 . Suppose the
uncertain term δ satisfies (10.23) for all t ≥ 0 and x ∈ D. Let v be given by
(10.26) and choose μ < 4θα3 (α2−1 (α1 (r)))/(1 − κ0 ) for 0 < θ < 1. Then, for any
x(t0 ) ∈ {V (x) ≤ α1 (r)}, the solution of the closed-loop system (10.24) satisfies
x(t) ≤ max {β1 (x(t0 ), t − t0 ), b(μ)} (10.29)
where β1 is a class KL function and b is a class K function defined by
b(μ) = α1−1 (α2 (μ0 )) = α1−1 (α2 (α3−1 (0.25μ(1 − κ0 )/θ)))
If all the assumptions hold globally and α1 belongs to class K∞ , then (10.29) holds
for any initial state x(t0 ). 3
254 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

In general, the continuous Lyapunov redesign (10.26) does not stabilize the origin
as its discontinuous counterpart (10.8) does. Nevertheless, it guarantees uniform
ultimate boundedness of the solutions. Since the ultimate bound b(μ) is a class K
function of μ, it can be made arbitrarily small by choosing μ small enough. Notice
that there is no analytical reason to require μ to be very small. The only analytical
restriction on μ is the requirement μ < 4θα3 (α2−1 (α1 (r)))/(1−κ0 ). This requirement
is satisfied for any μ when the assumptions hold globally and αi (i = 1, 2, 3) are
class K∞ functions. Of course, from a practical viewpoint, we would like to make
μ as small as feasible, because we would like to drive the state of the system to a
neighborhood of the origin that is as small as it could be. Exploiting the smallness of
μ in the analysis, we can arrive at a sharper result when the uncertainty δ vanishes
at the origin. Suppose there is a ball Ba = {x ≤ a}, a ≤ r, such that the following
inequalities are satisfied for all x ∈ Ba :

W (x) ≥ ϕ2 (x), β(x) ≥ β0 > 0, (x) ≤ 1 ϕ(x) (10.30)

where ϕ(x) is positive definite and 1 is a positive constant. Choosing μ < b−1 (a)
ensures that the trajectories of the closed-loop systems will be confined to Ba after
a finite time. When β(x)w ≤ μ, the derivative V̇ satisfies

β 2 (x)(1 − κ0 )
V̇ ≤ −W (x) − w2 + (x)w
μ
β 2 (1 − κ0 )
≤ −(1 − θ)W (x) − θϕ2 (x) − 0 w2 + 1 ϕ(x)w
μ

T

1 ϕ(x) 2θ −1 ϕ(x)

≤ −(1 − θ)W (x) −
2 w −1 2β02 (1 − κ0 )/μ w

where 0 < θ < 1. If μ < 4θβ02 (1 − κ0 )/21 the matrix of the quadratic form will be
positive definite; hence V̇ ≤ −(1 − θ)W (x). Since V̇ ≤ −W (x) ≤ −(1 − θ)W (x)
when β(x)w > μ, we conclude that V̇ ≤ −(1 − θ)W (x), which shows that the
origin is uniformly asymptotically stable. This conclusion is stated in the next
theorem.
Theorem 10.4 Assume (10.30) is satisfied, in addition to the assumptions of The-
orem 10.3. Then, for all μ < min{4θβ02 (1 − κ0 )/21 , b−1 (a)}, where 0 < θ < 1, the
origin of the closed-loop system (10.24) is uniformly asymptotically stable. If α1 to
α3 of (10.28) take the form αi (r) = ki rc , the origin will be exponentially stable. 3
Theorem 10.4 is particularly useful when the origin of the nominal closed-loop
system (10.21) is exponentially stable and the perturbation δ(t, x, u) is Lipschitz
in x and u and vanishes at (x = 0, u = 0). In this case, ϕ(x) is proportional to
x and the uncertain term satisfies (10.23) with (x) ≤ 1 ϕ(x). In general, the
inequality (x) ≤ 1 ϕ(x) may require more than just a vanishing perturbation at
the origin. For example if, in a scalar case, ϕ(x) = |x|3 and (x) = |x|, then (x)
cannot be bounded by 1 ϕ(x).
10.2. LYAPUNOV REDESIGN 255

The stabilization result of Theorem 10.4 is dependent on the choice of β to satisfy

β(x) ≥ β0 > 0. Under this condition, the feedback control law (10.26) acts in the
region βw ≤ μ as a high-gain feedback controller v = −βw/μ. Such high-gain
feedback can stabilize the origin when (10.30) is satisfied. It can be shown that if
β does not satisfy the condition β(x) ≥ β0 > 0, the feedback control may fail to
stabilize the origin.10

Example 10.5 Reconsider the pendulum equation of Example 9.4.

ẋ1 = x2 , ẋ2 = − sin(x1 + δ1 ) − b0 x2 + cu

where x1 = θ−δ1 , x2 = θ̇, and the uncertain parameters b0 and c satisfy 0 ≤ b0 ≤ 0.2
and 0.5 ≤ c ≤ 2. We want to stabilize the pendulum at θ = δ1 by stabilizing the
origin x = 0. This system is feedback linearizable and can be written in the form

ẋ = Ax + B[− sin(x1 + δ1 ) − b0 x2 + cu]

where the pair (A, B) represents a chain of two integrators. With nominal param-
eters b̂0 = 0 and ĉ, a nominal stabilizing feedback control can be taken as
 
1
φ(x) = [sin(x1 + δ1 ) − k1 x1 − k2 x2 ]
ĉ
 
where K = k1 k2 is chosen such that A − BK is Hurwitz. With u = φ(x) + v,
the uncertain term δ is given by
   
c − ĉ b0 c − ĉ
δ= [sin(x1 + δ1 ) − k1 x1 − k2 x2 ] − x2 + v
ĉ2 ĉ ĉ
Hence,
|δ| ≤ 0 + 1 |x1 | + 2 |x2 | + κ0 |v|
where
       
 (c − ĉ) sin δ1   c − ĉ  b0  c − ĉ   c − ĉ 
0 ≥  ,
 1 ≥  2  (1 + k1 ), 2 ≥ +  2  k2 , κ0 ≥  
ĉ2 ĉ ĉ ĉ ĉ 
Assuming κ0 < 1, we take β(x) as
0 + 1 |x1 | + 2 |x2 |
β(x) ≥ β0 + , β0 > 0
1 − κ0
The nominal closed-loop system has a Lyapunov function V (x) = xT P x, where P is
the solution of the Lyapunov equation P (A−BK)+(A−BK)T P = −I. Therefore,
w = 2xT B T P = 2(p12 x1 + p22 x2 ) and the control

u = φ(x) − β(x) sat (β(x)w/μ)

10 See Exercise 14.20 of [74].
256 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

√
achieves global ultimate boundedness with ultimate bound proportional to μ. If
sin δ1 = 0, we take 0 = 0 and the origin of the closed-loop system will be globally
exponentially stable.
In Example 9.4, we analyzed the same system under the control u = φ(x).
Comparing the results of the two examples shows the contribution of the additional
control component v. In Example 9.4, we had to restrict the uncertainty such that
  

 c − ĉ 
  1 + k2 + k2 ≤  1
 ĉ  1 2
2 p212 + p222
This restriction has now been removed. When sin δ1 = 0, we showed there that the
ultimate bound is proportional to | sin(δ1 )(c − ĉ)/ĉ|. With the current control, the
√
ultimate bound is proportional to μ; hence it can be made arbitrarily small by
choosing μ small enough.
Consider a case where the pendulum is to be stabilized at δ1 = π/2. To minimize
 ĉ = (2+0.5)/2 = 1.25, which allows us to take κ0 = |c−ĉ/ĉ| = 0.6.
|(c−ĉ)/ĉ|, take 11

With K = 1 2 , A − BK has multiple eigenvalues at −1 and the solution of the

Lyapunov equation yields w = x1 + x2 . With β0 = 0.3, β(x) is given by

β(x) = 2.4|x1 | + 2.8|x2 | + 1.5

and
u = 0.8(cos x1 − x1 − 2x2 ) − β(x) sat (β(x)w/μ)
The parameter μ is chosen small enough to meet the requirement on the ultimate
bound. For example, suppose it is required that x(t) be ultimately inside the set
{|x1 | ≤ 0.01, |x2 | ≤ 0.01}. We use the ultimate bound provided by Theorem 10.3
with θ = 0.9 to choose μ. The Lyapunov function V (x) = xT P x satisfies (10.28)
with α1 (r) = λmin (P )r2 , α2 (r) = λmax (P )r2 , and α3 (r) = r2 . Hence,
√
b(μ) = α1−1 (α2 (α3−1 (0.25μ(1 − κ0 )/θ))) ≈ 0.8 μ
√
Choosing μ < 0.01/0.8 ensures that the ball {x ≤ b} is inside the set {|x1 | ≤
0.01, |x2 | ≤ 0.01}. A less conservative estimate of μ can be obtained by examining
the trajectories inside {||x ≤ b}. It can be shown that the trajectories inside
this set converge to an equilibrium point (x̄1 ≈ μ(0.8c − 1)/(2.25c), x̄2 = 0).12
With c ∈ [0.5, 2], |μ(0.8c − 1)/(2.25c)| ≤ 0.53μ. Hence, it is sufficient to choose
μ < 0.01/0.53. Simulation results with b0 = 0.01, c = 0.5, and μ = 0.01 are shown
in Figure 10.9. Figure 10.9(a) shows the response of θ when θ(0) = θ̇(0) = 0.
Figure 10.9(b) shows that the phase portrait resembles sliding mode control where
trajectories reach the surface w = 0 in finite time, then slide on it towards the
origin. Even though the controller was not designed to make wẇ negative, it can
be shown that when β|w| ≥ μ, wẇ ≤ −w2 . 
11 Recall from Example 9.4 that for the same numbers, |(c − ĉ)/ĉ| has to be less than 0.3951.
12 Analysis inside {x ≤ b} use the fact that, in this example, wẇ ≤ −w2 whenever β|w| ≥ μ
and β is almost constant. Therefore, the trajectories enter the set {β|w| ≤ μ} in finite time.
10.3. HIGH-GAIN FEEDBACK 257

(a) (b)
3
π/2
1.5 2

1
1

x2
0
θ

−1
0.5
−2

0 −3
0 2 4 6 −3 −2 −1 0 1 2 3
Time x1

Figure 10.9: Simulation results for Example 10.5.

10.3 High-Gain Feedback

High-gain feedback is a classical tool for stabilization and disturbance rejection in
the face of model uncertainty. Both the sliding mode controller of Section 10.1
and the Lyapunov redesign controller of Section 10.2 can be replaced by high-gain
feedback controllers.
Starting with sliding mode control, suppose all the assumptions of Theorem 10.1
are satisfied. Replace the control v = −β(x) Sat(s/μ) of (10.13) by
β(x)s
v=− (10.31)
μ
The derivative of V = 12 sT s satisfies
β(x) T
V̇ = − s G(x)s + sT Δ(t, x, v)
μ
β(x) β(x)
≤ − λmin (G(x))s2 + λmin (G(x))(x)s + λmin (G(x))κ0 s2
μ μ
 

s
= λmin (G(x)) − − 1 β(x)(1 − κ0 )s − β0 (1 − κ0 )s
μ
≤ −λ0 β0 (1 − κ0 )s, for s ≥ μ
Hence, the trajectories reach the boundary layer {s ≤ μ} in finite time. From
this point on, the analysis is identical to sliding mode control because inside the
boundary layer Sat(s/μ) = s/μ. It follows that Theorems 10.1 and 10.2 hold for
the high-gain feedback control (10.31).
258 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

It is interesting to compare sliding mode control with high-gain feedback. For

convenience, we limit our comparison to single-input systems and to the case when
both controllers have ψ = 0. Both the sliding mode control u = −β(x) sat(s/μ) and
the high-gain feedback control u = −β(x)s/μ drive the trajectory to the boundary
layer {|s| ≤ μ} and maintains it there, but the high-gain feedback does it faster
because the ṡ equation takes the form

μṡ = −G(x)s + μΔ

The appearance of μ on the left-hand side shows that, for small μ, s will move
rapidly towards the boundary layer. This fast motion towards the boundary layer
comes at the expense of a large spike in the control signal in the initial transient.
This is different than sliding mode control where |u| ≤ β. Reducing the steady-state
error when u = −β sat(s/μ) can be achieved by decreasing μ without increasing
the control magnitude, while in the control u = −βs/μ, decreasing the steady-state
error by decreasing μ results in a large initial spike in the control. These differences
are illustrated in the next example. We note that in the special case when β is
constant, the sliding mode control u = −β sat(s/μ) can be viewed as saturated
high-gain feedback control, where the linear feedback −(β/μ)s passes through a
saturation function that saturates at ±β.
Example 10.6 In Example 10.1 we designed the sliding mode controller

u = −(2.5 + 2|θ̇|) sat(s/μ)

where s = θ − π/2 + θ̇, to practically stabilize the pendulum at θ = π/2. The

corresponding high-gain feedback controller is

u = −(2.5 + 2|θ̇|)(s/μ)

Figure 10.10 compares the performance of the two controllers when b = 0.01, c =
0.5, and μ = 0.1. Figure 10.10(a) demonstrates that s moves faster towards the
boundary layer under high-gain feedback. Figure 10.10(c) shows the spike in the
control signal under high-gain feedback, compared with the control signal under
sliding mode control, which is shown in Figure 10.10(b). 
Turning now to Lyapunov redesign, suppose all the assumptions of Theorem 10.3
are satisfied and replace v = −β(x) Sat(β(x)w/μ) by

β 2 (x)w
v=− (10.32)
μ

This control is used in Section 10.2 when β(x)w ≤ μ where it is shown that V̇
satisfies (10.27). The only difference here is that (10.32) is used for all w. Therefore,
(10.27) is still satisfied and Theorems 10.3 and 10.4 hold for the high-gain feedback
control (10.32).
10.4. EXERCISES 259

(a) (b) (c)

5
0 40
4
3 30
−0.5
2

n
20
s

−1 1
0 10
−1.5 −1
0
−2
0 0.5 1 0 2 4 0 0.2 0.4
Time Time Time

Figure 10.10: Comparison of the sliding mode controller u = −(2.5 + 2|θ̇|) sat(s/0.1)
(solid line) with the high-gain feedback controller u = −(2.5 + 2|θ̇|)(s/0.1) (dashed
line) for the pendulum equation of Example 10.6 when θ(0) = θ̇(0) = 0.

10.4 Exercises
10.1 For each of the following systems, use sliding mode control to design a locally
Lipschitz, globally stabilizing state feedback controller. The constants θ1 , θ2 , and
the function q(t) satisfy the bounds 0 ≤ θ1 ≤ a, 0 < b ≤ θ2 ≤ c, and |q(t)| ≤ d, for
known constants a, b, c, d. You need to verify that the controller will be stabilizing
for μ < μ∗ for some μ∗ > 0 but you do not need to estimate μ∗ .
(1) ẋ1 = x2 + 2 sin x1 , ẋ2 = θ1 x21 + θ2 u
(2) ẋ1 = x2 + sin x1 , ẋ2 = θ1 x1 x2 + u
(3) ẋ1 = (1 + x21 )x2 , ẋ2 = x1 + θ2 u
(4) ẋ1 = x21 + x2 , ẋ2 = θ1 h(x) + u, |h(x)| ≤ x
(5) ẋ1 = x31 + x2 , ẋ2 = q(t)x3 + u, ẋ3 = x1 − x3
(6) ẋ1 = x1 − θ2 u, ẋ2 = 2x2 + θ2 u

10.2 For each of the following systems, use sliding mode control to design a locally
Lipschitz state feedback controller that ensures global ultimate boundedness with
x(t) ultimately in the set {|x1 | ≤ δ1 , |x2 | ≤ δ1 } for a given δ1 > 0. The constants
θ1 , θ2 , and the function q(t) satisfy the bounds 0 ≤ θ1 ≤ a, 0 < b ≤ θ2 ≤ c, and
|q(t)| ≤ d, for known constants a, b, c, d. You need to choose μ in terms of δ1 .
(1) ẋ1 = x1 x2 , ẋ2 = x1 + θ2 u
260 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

(2) ẋ1 = x2 , ẋ2 = sin x1 + u + q(t)

(3) ẋ1 = sin x1 + x2 , ẋ2 = u + q(t)
(4) ẋ1 = x1 − θ2 u + q(t), ẋ2 = 2x2 + θ2 u − q(t)
(5) ẋ1 = −x1 + tanh(x2 ), ẋ2 = x2 + x3 , ẋ3 = u + q(t)
(6) ẋ1 = −x1 + tanh(x2 ), ẋ2 = θ1 x2 + u1 , ẋ3 = u2 + q(t)

10.3 Repeat Exercise 10.1 using Lyapunov redesign.

10.4 Repeat Exercise 10.2 using Lyapunov redesign.

10.5 A normalized model of a pendulum whose suspension point is subjected to

a time-varying bounded horizontal acceleration is given by

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + cu + h(t)

where b ∈ [0, 0.4], c ∈ [0.5, 1.5], and |h(t)| ≤ 1. It is required to design locally
Lipschitz state feedback controller such that, for any initial state, x(t) ∈ {|x1 | ≤
0.01, |x2 | ≤ 0.01} for all t ≥ T for some finite time T . Design the controller using

(a) sliding mode control; (b) Lyapunov redesign; (c) high-gain feedback

10.6 Consider the magnetic levitation system (A.29) with uncertain parameters
b ∈ [0, 0.1] and c ∈ [0.8, 1.2]. It is required to regulate x1 to r ∈ [0.5, 1.5], while
meeting the constraint u ∈ [−2, 0]. The steady-state regulation error should satisfy
|x1 − r| ≤ 0.01. Design a locally Lipschitz state feedback controller using

(a) sliding mode control; (b) Lyapunov redesign

10.7 Consider the magnetic levitation system (A.30)–(A.32) with α = 1.5, β = 2

and uncertain parameters b ∈ [0, 0.1] and c ∈ [0.8, 1.2]. It is required to regulate
x1 to 1 with steady-state error |x1 − 1| ≤ 0.01. Design a locally Lipschitz state
feedback controller. Hint: Transform the system into the controller form using the
nominal parameters b̂ = 0 and ĉ = 1.

10.8 Consider the biochemical reactor (A.19) with uncertain ν(x2 ), but with a
known upper bound νmax such that ν(x2 ) ≤ νmax over the domain of interest. It is
desired to design a state feedback controller to stabilize the system at x = col(1, 1).
(a) Show that the change of variables η = (x2 − α)/x1 , ξ = x2 , which is valid in
the region x1 > 0, transforms the system into the regular form.
(b) Let α = 23 and νmax = 1.5. Using sliding mode control, design a controller of
the form u = −k sat((x2 −1)/μ), with positive constants k and μ, such that the
system is stabilized at an equilibrium point that is O(μ) close to x = col(1, 1),
with a region of attraction that includes the set {0.1 ≤ x1 ≤ 2, 0 ≤ x2 ≤ 2}.
10.4. EXERCISES 261

(c) Construct the phase portrait of the closed-loop system in the region {x1 ≥
0, x2 ≥ 0} with ν(x2 ) defined by (A.20), α = 23, β = 0.39, γ = 0.57, and
μ = 0.01. For comparison, construct also the phase portrait when u = 1.

10.9 Consider the electrostatic microactuator (A.33) with uncertain ζ ∈ (0, 0.5]
and T ∈ [0.1, 0.5]. Using sliding mode control, design a locally Lipschitz state
feedback controller to stabilize the plate at x1 = r, where r ∈ [0.1, 0.9]. Simulate
the closed-loop system with r = 0.5, ζ = 0.1 and T = 0.2. Verify that the controller
achieves stabilization, not only ultimate boundedness.

10.10 Consider the inverted pendulum (A.47) with uncertain a ∈ [0.5, 1.5]. Using
sliding mode control, design a locally Lipschitz state feedback controller to stabilize
the pendulum at θ = 0. Use simulation with a = 1 to find the largest |x1 (0)|, when
x(0) = col(x1 (0), 0), for which the pendulum is stabilized.

10.11 A simplified model of an underwater vehicle in yaw is given by [39]

ψ̈ + aψ̇|ψ̇| = u

where ψ is the heading angle, u is a normalized torque, and a ∈ [0.5, 1.5] is a

parameter. Using sliding mode control, design a locally Lipschitz state feedback
controller to globally stabilize the system at the equilibrium point (ψ = ψr , ψ̇ = 0).

10.12 Consider the robot manipulator (A.34) and let M̂ , Ĉ and ĝ be nominal
models of M , C, and g, respectively. For the purpose of sliding mode control, take
s = Λ(q − qr ) + q̇, where qr is the desired set point and Λ = ΛT > 0.

(a) Show that sliding mode control that achieves global practical stabilization for
sufficiently small μ can be taken as

u = −β(x) Sat(s/μ) or u = Ĉ q̇ + ĝ − M̂ Λq̇ − β(x) Sat(s/μ)

and choose β in each case. Are there conditions under which either control
can achieve stabilization?

(b) Apply both controllers to the two-link manipulator defined by (A.36) and
(A.37) with nominal data (A.38) and actual data (A.39). It is desired to
regulate the manipulator to qr = (π/2, π/2) under the control constraints
|u1 | ≤ 6000 and |u2 | ≤ 5000. Compare the performance of the two controllers.

10.13 Consider the TORA system (A.49)–(A.52). In this exercise, we design a

sliding mode controller13 to stabilize the origin.
13 The design uses ideas from the passivity-based design of [121].
262 CHAPTER 10. ROBUST STATE FEEDBACK STABILIZATION

(a) Show that

mLx2 cos x1
η1 = x1 , η2 = x3 , η3 = x4 + , ξ = x2
m+M
is a global diffeomorphism that transforms the system into the regular form
−mLξ cos η1 −kη2
η̇1 = ξ, η̇2 = + η3 , η̇3 =
m+M m+M
1  
ξ˙ = (m + M )u − mL cos η1 (mLξ 2 sin η1 − kη2 )
Δ(η1 )
(b) Show that the derivative of

1 2 (m + M )k2 2 (m + M )2 k2 2
V0 (η) = 2 k1 η1 + η2 + η3
mL mLk
with positive constants k1 and k2 , is given by V̇0 = −φ(η)ξ, where φ(η) =
−k1 η1 + k2 η2 cos η1 .
(c) Verify that with ξ = φ(η), the origin η = 0 of the η̇-equation is globally asymp-
totically stable and locally exponentially stable.
(d) In view of (c), take
s = ξ − φ(η) = ξ + k1 η1 − k2 η2 cos η1 = x2 + k1 x1 − k2 x3 cos x1
Note that s is independent of the system parameters. Choose β(x) such that
u = −β(x) sat(s/μ) globally stabilizes the origin for sufficiently small μ.
10.14 Simulate the TORA system (A.49)–(A.52) with the (A.53) data to compare
the passivity-based control of Exercise 9.13 with the sliding mode control of the
previous exercise. In this comparison, the functions φ1 and φ2 of Exercise 9.13 are
taken as φi (y) = Ui sat(ki y) so that u = −U1 sat(k1 x1 ) − U2 sat(k2 x2 ). The sliding
mode control u = −β sat(s/μ) is taken with constant β. Therefore, it only achieves
regional stabilization versus global stabilization for the passivity-based control. The
comparison will be conducted for initial states in the region of attraction of the
sliding mode control. We limit both controllers by the control constraint |u| ≤ 0.1,
which is met by taking β = 0.1 and U1 + U2 = 0.1.
(a) With initial state x(0) = (π, 0, 0.025, 0), tune the parameters k1 , k2 , U1 , and
U2 of the passivity-based control to make the settling time as small as you
can. You should be able to achieve a settling time of about 30 sec.
(b) With initial state x(0) = (π, 0, 0.025, 0), tune the parameters k1 , k2 , and μ of
the sliding mode control to make the settling time as small as you can. You
should be able to achieve a settling time of about 4 sec.
(c) Compare the performance of the two controllers.
Chapter 11

Nonlinear Observers

In this chapter we study the design of observers to estimate the state x of the
nonlinear system
ẋ = f (x, u), y = h(x)
where u is a given input and y is the measured output. The observers will take the
general form
x̂˙ = f (x̂, u) + H(·)[y − h(x̂)]
where the observer gain H(·) could be a constant or time-varying matrix. The term
f (x̂, u) in the observer equation is a prediction term that duplicates the dynamics
of the system, while H[y − h(x̂)] is a correction term1 that depends on the error in
estimating the output y by h(x̂). For the linear system

ẋ = Ax + Bu, y = Cx

the estimation error x̃ = x − x̂ satisfies the linear equation

x̃˙ = (A − HC)x̃

If the pair (A, C) is detectable, that is, observable or having unobservable eigenval-
ues with negative real parts, the matrix H can be designed such that A − HC is
Hurwitz, which implies that limt→∞ x̃(t) = 0; hence, the estimate x̂(t) asymptoti-
cally approaches x(t) as t tends to infinity. For nonlinear systems, we present differ-
ent approaches to the design of nonlinear observers. The first two sections present
approaches that are based on linearization. The local observer of Section 11.1 is de-
signed by linearization about an equilibrium point of the system ẋ = f (x, u); hence
it is guaranteed to work only when x and u are sufficiently close to their equilibrium
values, in addition to requiring x̃(0) to be sufficiently small. The observer of Sec-
tion 11.2 is a deterministic version of the Extended Kalman Filter from nonlinear
1 Or innovation term in the language of estimation theory.

263
264 CHAPTER 11. NONLINEAR OBSERVERS

estimation theory.2 Here linearization is taken about the estimate x̂; therefore, the
observer gain is time varying and has to be calculated in real time since H(t) at
any time t depends on x̂(t). The observer still requires the initial estimation error
x̃(0) to be sufficiently small, but now x(t) and u(t) can be any well-defined trajec-
tories that have no finite escape time. The observer of Section 11.3 is developed for
nonlinear systems in the observer form of Section 8.3 and has linear error dynamics
identical to the case of linear observers. Hence, convergence of the estimation error
x̃ is global; that is, limt→∞ x̃(t) = 0 for all initial conditions x̃(0). It is also shown
that the same design will work for a more general system that has an additional
Lipschitz nonlinearity provided the Lipschitz constant is sufficiently small.3
The observers of the first three sections assume perfect knowledge of the non-
linear functions f and h. Because, in all three cases, we will show that the error
dynamics have an exponentially stable origin, we can see from the robustness anal-
ysis of Section 4.3 that for sufficiently small bounded perturbations of f or h, the
estimation error x̃(t) will be ultimately bounded by a small upper bound. How-
ever, there is no guarantee that these observers will function in the presence of
large perturbations. The high-gain observer of Section 11.4 applies to a special
class of nonlinear systems, but can be designed to be robust to a certain level of
uncertainty that cannot be tolerated in the observers of Section 11.1 to 11.3. The
robustness is achieved by designing the observer gain high enough, which makes the
estimation-error dynamics much faster than those of the system ẋ = f (x, u). The
fast dynamics play an important role when the observer is used in feedback control,
as we shall see in Section 12.4.

11.1 Local Observers

Consider the nonlinear system

ẋ = f (x, u), y = h(x) (11.1)

together with the observer

x̂˙ = f (x̂, u) + H[y − h(x̂)] (11.2)

where x ∈ Rn , u ∈ Rm , y ∈ Rp , f and h are twice continuously differentiable in

their arguments, and H is an n × p constant matrix. The estimation error x̃ = x − x̂
satisfies the equation

x̃˙ = f (x, u) − f (x̂, u) − H[h(x) − h(x̂)] (11.3)

Equation (11.3) has an equilibrium point at x̃ = 0 and our goal is to design the
observer gain H to make this equilibrium point exponentially stable. We seek a
2 See [47] or [127].
3 See [5, 46, 72, 123] for more approaches to nonlinear observers.
11.1. LOCAL OBSERVERS 265

local solution for sufficiently small x̃(0). Linearization of (11.3) at x̃ = 0 results

in the linear time-varying system

˙x̃ = ∂f (x(t), u(t)) − H ∂h (x(t)) x̃ (11.4)

∂x ∂x

Designing a constant matrix H to stabilize the origin of the foregoing time-varying

system is almost impossible, especially in view of the fact that it should work for
all initial states x(0) and all inputs u(t) in a certain domain. We assume that there
are vectors xss ∈ Rn and uss ∈ Rm such that the system (11.1) has an equilibrium
point x = xss when u = uss and that the output y = 0 at equilibrium; that is,

0 = f (xss , uss ), 0 = h(xss )

Furthermore, we assume that given ε > 0, there exist δ1 > 0 and δ2 > 0 such that

x(0) − xss  ≤ δ1 and u(t) − uss  ≤ δ2 ∀ t ≥ 0

ensure that x(t) is defined for all t ≥ 0 and satisfies x(t) − xss  ≤ ε for all t ≥ 0.
Let
∂f ∂h
A= (xss , uss ), C= (xss )
∂x ∂x
and assume that the pair (A, C) is detectable. Design H such that A − HC is
Hurwitz.
Lemma 11.1 For sufficiently small x̃(0), x(0) − xss , and supt≥0 u(t) − uss ,

lim x̃(t) = 0
t→∞

3
Proof: By (B.6), we have
 1
∂f
f (x, u) − f (x̂, u) = (x − σ x̃, u) dσ x̃
0 ∂x
Hence,

f (x, u) − f (x̂, u) − Ax̃ =

 1

∂f ∂f ∂f ∂f
(x − σ x̃, u) − (x, u) + (x, u) − (x , u ) dσ x̃
∂x ∂x ∂x ∂x
ss ss
0
≤ L1 ( 12 x̃ + x − xss  + u − uss )x̃

where L1 is a Lipschitz constant of [∂f /∂x]. Similarly, it can be shown that

h(x) − h(x̂) − C x̃ ≤ L2 ( 12 x̃ + x − xss )x̃

266 CHAPTER 11. NONLINEAR OBSERVERS

where L2 is a Lipschitz constant of [∂h/∂x]. Therefore, the estimation-error dy-

namics can be written as

x̃˙ = (A − HC)x̃ + Δ(x, u, x̃) (11.5)

where
Δ(x, u, x̃) ≤ k1 x̃2 + k2 (ε + δ2 )x̃
for some positive constants k1 and k2 . Let V = x̃T P x̃, where P is the positive
definite solution of the Lyapunov equation P (A − HC) + (A − HC)T P = −I. Using
V as a Lyapunov function candidate for (11.5), we obtain

V̇ ≤ −x̃2 + c4 k1 x̃3 + c4 k2 (ε + δ2 )x̃2

where c4 = 2P . Thus

V̇ ≤ − 13 x̃2 , for c4 k1 x̃ ≤ 1

3 and c4 k2 (ε + δ2 ) ≤ 1
3

which shows that for sufficiently small x̃(0), ε, and δ2 , the estimation error con-
verges to zero as t tends to infinity. Smallness of ε can be ensured by choosing δ1
and δ2 sufficiently small. 2

11.2 The Extended Kalman Filter

Consider the nonlinear system

ẋ = f (x, u), y = h(x) (11.6)

where x ∈ Rn , u ∈ Rm , y ∈ Rp , f and h are twice continuously differentiable in

their arguments, and x(t) and u(t) are bounded for all t ≥ 0. Consider the observer

x̂˙ = f (x̂, u) + H(t)[y − h(x̂)] (11.7)

with time-varying observer gain H(t). The estimation error x̃ = x − x̂ satisfies the
equation
x̃˙ = f (x, u) − f (x̂, u) − H(t)[h(x) − h(x̂)] (11.8)
Expanding the right-hand side in a Taylor series about x̃ = 0 and evaluating the
Jacobian matrices along x̂, the foregoing equation can be written as

x̃˙ = [A(t) − H(t)C(t)]x̃ + Δ(x̃, x, u) (11.9)

where
∂f ∂h
A(t) = (x̂(t), u(t)), C(t) = (x̂(t)) (11.10)
∂x ∂x
and
Δ = f (x, u) − f (x̂, u) − A(t)x̃ − H(t) [h(x) − h(x̂) − C(t)x̃]
11.2. EXTENDED KALMAN FILTER 267

As in Kalman filter design,4 the observer gain H is taken as

H(t) = P (t)C T (t)R−1 (11.11)

where P (t) is the solution of the differential Riccati equation

Ṗ = AP + P AT + Q − P C T R−1 CP, P (t0 ) = P0 (11.12)

and the constant matrices P0 , Q, and R are symmetric and positive definite. We
emphasize that the Riccati equation (11.12) and the observer equation (11.7) have
to be solved simultaneously because A(t) and C(t) depend on x̂(t).

Assumption 11.1 The solution P (t) of (11.12) exists for all t ≥ t0 and satisfies
the inequalities
α1 I ≤ P (t) ≤ α2 I (11.13)
for some positive constants α1 and α2 .

This assumption is crucial for the validity of the Extended Kalman Filter, yet it is
hard to verify.5 Although we know from the properties of the Riccati equation that
the assumption is satisfied if A(t) and C(t) are bounded and the pair (A(t), C(t))
is uniformly observable,6 the matrices A(t) and C(t) are generated in real time so
we cannot check their observability off line.

Lemma 11.2 Under Assumption 11.1, the origin of (11.9) is exponentially stable
and there exist positive constants c, k, and λ such that

x̃(0) ≤ c ⇒ x̃(t) ≤ ke−λ(t−t0 ) , ∀ t ≥ t0 ≥ 0 (11.14)

Proof: Since, by (B.6),



1
∂f ∂f
f (x, u) − f (x̂, u) − A(t)x̃ =
(σ x̃ + x̂, u) − (x̂, u) dσ x̃
≤ 2 L1 x̃
1 2
0 ∂x ∂x


1
∂h ∂h
h(x) − h(x̂) − C(t)x̃ =
(σ x̃ + x̂) − (x̂) dσ x̃
≤ 2 L2 x̃
1 2
0 ∂x ∂x
and
∂h ∂h
C(t) = (x − x̃) ≤ (0)
+ L2 (x + x̃)
∂x ∂x
4 See, for example, [82].
5 See [12, 79] for results on the convergence of the Extended Kalman Filter for special classes
of nonlinear systems.
6 See [82].
268 CHAPTER 11. NONLINEAR OBSERVERS

where L1 and L2 are Lipschitz constants of [∂f /∂x] and [∂h/∂x], respectively, there
exist positive constants k1 and k2 such that Δ(x̃, x, u) ≤ k1 x̃2 + k3 x̃3 . It
follows from Assumption 11.1 that P −1 (t) exists for all t ≥ t0 and satisfies
α3 I ≤ P −1 (t) ≤ α4 I (11.15)
for some positive constants α3 and α4 . Using V = x̃T P −1 x̃ as a Lyapunov function
candidate for (11.9) and noting that dP −1 /dt = −P −1 Ṗ P −1 , we have
d
V̇ = x̃T P −1 x̃˙ + x̃˙ T P −1 x̃ + x̃T P −1 x̃
dt
= x̃T P −1 (A − P C T R−1 C)x̃ + x̃T (AT − C T R−1 CP )P −1 x̃
− x̃T P −1 Ṗ P −1 x̃ + 2x̃T P −1 Δ
= x̃T P −1 (AP + P AT − P C T R−1 CP − Ṗ )P −1 x̃ − x̃T C T R−1 C x̃ + 2x̃T P −1 Δ
Substitution of Ṗ using (11.12) results in
V̇ = −x̃T (P −1 QP −1 + C T R−1 C)x̃ + 2x̃T P −1 Δ
The matrix P −1 QP −1 is positive definite uniformly in t, in view of (11.15), and
the matrix C T R−1 C is positive semidefinite. Hence their sum is positive definite
uniformly in t. Thus, V̇ satisfies the inequality
V̇ ≤ −c1 x̃2 + c2 k1 x̃||3 + c2 k2 x̃||4
for some positive constants c1 and c2 . Consequently,
V̇ ≤ − 12 c1 x̃2 , for x̃ ≤ r
where r is the positive root of − 21 c1 + c2 k1 y + c2 k2 y 2 . The foregoing inequality
shows that the origin is exponentially stable and completes the proof. 2
Example 11.1 Consider the system
ẋ = A1 x + B1 [0.25x21 x2 + 0.2 sin 2t], y = C1 x
where

0 1 0  
A1 = , B1 = , C1 = 1 0
−1 −2 1
T
We start by
investigating boundedness of x(t). Taking V1 (x) = x P1 x, where
3 1
P1 = 12 is the solution of the Lyapunov equation P1 A1 + AT1 P1 = −I, it can
1 1
be shown that
V̇1 = −xT x + 2xT P1 B1 [0.25x21 x2 + 0.2 sin 2t]
≤ −x2 + 0.5P1 B1 x21 x2 + 0.4P1 B1 x
x2 0.4 0.4
= −x2 + √1 x2 + √ x ≤ −0.5x2 + √ x
2 2 2 2
11.3. GLOBAL OBSERVERS 269
√ √ √
for x21 ≤ 2. Noting that minx2 =√2 xT P x = 2/(bT P −1 b) = 2, where b =
1 √ √
col(1, 0),7 we see that Ω = {V1 (x) ≤ 2} ⊂ {x21 ≤ 2}. Inside Ω, we have
0.4 0.4
V̇1 ≤ −0.5x2 + √ x ≤ −0.15x2 , ∀ x ≥ √ = 0.8081
2 0.35 2
√
With λmax (P1 ) = 1.7071, we have (0.8081)2 λmax (P1 ) < 2. Hence the ball {x ≤
0.8081} is in the interior of Ω. This shows that V̇ is negative on the boundary
∂Ω. Therefore, Ω is positively invariant and x(t) is bounded for all x(0) ∈ Ω. We
will now design an Extended Kalman Filter to estimate x(t) for x(0) ∈ Ω. Taking
Q = R = P (0) = I, the Riccati equation (11.12) is given by
Ṗ = AP + P AT + I − P C T CP, P (0) = I
where

0 1  
A(t) = and C = 1 0
−1 + 0.5x̂1 (t)x̂2 (t) −2 + 0.25x̂21 (t)

p p12
Taking P = 11 , it can be shown that the Extended Kalman Filter is defined
p12 p22
by the five simultaneous equations:
x̂˙ 1 = x̂2 + p11 (y − x̂1 )
x̂˙ 2 = −x̂1 − 2x̂2 + 0.25x̂21 x̂2 + 0.2 sin 2t + p12 (y − x̂1 )
ṗ11 = 2p12 + 1 − p211
ṗ12 = p11 (−1 + 0.5x̂1 x̂2 ) + p12 (−2 + 0.25x̂21 ) + p22 − p11 p12
ṗ22 = 2p12 (−1 + 0.5x̂1 x̂2 ) + 2p22 (−2 + 0.25x̂21 ) + 1 − p212
with initial conditions x̂1 (0), x̂2 (0), p11 (0) = 1, p12 (0) = 0, and p22 (0) = 1. Fig-
ure 11.1 shows simulation results for x1 (0) = 1, x2 (0) = −1, x̂1 (0) = x̂2 (0) = 0. We
note that the estimation error converges to zero for a relatively large initial error.


11.3 Global Observers

Consider a nonlinear system in the observer form
ẋ = Ax + ψ(u, y), y = Cx (11.16)
where x ∈ Rn , u ∈ Rm , y ∈ Rp , (A, C) is observable, ψ is locally Lipschitz, and
x(t) and u(t) are defined for all t ≥ 0. The observer for (11.16) is taken as
x̂˙ = Ax̂ + ψ(u, y) + H(y − C x̂) (11.17)
7 See (B.3).
270 CHAPTER 11. NONLINEAR OBSERVERS

(a) (b)
1
x1 1
p11

Components of P (t)
x2
Estimation Error

0.5 0.8
0.6
0 0.4 p22

0.2
−0.5 0
p12
−0.2
−1 −0.4
0 1 2 3 4 0 1 2 3 4
Time Time

Figure 11.1: Simulation results for Example 11.1. Figure (a) shows the estimation
errors x̃1 = x1 − x̂1 and x̃2 = x2 − x̂2 . Figure (b) shows the solution of the Riccati
equation.

The estimation error x̃ = x − x̂ satisfies the linear equation

x̃˙ = (A − HC)x̃

The design of H such that A − HC is Hurwitz ensures that limt→∞ x̃(t) = 0 for all
initial conditions x̃(0). We refer to (11.17) as observer with linear error dynamics.
Consider now the more general system

ẋ = Ax + ψ(u, y) + φ(x, u), y = Cx (11.18)

where (A, C) is observable, ψ and φ are locally Lipschitz, x(t) and u(t) are defined
for all t ≥ 0, and φ(x, u) is globally Lipschitz in x, uniformly in u; that is

φ(x, u) − φ(z, u) ≤ Lx − z

for all x, z, and u. The observer for (11.18) is taken as

x̂˙ = Ax̂ + ψ(u, y) + φ(x̂, u) + H(y − C x̂) (11.19)

where H is designed such that A − HC is Hurwitz. The estimation error satisfies

the equation
x̃˙ = (A − HC)x̃ + φ(x, u) − φ(x̂, u) (11.20)
Using V = x̃T P x̃, where P is the positive definite solution of the Lyapunov equation
P (A − HC) + (A − HC)T P = −I, as a Lyapunov function candidate for (11.20),
we obtain

V̇ = −x̃T x̃ + 2x̃T P [φ(x, u) − φ(x̂, u)] ≤ −x̃2 + 2LP x̃2

11.4. HIGH-GAIN OBSERVERS 271

Hence, the origin of (11.20) is globally exponentially stable if

1
L< (11.21)
2P 
Since the bound on L depends on P , which in turn depends on H, one may seek to
design H to satisfy (11.21) for a given L or to make 1/(2P ) as large as possible.8

11.4 High-Gain Observers

We start our study of high-gain observers with a motivating example.
Example 11.2 Consider the two-dimensional system
ẋ1 = x2 , ẋ2 = φ(x, u), y = x1
where x = col(x1 , x2 ), φ is locally Lipschitz, and x(t) and u(t) are bounded for all
t ≥ 0. We take the observer as
x̂˙ 1 = x̂2 + h1 (y − x̂1 ), x̂˙ 2 = φ0 (x̂, u) + h2 (y − x̂1 )
where φ0 (x, u) is a nominal model φ(x, u). Unlike the observers of the previous
three sections, we do not require φ0 to be the same as φ. We may even take φ0 = 0,
which simplifies the observer to a linear one. Whatever the choice of φ0 is, we
assume that
|φ0 (z, u) − φ(x, u)| ≤ Lx − z + M
for some nonnegative constants L and M , for all (x, z, u) in the domain of interest.
In the special case when φ0 = φ and φ is Lipschitz in x uniformly in u, the foregoing
inequality holds with M = 0. The estimation error x̃ = x − x̂ satisfies the equation

−h1 1 0
x̃˙ = Ao x̃ + Bδ(x, x̃, u), where Ao = , B=
−h2 0 1
and δ(x, x̃, u) = φ(x, u) − φ0 (x̂, u). We view this equation as a perturbation of
the linear system x̃˙ = Ao x̃. In the absence of δ, asymptotic error convergence is
achieved by designing H = col(h1 , h2 ) such that Ao is Hurwitz. In the presence of
δ, we need to design H with the additional goal of rejecting the effect of δ on x̃.
This is ideally achieved, for any δ, if the transfer function from δ to x̃:

1 1
Go (s) = 2
s + h1 s + h2 s + h1
is identically zero. While this is not possible, we can make supω∈R Go (jω) arbi-
trarily small by choosing h2  h1  1. In particular, taking
α1 α2
h1 = , h2 = 2
ε ε
8 Algorithms that seek this goal are given in [108, 109].
272 CHAPTER 11. NONLINEAR OBSERVERS

for some positive constants α1 , α2 , and ε, with ε  1, it can be shown that

ε ε
Go (s) =
(εs)2 + α1 εs + α2 εs + α1

Hence, limε→0 Go (s) = 0. The disturbance rejection property of the high-gain

observer can be seen in the time domain by scaling the estimation error. Let
x̃1
η1 = , η2 = x̃2 (11.22)
ε
Then

−α1 1
εη̇ = F η + εBδ, where F = (11.23)
−α2 0
The matrix F is Hurwitz because α1 and α2 are positive. The matrices Ao and
F/ε are related by the similarity transformation (11.22). Therefore, the eigenvalues
of Ao are 1/ε times the eigenvalues of F . From equation (11.23) and the change
of variables (11.22) we can make some important observations about the behavior
of the estimation error. Using the bound |δ| ≤ Lx̃ + M ≤ Lη + M and the
Lyapunov function candidate V = η T P η, where P is the solution of P F + F T P =
−I, we obtain

εV̇ = −η T η + 2εη T P Bδ ≤ −η2 + 2εLP Bη2 + 2εM P Bη

For εLP B ≤ 14 ,
εV̇ ≤ − 12 η2 + 2εM P Bη
Therefore, by Theorem 4.5, η, and consequently x̃, is ultimately bounded by
εcM for some c > 0, and
! "
η(t) ≤ max ke−at/ε η(0), εcM , ∀t≥0

for some positive constants a and k. Hence η(t) approaches the ultimate bound
exponentially fast, and the smaller ε the faster the rate of decay, which shows
that for sufficiently small ε the estimation error x̃ will be much faster than x. The
ultimate bound can be made arbitrarily small by choosing ε small enough. If M = 0,
which is the case when φ0 = φ, then x̃(t) converges to zero as t tends to infinity.
Notice, however, that whenever x1 (0) = x̂1 (0), η1 (0) = O(1/ε). Consequently, the
solution of (11.23) will contain a term of the form (1/ε)e−at/ε for some a > 0.
While this exponential mode decays rapidly for small ε, it exhibits an impulsive-
like behavior where the transient peaks to O(1/ε) values before it decays rapidly
towards zero. In fact, the function (a/ε)e−at/ε approaches an impulse function as
ε tends to zero. This behavior is known as the peaking phenomenon. It has a
serious impact when the observer is used in feedback control, as we shall see in
Section 12.4. We use numerical simulation to illustrate the foregoing observations.
11.4. HIGH-GAIN OBSERVERS 273
√
We saw in Example 11.1 that for all x(0) ∈ Ω = {1.5x21 + x1 x2 + 0.5x22 ≤ 2}, the
state x(t) of the system

ẋ1 = x2 , ẋ2 = −x1 − 2x2 + ax21 x2 + b sin 2t, y = x1

with a = 0.25 and b = 0.2, is bounded. We use the high-gain observer

2 1
x̂˙ 1 = x̂2 + (y − x̂1 ), x̂˙ 2 = −x̂1 − 2x̂2 + âx̂21 x̂2 + b̂ sin 2t + 2 (y − x̂1 )
ε ε
with two different choice of the pair (â, b̂). When a = 0.25 and b = 0.2 are known,
we take â = 0.25 and b̂ = 0.2. This is a case with no model uncertainty and
φ0 = φ. The other case is when the coefficients a and b are unknown. In this
case we take â = b̂ = 0. Figure 11.2 shows simulation results for both cases.
Figures 11.2(a) and (b) show the state x and its estimate x̂ in the no-uncertainty
case for different values of ε. The estimate x̂2 illustrates the peaking phenomenon.
We note that the peaking phenomenon is not present in x̂1 . While peaking is
induced by the error x1 (0) − x̂1 (0), it does not appear in x̂1 because x̃1 = εη1 .
Figures 11.2(c) and (d) show the estimation error x̃2 for the uncertain model case
when â = b̂ = 0. Comparison of Figures 11.2(b) and (c) shows that the presence of
uncertainty has very little effect on the performance of the observer. Figure 11.2(d)
demonstrates the fact that the ultimate bound on the estimation error is of the
order O(ε). We conclude the example by discussing the effect of measurement
noise on the performance of high-gain observers.9 Suppose the measurement y is
corrupted by bounded measurement noise v; that is, y = x1 + v, and let |v(t)| ≤ N .
Equation (11.23) will take the form

1 α1
εη̇ = F η + εBδ − Ev, where E=
ε α2

and the inequality satisfied by V̇ changes to

2N
εV̇ ≤ − 12 η2 + 2εM P Bη + P Eη
ε
Therefore, the ultimate bound on x̃ takes the form
c2 N
x̃ ≤ c1 M ε + (11.24)
ε
for some positive constants c1 and c2 . This inequality shows a tradeoff between
model uncertainty and measurement noise. An illustration of the ultimate bound
9 Other factors that limit the performance of high-gain observers in digital implementation are

the sampling rate and the computer wordlength. The sampling rate needs to be high enough to
capture the fast changing estimates, and the wordlength should be large enough to represent the
very large (or very small) numbers encountered in the observer. Because of technology advances,
these factors are not as important as the effect of measurement noise.
 274 CHAPTER 11. NONLINEAR OBSERVERS

(a) (b)
1.2 40
ε = 0.01
1
ε = 0.1 30
x1 and Estimates

x2 and Estimates
ε = 0.01
0.8

0.6 20

0.4
10
0.2 ε = 0.1

0 0

0 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Time Time
(c) (d)
0.04
0
Estimation Error of x2

Estimation Error of x2

ε = 0.1 ε = 0.1
0.02
−10
ε = 0.01 0 ε = 0.01
−20

−30 −0.02

−40 −0.04
0 0.1 0.2 0.3 0.4 5 6 7 8 9 10
Time Time

Figure 11.2: Simulation results for Example 11.2. Figures (a) and (b) show the states
x1 and x2 (solid) and their estimates x̂1 and x̂2 (dashed) in the case â = a and b̂ = b.
Figures (c) and (d) show the transient and steady state behavior of x̃2 in the case
â = b̂ = 0.

in Figure 11.3 shows

 that decreasing ε reduces the ultimate bound until we reach
the value ε1 = c2 N/(c1 M ). Reducing ε beyond this point increases the ultimate
bound. For the high-gain observer to be effective, the ratio N/M should be relatively
small so that ε can be chosen to attenuate the effect of uncertainty and make the
observer sufficiently fast. Even if there was no model uncertainty; that is, M = 0,
we still need N to be relatively small so we can design the observer to be sufficiently
fast without bringing the ultimate bound on the estimation error to unacceptable
levels. 
11.4. HIGH-GAIN OBSERVERS 275

ε c1 M + c2N/ε

ε1 ε

Figure 11.3: An illustration of the ultimate bound (11.24).

The high-gain observer will now be developed for a nonlinear system of the form

ẇ = f0 (w, x, u) (11.25)
ẋi = xi+1 + ψi (x1 , . . . , xi , u), for 1 ≤ i ≤ ρ − 1 (11.26)
ẋρ = φ(w, x, u) (11.27)
y = x1 (11.28)

where w ∈ R and x = col(x1 , x2 , . . . , xρ ) ∈ Rρ form the state vector, u ∈ Rm is

the input and y ∈ R is the measured output. It can be seen that the normal form
of Section 8.1 is a special case of this system where ψi = 0 for 1 ≤ i ≤ ρ − 1, f0
is independent of u, and φ(w, x, u) = φ1 (w, x) + φ2 (w, x)u. Moreover, the strict
feedback form (9.17), with gi = 1 for 1 ≤ i ≤ k − 1, is also a special case where
f0 and ψi are independent of u, f0 (w, x) = f01 (w) + f02 (w)x1 , and φ(w, x, u) =
φ1 (w, x) + φ2 (w, x)u. We assume that, over the domain of interest, the functions
f0 , ψ1 , . . . , ψρ−1 and φ are locally Lipschitz in their arguments, and ψ1 , . . . , ψρ−1
are Lipschitz in x uniformly in u; that is,

i
|ψi (x1 , . . . , xi , u) − ψi (z1 , . . . , zi , u)| ≤ Li |xk − zk | (11.29)
k=1

Furthermore, we assume that w(t), x(t), and u(t) are bounded for all t ≥ 0.
A partial state observer that estimates x is taken as
αi
x̂˙ i = x̂i+1 + ψi (x̂1 , . . . , x̂i , u) + (y − x̂1 ), for 1 ≤ i ≤ ρ − 1 (11.30)
εi
αρ
x̂˙ ρ = φ0 (x̂, u) + (y − x̂1 ) (11.31)
ερ
where φ0 is Lipschitz in x uniformly in u, ε is a sufficiently small positive constant,
and α1 to αρ are chosen such that the roots of

sρ + α1 sρ−1 + · · · + αρ−1 s + αρ = 0 (11.32)

276 CHAPTER 11. NONLINEAR OBSERVERS

have negative real parts. The function φ0 in (11.31) is a nominal model of φ. We

assume that
φ(w, x, u) − φ0 (z, u) ≤ L x − z + M (11.33)
Because

φ(w, x, u) − φ0 (z, u) = φ(w, x, u) − φ0 (x, u) + φ0 (x, u) − φ0 (z, u)

and φ0 is Lipschitz, (11.33) requires the modeling error φ(w, x, u) − φ0 (x, u) to be

bounded.

Lemma 11.3 Under the stated assumptions, there is ε∗ > 0 such that for 0 < ε ≤
ε∗ , the estimation errors x̃i = xi − x̂i , for 1 ≤ i ≤ ρ, of the high-gain observer
(11.30)–(11.31) satisfy the bound
 
b −at/ε
|x̃i | ≤ max e , ερ+1−i
cM (11.34)
εi−1

for some positive constants a, b, c. 3

Proof: Define the scaled estimation errors

x1 − x̂1 x2 − x̂2 xρ−1 − x̂ρ−1
η1 = ρ−1
, η2 = ρ−2
, . . . , ηρ−1 = , ηρ = xρ − x̂ρ (11.35)
ε ε ε
It can be shown that η = col(η1 , η2 , · · · , ηρ ) satisfies the equation

εη̇ = F η + εδ(w, x, x̃, u) (11.36)

where δ = col(δ1 , δ2 , · · · , δρ ),
⎡ ⎤
−α1 1 0 ··· 0
⎢ −α2 0 1 ··· 0⎥
⎢ ⎥
⎢ .. ⎥ ,
F = ⎢ ... ..
. .⎥ δρ = φ(w, x, u) − φ0 (x̂, u)
⎢ ⎥
⎣−αρ−1 0 1⎦
−αρ 0 ··· ··· 0

and
1
δi = [ψi (x1 , . . . , xi , u) − ψi (x̂1 , . . . , x̂i , u)], for 1≤i≤ρ−1
ερ−i
The matrix F is Hurwitz by design because its characteristic equation is (11.32).
Using (11.29) we see that δ1 to δρ−1 satisfy

Li  Li  ρ−k 
i i i
|δi | ≤ ρ−i |xk − x̂k | = ρ−i ε |ηk | = Li εi−k |ηk |
ε ε
k=1 k=1 k=1
11.5. EXERCISES 277

The preceding inequality and (11.33) show that

δ ≤ Lδ η + M (11.37)

where Lδ is independent of ε for all ε ≤ ε∗ for any given ε∗ > 0. Let V = η T P η,

where P = P T > 0 is the solution of the Lyapunov equation P F + F T P = −I.
Then
εV̇ = −η T η + 2εη T P δ
Using (11.37) we obtain

εV̇ ≤ −η2 + 2εP Lδ η2 + 2εP M η

For P Lδ ≤ 14 ,

εV̇ ≤ − 12 η2 + 2εP M η ≤ − 14 η2 , ∀ η ≥ 8εP M

We conclude by Theorem 4.5 that

! "
η(t) ≤ max ke−at/ε η(0), εcM , ∀t≥0

for some positive constants a, c, and k. From (11.35) we see that η(0) ≤ β/ερ−1 ,
for some β > 0, and |x̃i | ≤ ερ−i |ηi |, which yields (11.34). 2

11.5 Exercises
11.1 Consider the van der Pol equation

ẋ1 = x2 , ẋ2 = −x1 + (1 − x21 )x2 , y = x1

(a) Design an Extended Kalman Filter.

(b) Design a high-gain observer with ε = 0.01.
(c) Compare the performance of the two observers using simulation.

11.2 A normalized state model of the Duffing equation (A.4) is given by

ẋ1 = x2 , ẋ2 = −x1 (1 + x21 ) − x2 + a cos(t), y = x1

(a) Design an observer with linear error dynamics when a = 1.

(b) Design a high-gain observer with ε = 0.01.
(c) Using simulation, compare the performance of the two observers when a is
perturbed. (Keep a = 1 in the observer but gradually increase it in the
system’s equation.)
278 CHAPTER 11. NONLINEAR OBSERVERS

11.3 The Wien-Bridge oscillator of Exercise 2.5, with C1 = C2 = C and R1 =

R2 = R, is modeled in the τ = t/(CR) time scale by the equations

ẋ1 = − 12 x2 , ẋ2 = 2(x1 − x2 ) + g(x2 ) − x2

where g(v) = 3.234v − 2.195v 3 + 0.666v 5 . Suppose the measured output is x2 .

(a) Design an Extended Kalman Filter.
(b) Design an observer with linear error dynamics.
(c) Design a high-gain observer with ε = 0.01. Hint: Transform the system into
the form (11.26)–(11.28).
(d) Using simulation, compare the performance of the three observers.
(e) Repeat (d) when g in the system’s equation is perturbed to 1.2g.

11.4 Consider the electrostatic microactuator

√ (A.33) with ζ = 0.1 and T = 0.2.
Under the constant input u = 5/(4 2),√ the system has an asymptotically stable
equilibrium point at x̄ = col(1/6, 0, 1/ 2). Assume that the measured outputs
are x1 and x3 . The state x2 can be estimated by one of the following high-gain
observers:

x̂˙ 1 = x̂2 + (α1 /ε)(x1 − x̂1 ), x̂˙ 2 = −x̂1 + 13 x23 + (α2 /ε2 )(x1 − x̂1 )

or
x̂˙ 1 = x̂2 + (α1 /ε)(x1 − x̂1 ), x̂˙ 2 = −x̂1 + (α2 /ε2 )(x1 − x̂1 )
where in the first observer 13 x23 is treated as a given input, while in the second one
it is treated as disturbance. Use simulation to compare the performance of the two
observers for ε = 1 and 0.1. Let α1 = 2 and α2 = 1. Limit x(0) to the region of
attraction of x̄.

11.5 Consider the system

ẋ1 = β1 (x1 ) + x2 , ẋ2 = β2 (x1 ) + β3 (x1 )x2 + u, y = x1

where β1 to β3 are smooth functions.

(a) Design an observer with linear error dynamics.
(b) Design a high-gain observer.
(c) Repeat part (b) when the function β2 and β3 are unknown.

11.6 Consider the single link manipulator with flexible joint from Example 8.15.
(a) Design an observer with linear error dynamics when y = x1 .
11.5. EXERCISES 279

(b) Design an observer with linear error dynamics when y = col(x1 , x3 ).

(c) Design a high-gain observer when y = x1 .
(d) Design a high-gain observer when y = col(x1 , x3 ) and the parameters a, b, c, d
are unknown.

11.7 Consider the high-gain observer (11.30)–(11.31) when y(t) = x1 (t)+v(t) with
|v(t)| ≤ N . Show that the bound (11.34) changes to
 
b −at/ε c2 N
|x̃i | ≤ max e , ερ+1−i
cM + i−1
εi−1 ε

for some c2 > 0.

11.8 A modified Extended Kalman Filter in which the estimation error decays to
zero faster than an exponential ke−αt with a prescribed positive constant α [110]
can be obtained by changing the Riccati equation (11.12) to

Ṗ = (A + αI)P + P (AT + αI) + Q − P C T R−1 CP, P (t0 ) = P0

Redesign the Extended Kalman Filter of Example 11.1 with this modification and
use simulation to examine the effect of α on the speed of convergence.

11.9 Consider the system

ẋ1 = x2 , ẋ2 = x1 + a sin x1 − x2 , y = x2

What is the largest |a| for which you can design a global observer?

Nonlinear observers can be used as a tool to estimate unknown parameters by

treating them as state variables in an extended state model. The next three exercises
use this tool.

11.10 Consider the harmonic oscillator

ẋ1 = x2 , ẋ2 = −ax1 , y = x1

with unknown positive constant a. Extend the state model with x3 = a and ẋ3 = 0.
Design an Extended Kalman Filter with Q = R = P0 = I and simulate it with
a = 1, x(0) = col(1, 0), and x̂(0) = col(0, 0, 0).

11.11 Consider the pendulum equation

ẋ1 = x2 , ẋ2 = − sin x1 + cu, y = x1

with unknown positive constant c and known constant input u. Extend the state
model with x3 = cu and ẋ3 = 0.
280 CHAPTER 11. NONLINEAR OBSERVERS

(a) Design an observer with linear error dynamics having eigenvalues −1, −1, −1.

(b) Design a high-gain observer with ε = 0.1.

(c) Design an Extended Kalman Filter with Q = R = P0 = I.

(d) Compare the performance of the three observers using simulation with u = 12 ,
c = 1, x(0) = col(π/3, 0), and x̂(0) = col(0, 0, 0).

11.12 Consider the boost converter of Section A.5 and suppose that, due to
 the resistance R is unknown. Changing the definition of x1 to
unknown load,
x1 = (iL /E) L/C, the state model (A.16) is modified to

ẋ1 = (x2 + k)u − (1/k)x2 , ẋ2 = (1/k)x1 − αx2 − x1 u − αk


where α = R1 L/C is unknown. It is required to design an observer to estimate α.
Extend the state model with x3 = α and ẋ3 = 0, and take u = 0.

(a) Assuming that y = x1 , design an Extended Kalman Filter. Simulate the re-
sponse of the observer with α = 0.3 and k = 2.

(b) Repeat (a) when y = x2 . Does x̂3 converge to α? If not, explain why.
Chapter 12

Output Feedback Stabilization

The feedback stabilization techniques we studied in Chapters 9 and 10 require mea-

surement of all state variables. In this chapter we study feedback stabilization when
we can only measure an output vector, which could be a nonlinear function of the
state and whose dimension is typically lower than the state’s dimension.
We start in Section 12.1 by linearization, which is the simplest method to de-
sign output feedback control, but of course is guaranteed to work only locally. In
Section 12.2 we present a passivity-based controller when the map from the input
to the derivative of the output is passive. The technique is useful in many physical
systems, and this is illustrated by application to a robot manipulator. The next
two sections deal with observer-based control, where the stabilization problem is
separated in two steps. First a state feedback controller is designed to stabilize the
origin; then an observer is designed to estimate the state. The output feedback
controller results from replacing the state by its estimate. We do not discuss here
how the state feedback controller is designed, but this could be done by one of the
techniques of Chapters 9 and 10. In Section 12.3 we show how state feedback stabi-
lization can be recovered by the local or global observers of Sections 11.1 and 11.3.
As we saw in Chapter 11, those observers require perfect knowledge of the nonlinear
state model and may tolerate limited model uncertainty. The high-gain observer
of Section 11.4 can accommodate large matched uncertainty. Its use in feedback
control is shown in Section 12.4. We introduce the idea of saturating the control as
a tool for overcoming the peaking phenomenon of the observer. Then we show that
the combination of saturation with fast convergence of the estimation error enables
us to prove that the trajectories under output feedback can be made arbitrarily
close to those under state feedback by making the observer fast enough. Finally,
in Section 12.5 we present robust output feedback stabilization of minimum phase
systems, extending the robust state feedback stabilization techniques of Chapter 10.
Our presentation uses sliding mode control but the same procedures could be used
with Lyapunov redesign or high-gain feedback.

281
282 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

12.1 Linearization
Consider the system
ẋ = f (x, u), y = h(x) (12.1)
where x ∈ Rn , u ∈ Rm , y ∈ Rp , f (0, 0) = 0, h(0) = 0, and f (x, u), h(x) are
continuously differentiable in a domain Dx × Du ⊂ Rn × Rm that contains the
origin (x = 0, u = 0). We want to design an output feedback controller to stabilize
the system at x = 0. Linearization of (12.1) about (x = 0, u = 0) results in the
linear system
ẋ = Ax + Bu, y = Cx (12.2)
where
  
∂f  ∂f  ∂h 
A= (x, u) , B= (x, u) , C= (x)
∂x x=0,u=0 ∂u x=0,u=0 ∂x x=0

Assume (A, B) is stabilizable and (A, C) is detectable, and design a linear dynamic
output feedback controller

ż = F z + Gy, u = Lz + M y (12.3)

such that the closed-loop matrix

A + BM C BL
(12.4)
GC F

is Hurwitz. An example of such design is the observer-based controller

x̂˙ = Ax̂ + Bu + H(y − C x̂), u = −K x̂

which takes the form (12.3) with

z = x̂, F = A − BK − HC, G = H, L = −K, M = 0

In this case the matrix (12.4) is Hurwitz when K and H are designed such that
A − BK and A − HC are Hurwitz. Another example is the static output feedback
controller u = M y, where M is designed such that A + BM C is Hurwitz. It takes
the form (12.3) with L = 0 after dropping out the ż-equation. When the controller
(12.3) is applied to the nonlinear system (12.1) it results in the closed-loop system

ẋ = f (x, Lz + M h(x)), ż = F z + Gh(x) (12.5)

It can be verified that the origin (x = 0, z = 0) is an equilibrium point of the

closed-loop system (12.5) and linearization at the origin results in the Hurwitz
matrix (12.4). Thus, we conclude by Theorem 3.2 that the origin is exponentially
stable.
12.2. PASSIVITY-BASED CONTROL 283

Example 12.1 Reconsider the pendulum equation of Example 9.2, and suppose
we measure the angle θ, but not the angular velocity θ̇. The output is taken as y =
x1 = θ − δ1 , and the state feedback controller of Example 9.2 can be implemented
using the observer
x̂˙ = Ax̂ + Buδ + H(y − x̂1 )
where H = col(h1 , h2 ). It can be verified that A − HC will be Hurwitz if

h1 + b > 0, h1 b + h2 + cos δ1 > 0

The output feedback controller is u = (1/c) sin δ1 − K x̂ 

12.2 Passivity-Based Control

In Section 9.6 we saw that the m-input–m-output system

ẋ = f (x, u), y = h(x) (12.6)

where f (0, 0) = 0 and h(0) = 0, can be globally stabilized by the output feedback
controller u = −φ(y), where φ(0) = 0 and y T φ(y) > 0 for all y = 0, provided the
system is passive with a radially unbounded positive definite storage function and
zero-state observable. In this section we extend this result to the case when the
preceding two properties hold for a map from the input, u, to the derivative of the
output, ẏ.
Consider the system (12.6) where f (0, 0) = 0, h(0) = 0, f is locally Lipschitz in
(x, u), and h is continuously differentiable for all x ∈ Rn . Consider the auxiliary
system
∂h def
ẋ = f (x, u), ẏ = f (x, u) = h̃(x, u) (12.7)
∂x
where ẏ is the output. Suppose the auxiliary system is passive with a radially un-
bounded positive definite storage function V (x) and zero state observable. The idea
of designing a stabilizing output feedback controller for the system (12.6) is illus-
trated in Figure 12.1 for the single-input–single-output case. The block diagram in
Figure 12.1(a) shows that the controller for the plant (12.6) is constructed by pass-
ing the output y through the first order transfer function s/(τ s + 1) with a positive
constant τ . The output of the transfer function, z, drives a passive nonlinearity
φ, which provides the feedback signal that closes the loop. Figure 12.1(b) shows
an equivalent representation where the system in the forward path is the auxiliary
system (12.7) while the transfer function in the feedback path is 1/(τ s + 1). The
system now is a feedback connection of two passive systems, where the system in
the forward path is passive by assumption and its storage function is V (x). The sys-
tem in the feedback path was shown in Example 5.3 to be passive with the storage
z
function τ 0 φ(σ) dσ. The sum of the two storage functions is a storage function
for the feedback connection and will be used as a Lyapunov function candidate to
284 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

+ u y + u y y
.
Plant Plant s
− −

z s z
φ(.) φ(.)
1
τs+1 τs+1

(a) (b)

Figure 12.1: Passivity-based control for the system (12.6).

show asymptotic stability. The transfer function s/(τ s + 1) can be realized by the
state model
τ ẇ = −w + y, z = (−w + y)/τ
whose output satisfies τ ż = −z + ẏ.
Lemma 12.1 Consider the system (12.6) and the output feedback controller
ui = −φi (zi ), τi ẇi = −wi + yi , zi = (−wi + yi )/τi , for 1 ≤ i ≤ m (12.8)
where τi > 0, φi is locally Lipschitz, φi (0) = 0, and zi φi (zi ) > 0 for all zi = 0.
Suppose the auxiliary system (12.7) is
• passive with a positive definite storage function V (x):
∂V
uT ẏ ≥ V̇ = f (x, u), ∀ (x, u)
∂x
• zero-state observable: with u = 0, ẏ(t) ≡ 0 ⇒ x(t) ≡ 0.
Then the origin of the closed-loop system is asymptotically
z stable. It is globally
asymptotically stable if V (x) is radially unbounded and 0 i φi (σ) dσ → ∞ as |zi | →
∞. 3
Proof: Using x and z1 , . . . , zm as the state variables, the closed-loop system is
represented by
ẋ = f (x, −φ(z))
τi żi = −zi + ẏi = −zi + h̃i (x, −φ(z)), for 1 ≤ i ≤ m
where z = col(z1 , . . . , zm ) and φ = col(φ1 , . . . , φm ). Let
 m  zi
W (x, z) = V (x) + τi φi (σ) dσ
i=1 0

The function W is positive definite. Its derivative is


m 
m 
m
Ẇ = V̇ + τi φi (zi )żi ≤ u ẏ −
T
zi φi (zi ) − u ẏ = −
T
zi φi (zi )
i=1 i=1 i=1
12.2. PASSIVITY-BASED CONTROL 285

Hence, Ẇ is negative semidefinite. Moreover,

Ẇ ≡ 0 ⇒ z(t) ≡ 0 ⇒ u(t) ≡ 0 and ẏ(t) ≡ 0

It follows from zero state observability of the auxiliary system (12.7) that x(t) ≡ 0.
By the invariance principle we conclude
z that the origin is asymptotically stable. If
V (x) is radially unbounded and 0 i φi (σ) dσ → ∞ as |zi | → ∞, the function W
will be radially unbounded and the origin will be globally asymptotically stable. 2
Similar to our discussion in Section 9.6, if the system (12.7) is not passive we
may be able to turn it into a passive one by feedback passivation. Here, however,
passivation will have to be via output feedback.

Example 12.2 Reconsider the regulation problem of an m-link robot from Exam-
ple 9.15 when we can measure q but not q̇. The regulation error e = q − qr satisfies
the equation
M (q)ë + C(q, q̇)ė + Dė + g(q) = u
with the output y = e. The regulation task is achieved by stabilizing the system at
(e = 0, ė = 0). With
u = g(q) − Kp e + v
where Kp is a positive definite symmetric matrix, the system is given by

M (q)ë + C(q, q̇)ė + Dė + Kp e = v, y=e

which has an equilibrium point at (e = 0, ė = 0) when v = 0. Taking the energy

V = 12 ėT M (q)ė + 12 eT Kp e

as the storage function, we have

V̇ = ėT M ë + 12 ėT Ṁ ė + eT Kp ė
= 1 T
2 ė (Ṁ − 2C)ė − ėT Dė − ėT Kp e + ėT v + eT Kp ė ≤ ėT v

Hence, the auxiliary system from v to ė is passive. It is also zero-state observable

because, with v = 0,

ė(t) ≡ 0 ⇒ ë(t) ≡ 0 ⇒ Kp e(t) ≡ 0 ⇒ e(t) ≡ 0

According to Lemma 12.1, the system can be globally stabilized by the control
v = −Kd z where Kd is a positive diagonal matrix and zi is the output of the linear
system
τi ẇi = −wi + ei , zi = (−wi + ei )/τi , for 1 ≤ i ≤ m
The overall control u is given by

u = g(q) − Kp (q − qr ) − Kd z
286 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

Comparing this control with the state feedback control

u = g(q) − Kp (q − qr ) − Kd q̇

of Example 9.15 shows that the derivative q̇ of the PD controller is replaced by

z, which is obtained by passing ei through the transfer function s/(τi s + 1). We
emphasize that zi is not necessarily an approximation of q̇i because we do not require
the time constant τi to be small. 

12.3 Observer-Based Control

Consider the system
ẋ = f (x, u), y = h(x) (12.9)
where x ∈ Rn , u ∈ Rm , and y ∈ Rp are the state, control input, and measured
output, respectively, f (0, 0) = 0, h(0) = 0, and f (x, u), h(x) are locally Lipschitz.
Let u = γ(x) be a locally Lipschitz state feedback controller that stabilizes the
origin of the closed-loop system ẋ = f (x, γ(x)). The observer for (12.9) is taken as

x̂˙ = f (x̂, u) + H[y − h(x̂)] (12.10)

The estimation error x̃ = x − x̂ satisfies the equation

def
x̃˙ = f (x, u) − f (x̂, u) − H[h(x) − h(x̂)] = g(x, x̃) (12.11)

The equation (12.11) has an equilibrium point at x̃ = 0. We assume that the

observer gain H is designed such that x̃ = 0 is exponentially stable and there is a
Lyapunov function V1 (x̃) that satisfies the inequalities

∂V1 ∂V1
c1 x̃ ≤ V1 (x̃) ≤ c2 x̃ ,
2 2
g(x, x̃) ≤ −c3 x̃ ,
2 ≤ c4 x̃ (12.12)
∂ x̃ ∂ x̃

for all x ∈ D1 and x̃ ∈ D2 for some positive constants c1 to c4 , where D1 ⊂ Rn

and D2 ⊂ Rn are domains that contain the origin. We have seen in Sections 11.1
and 11.3 observers for which (12.12) is satisfied locally or globally. The closed-loop
system under output feedback is given by

ẋ = f (x, γ(x − x̃)), x̃˙ = g(x, x̃) (12.13)

The task now is to show that the origin (x = 0, x̃ = 0) is an asymptotically stable

equilibrium point of (12.13).

Theorem 12.1 Consider the system (12.13) and suppose that (12.12) is satisfied.
• If the origin of ẋ = f (x, γ(x)) is asymptotically stable, then the origin of
(12.13) is asymptotically stable.
12.3. OBSERVER-BASED CONTROL 287

• If the origin of ẋ = f (x, γ(x)) is exponentially stable, then the origin of (12.13)
is exponentially stable.

• If (12.12) holds globally and the system ẋ = f (x, γ(x − x̃)), with input x̃,
is input-to-state stable, then the origin of (12.13) is globally asymptotically
stable.

The first two bullets give local results. The third bullet is a global result that re-
quires the stronger condition of input-to-state stability. The proof of the theorem is
very similar to the Lyapunov analysis of cascade systems in Section C.1. Although
the system (12.13) is not a cascade connection, the origin of x̃˙ = g(x, x̃) is expo-
nentially stable uniformly in x.

Proof: If the origin of ẋ = f (x, γ(x)) is asymptotically stable, then by (the con-
verse Lyapunov) Theorem 3.9, there are positive definite functions V0 (x) and W0 (x)
such that
∂V0
f (x, γ(x)) ≤ −W0 (x)
∂x
in some neighborhood of x = 0. On a bounded neighborhood of the origin, we can
use the local Lipschitz property of f and γ to obtain

∂V0
[f (x, γ(x)) − f (x, γ(x − x̃))] ≤ L x̃
∂x

for some positive constant L. Using V (x, x̃) = bV0 (x) + V1 (x̃), with b > 0, as a
Lyapunov function candidate for (12.13), we obtain

∂V0 1 ∂V1
V̇ = b f (x, γ(x − x̃)) + √ g(x, x̃)
∂x 2 V1 ∂ x̃
∂V0 ∂V0 1 ∂V1
= b f (x, γ(x)) + b [f (x, γ(x − x̃)) − f (x, γ(x))] + √ g(x, x̃)
∂x ∂x 2 V1 ∂ x̃
c3 x̃ 2
≤ −bW0 (x) + bL x̃ − √
2 V1
Since
−1 −1
V1 ≤ c2 x̃ 2 ⇒ √ ≤√
V1 c2 x̃

we have
c3
V̇ ≤ −bW0 (x) + bL x̃ − √ x̃
2 c2
√
Choosing b < c3 /(2L c2 ) ensure that V̇ is negative definite, which completes the
proof of the first bullet.
288 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

If the origin of ẋ = f (x, γ(x)) is exponentially stable, then by (the converse Lya-
punov) Theorem 3.8, there is a Lyapunov function V0 (x) that satisfies the inequal-
ities

∂V0 ∂V0
a1 x ≤ V0 (x) ≤ a2 x ,
2 2
f (x, γ(x)) ≤ −a3 x ,
2 ≤ a4 x
∂x ∂x

in some neighborhood of the origin, for some positive constants a1 to a4 . Using

V (x, x̃) = bV0 (x) + V1 (x̃), with b > 0, as a Lyapunov function candidate for (12.13),
we obtain
∂V0 ∂V0 ∂V1
V̇ = b f (x, γ(x)) + b [f (x, γ(x − x̃)) − f (x, γ(x))] + g(x, x̃)
∂x ∂x ∂ x̃
≤ −ba3 x 2 + ba4 L1 x x̃ − c3 x̃ 2

T

x ba3 −ba4 L1 /2 x
= −
x̃ −ba4 L1 /2 c3 x̃

where L1 is a Lipschitz constant of f (x, γ(x − x̃)) with respect to x̃. Choosing
b < 4a3 c3 /(a4 L1 )2 ensures that the matrix in the foregoing inequality is positive
definite, which completes the proof of the second bullet. The proof of the third
bullet follows from the proof of Lemma 4.6.1 2
If the Lyapunov functions V0 (x) for the closed-loop system under state feedback
and V1√ (x̃) for the observer are known, the composite Lyapunov functions V =
bV0 + V1 , for the first bullet, and V = bV0 + V1 , for the second bullet, can be
used to estimate the region of attraction of the closed-loop system under output
feedback.

12.4 High-Gain Observers and the Separation Principle

As we did in Section 11.4, we start our study of high-gain observers in feedback
control with a motivating example.
Example 12.3 Consider the two-dimensional system

ẋ1 = x2 , ẋ2 = φ(x, u), y = x1 (12.14)

where x = col(x1 , x2 ). Suppose u = γ(x) is a locally Lipschitz state feedback

controller that stabilizes the origin of the closed-loop system

ẋ1 = x2 , ẋ2 = φ(x, γ(x)) (12.15)

To implement this control with output feedback we use the high-gain observer

x̂˙ 1 = x̂2 + (α1 /ε)(y − x̂1 ), x̂˙ 2 = φ0 (x̂, u) + (α2 /ε2 )(y − x̂1 ) (12.16)
1 See [130].
12.4. HIGH-GAIN OBSERVERS AND THE SEPARATION PRINCIPLE 289

where φ0 is a nominal model of φ, and α1 , α2 , and ε are positive constants with

ε  1. We saw in Lemma 11.3 that if
|φ0 (z, u) − φ(x, u)| ≤ L x − z + M
over the domain of interest, then for sufficiently small ε, the estimation errors
x̃1 = x1 − x̂1 and x̃2 = x2 − x̂2 satisfy the inequalities
! "  
−at/ε b −at/ε
|x̃1 | ≤ max be 2
, ε cM , |x̃2 | ≤ e , εcM (12.17)
ε
for some positive constants a, b, c. These inequalities show that reducing ε di-
minishes the effect of model uncertainty. They show also that, for small ε, the
estimation error will be much faster than x. The bound on x̃2 demonstrates the
peaking phenomenon that was described in Section 11.4; namely, x̃2 might peak
to O(1/ε) values before it decays rapidly towards zero. The peaking phenomenon
might destabilize the closed-loop system. This fact is illustrated by simulating the
system
ẋ1 = x2 , ẋ2 = x32 + u, y = x1
which can be globally stabilized by the state feedback controller
u = −x32 − x1 − x2
The output feedback controller is taken as
u = −x̂32 − x̂1 − x̂2 , x̂˙ 1 = x̂2 + (2/ε)(y − x̂1 ), x̂˙ 2 = (1/ε2 )(y − x̂1 )
where we took φ0 = 0. Figure 12.2 shows the performance of the closed-loop system
under state and output feedback. Output feedback is simulated for three different
values of ε. The initial conditions are x1 (0) = 0.1, x2 (0) = x̂1 (0) = x̂2 (0) =
0. Peaking is induced by [x1 (0) − x̂1 (0)]/ε = 0.1/ε when ε is sufficiently small.
Figure 12.2 shows a counter intuitive behavior as ε decreases. Since decreasing
ε causes the estimation error to decay faster toward zero, one would expect the
response under output feedback to approach the response under state feedback as ε
decreases. Figure 12.2 shows the opposite behavior where the response under output
feedback deviates from the response under state feedback as ε decreases. This is
the impact of the peaking phenomenon. The same figure shows the control u on a
much shorter time interval to exhibit peaking. This control peaking is transmitted
to the plant causing its state to peak. Figure 12.3 shows that as we decrease ε to
0.004, the system has a finite escape time shortly after t = 0.07.
Fortunately, we can overcome the peaking phenomenon by saturating the control
outside a compact set of interest in order to create a buffer that protects the plant
from peaking. Writing the closed-loop system under state feedback as ẋ = Ax and
solving the Lyapunov equation P A + AT P = −I, we have

0 1 1.5 0.5
A= and P =
−1 −1 0.5 1
290 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

0.5
0 SFB
−0.5 OFB ε = 0.1
−1 OFB ε = 0.01
x1

−1.5 OFB ε = 0.005

−2
0 1 2 3 4 5 6 7 8 9 10
1
0
−1
x2

−2
−3
0 1 2 3 4 5 6 7 8 9 10
0
−100
−200
u

−300
−400
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
t

Figure 12.2: Performance under state (SFB) and output (OFB) feedback.

Then, V (x) = xT P x is a Lyapunov function for ẋ = Ax and V̇ (x) = −xT x. Suppose

all initial conditions of interest belong to the set Ω = {V (x) ≤ 0.3}. Because Ω is
positively invariant, x(t) ∈ Ω for all t ≥ 0. By maximizing |x1 + x2 | and |x2 | over
Ω, it can be shown by using (B.4) that for all x ∈ Ω, |x1 + x2 | ≤ 0.6 and |x2 | ≤ 0.6.
Hence, |u| ≤ |x2 |3 + |x1 + x2 | ≤ 0.816. Saturating u at ±1 results in the globally
bounded state feedback control
u = sat(−x32 − x1 − x2 )
For all x(0) ∈ Ω, the saturated control produces the same trajectories as the unsat-
urated control because for x ∈ Ω, |u| < 1 and the saturation will never be effective.
In output feedback, the state x in the foregoing saturated control is replaced by its
estimate x̂; namely,
u = sat(−x̂32 − x̂1 − x̂2 )
During the peaking period, the control saturates. Figure 12.4 shows the perfor-
mance of the closed-loop system under saturated state and output feedback. The
control u is shown on a shorter time interval that exhibits control saturation during
peaking. The peaking period decreases with ε. The states x1 and x2 exhibit the
intuitive behavior we expected earlier; namely, the response under output feedback
approaches the response under state feedback as ε decreases. Note that we decrease
ε to 0.001, beyond the value 0.004 where instability was detected in the unsatu-
rated case. Not only does the system remain stable, but the response under output
feedback is almost indistinguishable from the response under state feedback. What
12.4. HIGH-GAIN OBSERVERS AND THE SEPARATION PRINCIPLE 291

0.2
0
−0.2
x1

−0.4
−0.6
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
0

−200
x2

−400

−600
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
2000

1000
u

−1000
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
t

Figure 12.3: Instability induced by peaking at ε = 0.004.

is also interesting is that the region of attraction under output feedback approaches
the region of attraction under saturated state feedback as ε tends to zero. This is
shown in Figures 12.5 and 12.6. The first figure shows the phase portrait of the
closed-loop system under u = sat(−x32 − x1 − x2 ). It has a bounded region of at-
traction enclosed by a limit cycle. The second figure shows that the intersection of
the boundary of the region of attraction under u = sat(−x̂32 − x̂1 − x̂2 ) with the
x1 –x2 plane approaches the limit cycle as ε tends to zero.
The behavior we saw in Figures 12.4 and 12.6 will be realized with any globally
bounded stabilizing function γ(x). During the peaking period, the control γ(x̂) sat-
urates. Since the peaking period shrinks to zero as ε tends to zero, for sufficiently
small ε the peaking period becomes so small that the state of the plant x remains
close to its initial value. After the peaking period, the estimation error becomes
O(ε) and the feedback control γ(x̂) becomes close to γ(x). Consequently, the tra-
jectories of the closed-loop system under output feedback asymptotically approach
its trajectories under state feedback as ε tends to zero. This leads to recovery of the
performance achieved under state feedback. The global boundedness of γ(x) can
be always achieved by saturating the state feedback control, or the state estimates,
outside a compact set of interest.
The analysis of the closed-loop system under output feedback starts by repre-
senting the system in the form

ẋ1 = x2 , ẋ2 = φ(x, γ(x − Dη)) (12.18)

292 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

SFB
0.15 OFB ε = 0.1
0.1 OFB ε = 0.01
OFB ε = 0.001
0.05
x1

0
−0.05
0 1 2 3 4 5 6 7 8 9 10
0.05
0
−0.05
x2

−0.1

0 1 2 3 4 5 6 7 8 9 10

−0.5
u

−1
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
t

Figure 12.4: Performance under state (SFB) and output (OFB) feedback with satura-
tion.

εη̇1 = −α1 η1 + η2 , εη̇2 = −α2 η1 + εδ(x, x̃) (12.19)

ε 0
where D = and η = D−1 x̃. For sufficiently small ε, η will be much faster
0 1
than x. Setting ε = 0 on the right-hand side of (12.19) shows that the motion of
the fast variable η can be approximated by the model

−α1 1 def
εη̇ = η = Fη (12.20)
−α2 0

whose solution converges to zero rapidly. When η = 0, equation (12.18) reduces to

ẋ1 = x2 , ẋ2 = φ(x, γ(x))

which is the closed-loop system under state feedback (12.15). Let V (x) be a Lya-
punov function for (12.15) and W (η) = η T P0 η, where P0 is the solution of the
Lyapunov equation P0 F + F T P0 = −I. We will use V and W to analyze the
stability of the closed-loop system (12.18)–(12.19). Define the sets Ωc and Σ by
Ωc = {V (x) ≤ c} and Σ = {W (η) ≤ ρ ε2 }, where c > 0 is chosen such that Ωc is
in the interior of the region of attraction of (12.15). The analysis can be divided in
two steps. In the first step we show that for sufficiently large ρ, there is ε∗1 > 0 such
12.4. HIGH-GAIN OBSERVERS AND THE SEPARATION PRINCIPLE 293

2 1

1 0.5

x2
0
x2

−0.5
−1

−1
−2 −2 −1 0 1 2
−3 −2 −1 0 1 2 3
x1 x1

Figure 12.5: Phase portrait of Figure 12.6: Output feedback with

the closed-loop system under u = ε = 0.08 (dashed) and ε = 0.01 (dash-
sat(−x32 − x1 − x2 ). dot).

that, for every 0 < ε ≤ ε∗1 , the origin of the closed-loop system is asymptotically
stable and the set Ωc × Σ is a positively invariant subset of the region of attraction.
The proof makes use of the fact that in Ωc × Σ, η is O(ε). In the second step we
show that for any bounded x̂(0) and any x(0) ∈ Ωb , where 0 < b < c, there exists
ε∗2 > 0 such that, for every 0 < ε ≤ ε∗2 , the trajectory enters the set Ωc × Σ in
finite time. The proof makes use of the fact that Ωb is in the interior of Ωc and
γ(x̂) is globally bounded. Hence, there exits T1 > 0, independent of ε, such that
any trajectory starting in Ωb will remain in Ωc for all t ∈ [0, T1 ]. Using the fact
that η decays faster than an exponential function of the form (k/ε)e−at/ε , we can
show that the trajectory enters the set Ωc × Σ within a time interval [0, T (ε)], where
limε→0 T (ε) = 0. Thus, by choosing ε small enough, we can ensure that T (ε) < T1 .

We turn now to the general case. Consider the system

ẇ = ψ(w, x, u) (12.21)
ẋi = xi+1 + ψi (x1 , . . . , xi , u), 1≤i≤ρ−1 (12.22)
ẋρ = φ(w, x, u) (12.23)
y = x1 (12.24)
z = q(w, x) (12.25)

where u ∈ Rm is the control input, y ∈ R and z ∈ Rs are measured outputs, and

w ∈ R and x ∈ Rρ constitute the state vector. The functions ψ, φ, and q are locally
Lipschitz in their arguments for (w, x, u) ∈ Dw × Dx × Rm , where Dw ⊂ Rs and
Dx ⊂ Rρ are domains that contain their respective origins. Moreover, φ(0, 0, 0) = 0,
ψ(0, 0, 0) = 0, and q(0, 0) = 0. Our goal is to design an output feedback controller
to stabilize the origin.
294 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

When ψ1 = · · · = ψρ = 0, equations (12.22)–(12.23) represent a chain of ρ

integrators. The two main sources for the model (12.21)–(12.25) in this case are
the normal form of Section 8.1 and models of mechanical and electromechanical
systems where displacement variables are measured while their derivatives (veloc-
ities, accelerations, etc.) are not measured. When ψ1 to ψρ−1 are independent
of u, equations (12.22)–(12.23) are in the strict feedback form (9.17). If y is the
only measured variable, we can drop equation (12.25). However, in many problems,
we can measure some state variables in addition to y. For example, the magnetic
levitation system of Section A.8 is modeled by

ẋ1 = x2
4cx23
ẋ2 = −bx2 + 1 −
(1 + x1 )2

1 βx2 x3
ẋ3 = −x3 + u +
T (x1 ) (1 + x1 )2
where the normalized variables x1 , x2 , and x3 are the ball position, its velocity, and
the electromagnet current, respectively. Typically, we measure the ball position x1
and the current x3 . The model fits the form (12.21)–(12.25) with (x1 , x2 ) as the
x component and x3 as the w component. The measured outputs are y = x1 and
z = x3 .
We use a two-step approach to design the output feedback controller. First, a
partial state feedback controller that uses measurements of x and z is designed to
asymptotically stabilize the origin. Then, a high-gain observer is used to estimate
x from y. The state feedback controller is allowed to be a dynamical system of the
form
ϑ̇ = Γ(ϑ, x, z), u = γ(ϑ, x, z) (12.26)
where γ and Γ are locally Lipschitz functions in their arguments over the domain
of interest and globally bounded functions of x. Moreover, γ(0, 0, 0) = 0 and
Γ(0, 0, 0) = 0. A static state feedback controller u = γ(x, z) is a special case of
the foregoing equation by dropping the ϑ̇-equation. The more general form (12.26)
allows us to include, among other things, integral control as we shall see in Chap-
ter 13. If the functions γ and Γ are not globally bounded in x we can saturate
them or their x entries outside a compact set of x. The saturation levels can be
determined by analytically calculated bounds, as it was done in Example 12.3, or
by extensive simulations that cover several initial conditions in the compact set of
interest.
For convenience, we write the closed-loop system under state feedback as

Ẋ = f (X ) (12.27)

where X = col(w, x, ϑ). The output feedback controller is taken as

ϑ̇ = Γ(ϑ, x̂, z), u = γ(ϑ, x̂, z) (12.28)

12.4. HIGH-GAIN OBSERVERS AND THE SEPARATION PRINCIPLE 295

where x̂ is generated by the high-gain observer

αi
x̂˙ i = x̂i+1 + ψi (x̂1 , . . . , x̂i , u) + i (y − x̂1 ), for 1 ≤ i ≤ ρ − 1 (12.29)
ε
˙x̂ρ = φ0 (x̂, z, u) + αρ (y − x̂1 ) (12.30)
ερ
where ε is a sufficiently small positive constant, and α1 to αρ are chosen such that
the roots of
sρ + α1 sρ−1 + · · · + αρ−1 s + αρ = 0 (12.31)
have negative real parts. The function φ0 (x̂, z, u) implements φ0 (x, z, u), which is a
nominal model of φ(w, x, u) that is required to be locally Lipschitz in its arguments
over the domain of interest and globally bounded in x. Moreover, φ0 (0, 0, 0) = 0.
The following theorem states the properties of the closed-loop system under the
output feedback controller.2
Theorem 12.2 Consider the closed-loop system of the plant (12.21)–(12.25) and
the output feedback controller (12.28)–(12.30). Suppose the origin of (12.27) is
asymptotically stable and R is its region of attraction. Let S be any compact set
in the interior of R and Q be any compact subset of Rρ . Then, given any μ > 0
there exist ε∗ > 0 and T ∗ > 0, dependent on μ, such that for every 0 < ε ≤ ε∗ , the
solutions (X (t), x̂(t)) of the closed-loop system, starting in S × Q, are bounded for
all t ≥ 0 and satisfy
X (t) ≤ μ and x̂(t) ≤ μ, ∀ t ≥ T∗ (12.32)
X (t) − Xr (t) ≤ μ, ∀t≥0 (12.33)
where Xr is the solution of (12.27), starting at X (0). Moreover, if the origin of
(12.27) is exponentially stable, then the origin of the closed-loop system is exponen-
tially stable and S × Q is a subset of its region of attraction. 3
The theorem shows a number of properties of the output feedback controller when
ε is sufficiently small. First, (12.32) shows that the trajectories can be brought to
an arbitrarily small neighborhood of the origin by choosing ε small enough. Second,
(12.33) shows that the solution X (t) under output feedback approaches the solution
Xr (t) under state feedback as ε tends to zero. Third, the output feedback controller
recovers the region of attraction of the state feedback controller in the sense that
the foregoing two properties hold for any compact set in the interior of the region
of attraction. Finally, it recovers exponential stability when the origin under state
feedback is exponentially stable.3 As a corollary of the theorem, it is clear that if
2 The theorem is proved in [74, Theorem 14.6] for the case when ψ = · · · = ψ
1 ρ−1 = 0. The
presence of ψi ’s does not alter the proof in view of the fact, shown in Section 11.4, that the scaled
estimation error η satisfies equation (11.36) where δ satisfies (11.37). The steps of the proof are
outlined in Example 12.3.
3 For convenience, recovery of asymptotic stability is shown only for the exponentially stable

case. See [9] for the more general case when the origin is asymptotically, but not exponentially,
stable.
296 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

the state feedback controller achieves global or semiglobal stabilization with local
exponential stability, then for sufficiently small ε, the output feedback controller
achieves semiglobal stabilization with local exponential stability.
The trajectory recovery property (12.33) has a significant practical implication
because it allows the designer to design the state feedback controller to meet tran-
sient response specifications, constraints on the state variables, and/or constraints
on the control inputs. Then, by saturating the state estimate x̂ and/or the control
u outside compact sets of interest to make the functions γ(ϑ, x̂, z), Γ(ϑ, x̂, z), and
φ0 (x̂, z, u) globally bounded function in x̂, he/she can proceed to tune the parame-
ter ε by decreasing it monotonically to bring the trajectories under output feedback
close enough to the ones under state feedback. This establishes a separation prin-
ciple where the state feedback design is separated from the observer design.

12.5 Robust Stabilization of Minimum Phase Systems

In this section we consider a single-input–single-output nonlinear system that has
relative degree ρ and can be transformed into the normal form of Section 8.1. We
assume that the system is minimum phase. Our goal is to design a feedback con-
troller to stabilize the origin in the presence of matched model uncertainty and
time-varying disturbance. In Section 12.5.1, the system has relative degree one.
The design of a robust stabilizing controller in this case follows in a straightforward
way from the state feedback stabilization techniques of Chapter 10. We design
a sliding mode controller. In Section 12.5.2 we consider a system of relative de-
gree higher than one and show how high-gain observers can be used to reduce the
stabilization problem to that of a relative-degree-one system.

12.5.1 Relative Degree One

When ρ = 1, the normal form is given by

η̇ = f0 (η, y), ẏ = a(η, y) + b(η, y)u + δ(t, η, y, u) (12.34)

where η ∈ Rn−1 , u ∈ R, y ∈ R, the functions f0 , a and b are locally Lipschitz,

f0 (0, 0) = 0, and a(0, 0) = 0. Equation (12.34) includes a matched time-varying
disturbance δ, which is assumed to be piecewise continuous in t and locally Lipschitz
in (η, y, u). Let D ⊂ Rn be a domain that contains the origin and assume that for
all (η, y) ∈ D, b(η, y) ≥ b0 > 0. Equation (12.34) is in the regular form (10.4)
with ξ = y. Because the system is minimum phase, the origin of η̇ = f0 (η, 0) is
asymptotically stable. Therefore, we can take the sliding surface as y = 0. We
assume that there is a (continuously differentiable) Lyapunov function V (η) that
satisfies the inequalities

α1 (η) ≤ V (η) ≤ α2 (η) (12.35)

12.5. ROBUST STABILIZATION OF MINIMUM PHASE SYSTEMS 297

∂V
f0 (η, y) ≤ −α3 (η), ∀ η ≥ α4 (|y|) (12.36)
∂η
for all (η, y) ∈ D, where α1 to α4 are class K functions. By Theorem 4.7, inequality
(12.36) implies regional input-to-state stability of the system η̇ = fa (η, y) with input
y. Similar to (10.7), we take the control as

u = ψ(y) + v

where ψ could be zero or could be chosen to cancel part of the known term on the
right-hand side of ẏ, which depends only on y. We assume that we know a locally
Lipschitz function (y) ≥ 0 and a constant κ0 ∈ [0, 1) such that the inequality
 
 a(η, y) + b(η, y)ψ(y) + δ(t, η, y, ψ(y) + v) 
  ≤ (y) + κ0 |v| (12.37)
 b(η, y) 

holds for all t ≥ 0, (η, y) ∈ D, and u ∈ R. Similar to (10.12), take v as

 
y
v = −β(y) sat
μ
where β(y) is a locally Lipschitz function that satisfies
(y)
β(y) ≥ + β0 (12.38)
1 − κ0
for some β0 > 0. The overall control
 
y
u = ψ(y) − β(y) sat (12.39)
μ
is similar to the state feedback case except that we take s = y and restrict β and ψ
to be functions of y rather than the whole state vector. The following two theorems
are corollaries of Theorems 10.1 and 10.2.
Theorem 12.3 Consider the system (12.34). Suppose there exist V (η), (y), and
κ0 , which satisfy (12.35), (12.36), and (12.37). Let u be given by (12.39) where β
satisfies (12.38). Define the class K function α by α(r) = α2 (α4 (r)) and suppose μ,
c > μ, and c0 ≥ α(c) are chosen such that the set

Ω = {V (η) ≤ c0 } × {|y| ≤ c}, with c0 ≥ α(c) (12.40)

is compact and contained in D. Then, Ω is positively invariant and for any initial
state in Ω, the state is bounded for all t ≥ 0 and reaches the positively invariant set

Ωμ = {V (η) ≤ α(μ)} × {|y| ≤ μ} (12.41)

in finite time. If the assumptions hold globally and V (η) is radially unbounded, the
foregoing conclusion will hold for any initial state. 3
298 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

Theorem 12.4 Suppose all the assumptions of Theorem 12.3 are satisfied with
(0) = 0 and the origin of η̇ = f0 (η, 0) is exponentially stable. Then, there exists
μ∗ > 0 such that for all 0 < μ < μ∗ , the origin of the closed-loop system is ex-
ponentially stable and Ω is a subset of its region of attraction. If the assumptions
hold globally and V (η) is radially unbounded, the origin will be globally uniformly
asymptotically stable. 3

12.5.2 Relative Degree Higher Than One

The normal form of a system having relative degree ρ > 1 is given by

η̇ = f0 (η, ξ) (12.42)
ξ˙i = ξi+1 , for 1 ≤ i ≤ ρ − 1 (12.43)
ξ˙ρ = a(η, ξ) + b(η, ξ)u + δ(t, η, ξ, u) (12.44)
y = ξ1 (12.45)

where η ∈ Rn−ρ , ξ = col(ξ1 , ξ2 , . . . , ξρ ), u ∈ R, y ∈ R, and the functions f0 , a and

b are locally Lipschitz. The matched disturbance δ is piecewise continuous in t and
locally Lipschitz in (η, ξ, u). Let D ⊂ Rn be a domain that contains the origin and
assume that for all (η, ξ) ∈ D, b(η, ξ) ≥ b0 > 0. Because the system is minimum
phase, the origin of η̇ = f0 (η, 0) is asymptotically stable.
From the properties of high-gain observers, presented in Section 11.4, and the
separation principle of Section 12.4, we expect that if we can design a partial state
feedback controller, in terms of ξ, which stabilizes or practically stabilizes the origin,
we will be able to recover its performance by using a high-gain observer to estimate
ξ. Based on this expectation we will proceed to design a controller that uses feedback
from ξ.4
With measurement of ξ we can convert the relative-degree-ρ system into a
relative-degree-one system with respect to the output

s = k1 ξ1 + k2 ξ2 + · · · + kρ−1 ξρ−1 + ξρ (12.46)

The normal form of this system is given by

ż = f¯0 (z, s), ṡ = ā(z, s) + b̄(z, s)u + δ̄(t, z, s, u) (12.47)

where
⎡ ⎤ ⎡ ⎤
η f0 (η, ξ)
⎢ ⎥ξ1 ⎢ ξ2 ⎥
⎢ ⎥ ⎢ ⎥ 
ρ−1
⎢ ⎥ .. ⎢ ⎥
z=⎢ ⎥, . f0 (z, s) = ⎢ ... ⎥ ,
¯ ā(z, s) = ki ξi+1 + a(η, ξ)
⎢ ⎥ ⎢ ⎥
⎣ξρ−2 ⎦ ⎣ ξρ−1 ⎦ i=1

ξρ−1 ξρ
4 Alternatively, robust stabilization can be achieved using sliding mode observers [85, 124].
12.5. ROBUST STABILIZATION OF MINIMUM PHASE SYSTEMS 299

b̄(z, s) = b(η, ξ), δ̄(t, z, s, u) = δ(t, η, ξ, u)

)ρ−1
with ξρ = s − i=1 ki ξi . The functions f¯0 , ā, b̄, and δ̄ inherit the properties of
f0 , a, b, and δ, respectively. In particular, b̄(z, s) ≥ b0 > 0 for all (η, ξ) ∈ D. The
zero dynamic of (12.47) is ż = f¯0 (z, 0). This system can be written as the cascade
connection 


η̇ = f0 (η, ξ)  , ζ̇ = F ζ (12.48)
 ρ−1
ξρ =− i=1 k i ξi

where ⎡ ⎤
⎡ ⎤ 0 1 0 ··· 0
ξ1 ⎢ ⎥
⎢ ξ2 ⎥ ⎢ 0 0 1 ··· 0 ⎥
⎢ ⎥ ⎢ .. .. ⎥
ζ = ⎢ . ⎥, F =⎢ . . ⎥
⎣ .. ⎦ ⎢ ⎥
⎣ 0 ··· 0 1 ⎦
ξρ−1
−k1 −k2 ··· −kρ−2 −kρ−1
Because the origin η = 0 of η̇ = f0 (η, 0) is asymptotically stable by the mini-
mum phase assumption, the origin z = 0 of the cascade connection (12.48) will be
asymptotically stable if the matrix F is Hurwitz. Therefore, k1 to kρ−1 in (12.46)
are chosen such that the polynomial

λρ−1 + kρ−1 λρ−2 + · · · + k2 λ + k1 (12.49)

is Hurwitz; that is, all its roots have negative real parts. Consequently, k1 to kρ−1
are positive constants. As in Section C.1, asymptotic stability of the origin of the
cascade
 connection (12.48) can be shown by the Lyapunov function V1 (z) = cV0 (η)+
T
ζ P ζ, where V0 is a Lyapunov function of η̇ = f0 (η, 0), which is guaranteed to
exist by (the converse Lyapunov) Theorem 3.9, P is the positive definite solution
of the Lyapunov equation P F + F T P = −I, and c is a sufficiently small positive
constant. The function V1 (z) is not continuously differentiable in the neighborhood
of the origin because it is not continuously differentiable on the manifold ζ = 0.
However, once we establish asymptotic stability of the origin of (12.48), Theorem 3.9
ensures the existence of a smooth Lyapunov function. Using the local Lipschitz
property of f¯0 (z, s) with respect to s, it is reasonable to assume that there is a
(continuously differentiable) Lyapunov function V (z) that satisfies the inequalities

α1 (z) ≤ V (z) ≤ α2 (z) (12.50)

∂V ¯
f0 (z, s) ≤ −α3 (z), ∀ z ≥ α4 (|s|) (12.51)
∂η
for all (η, ξ) ∈ D, where α1 to α4 are class K functions. Another property of the
cascade connection (12.48) is that its origin z = 0 will be exponentially stable if the
origin η = 0 of η̇ = f0 (η, 0) is exponentially stable. This is shown in Section C.1.
We note that if ρ = n the zero dynamics of (12.47) will be ζ̇ = F ζ.
300 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

Thus, we have converted a relative-degree-ρ system into a relative-degree-one

system that satisfies the assumptions of Section 12.5.1. We proceed to design a
sliding mode controller as in Section 12.5.1, with the exception that the functions
ψ, , and β will be allowed to depend on the whole vector ξ and not only the output
s since we are working under the assumption that ξ is measured. Let

u = ψ(ξ) + v

for some locally Lipschitz function ψ, with ψ(0) = 0, and suppose we know a locally
Lipschitz function (ξ) ≥ 0 and a constant κ0 ∈ [0, 1) such that the inequality
) 
 ρ−1 k ξ 
 i=1 i i+1 + a(η, ξ) + b(η, ξ)ψ(ξ) + δ(t, η, ξ, ψ(ξ) + v) 
  ≤ (ξ) + κ0 |v| (12.52)
 b(η, ξ) 

holds for all t ≥ 0, (η, ξ) ∈ D, and u ∈ R. We note that the left-hand side of (12.52)
is the same as  
 ā(z, s) + b̄(z, s)ψ(ξ) + δ̄(t, z, s, ψ(ξ) + v) 
 
 b̄(z, s) 

Similar to (10.12), take v as

 
s
v = −β(ξ) sat
μ

where β(ξ) is a locally Lipschitz function that satisfies

(ξ)
β(ξ) ≥ + β0 (12.53)
1 − κ0
for some β0 > 0. The overall control
 
s def
u = ψ(ξ) − β(ξ) sat = γ(ξ) (12.54)
μ

is similar to the state feedback case except that we restrict β and ψ to be functions
of ξ rather than the whole state vector. Define the class K function α by α(r) =
α2 (α4 (r)) and suppose μ, c > μ, and c0 ≥ α(c) are chosen such that the set

Ω = {V (z) ≤ c0 } × {|s| ≤ c}, with c0 ≥ α(c) (12.55)

is compact and contained in D. By Theorem 10.1 we conclude that Ω is positively

invariant and for all initial states in Ω, the trajectories enter the positively invariant
set
Ωμ = {V (z) ≤ α(μ)} × {|s| ≤ μ} (12.56)
in finite time. If (0) = 0 and the origin of η̇ = f0 (η, 0) is exponentially stable,
then Theorem 10.2 shows that there exists μ∗ > 0 such that for all 0 < μ < μ∗ ,
12.5. ROBUST STABILIZATION OF MINIMUM PHASE SYSTEMS 301

the origin of the closed-loop system is exponentially stable and Ω is a subset of its
region of attraction.
To implement the partial state feedback controller (12.54) using a high-gain
observer, we start by saturating the control outside the compact set Ω to make it a
globally bounded function of ξ. This is needed to overcome the peaking phenomenon
of high-gain observers.5 There are different ways to saturate the control. We can
saturate each component of ξ in the functions ψ and β. Let

Mi = max{|ξi |}, 1≤i≤ρ

Ω

and take ψs and βs as the functions ψ and β with ξi replaced Mi sat(ξi /Mi ).
Alternatively, let

Mψ = max{|ψ(ξ)|}, Mβ = max{|β(ξ)|}
Ω Ω

and take ψs (ξ) = Mψ sat(ψ(ξ)/Mψ ) and βs (ξ) = Mβ sat(β(ξ)/Mβ ). In either case,

the control is taken as  
s
u = ψs (ξ) − βs (ξ) sat
μ
We may also saturate the control signal itself. Let

Mu = max{|ψ(ξ) − β(x) sat(s/μ)|}

Ω

and take the control as

 
ψ(ξ) − β(ξ) sat(s/μ)
u = Mu sat
Mu

Irrespective of how we saturate the control (12.54), the saturation will not be active
for all initial states in Ω because the saturated control coincides with the original
control (12.54) over Ω. We use the high-gain observer
˙ αi
ξˆi = ξˆi+1 + i (y − ξˆ1 ), 1 ≤ i ≤ ρ − 1 (12.57)
ε
˙
ξˆρ ˆ ˆ + αρ (y − ξˆ1 )
= a0 (ξ) + b0 (ξ)u (12.58)
ερ
ˆ where ε is a sufficiently small positive constant, α1 to αρ are
to estimate ξ by ξ,
chosen such that the polynomial

sρ + α1 sρ−1 + · · · + αρ−1 s + αρ (12.59)

is Hurwitz, and a0 (ξ) and b0 (ξ) are locally Lipschitz, globally bounded functions of
ξ, which serve as nominal models of a(η, ξ) and b(η, ξ), respectively. The functions
5 See Sections 11.4 and 12.4.
302 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

a0 and b0 are not allowed to depend on η because it is not estimated by the observer
(12.57)–(12.58). The choice a0 = b0 = 0 results in a linear observer. This choice is
typically used when no good models of a and b are available, or when it is desired
to use a linear observer. The output feedback controller is taken as
ˆ
u = γs (ξ) (12.60)
ˆ is given by
where γs (ξ)
+ ,
ˆ − β(ξ)
ψ(ξ) ˆ sat(ŝ/μ)
ˆ − βs (ξ)
ψs (ξ) ˆ sat(ŝ/μ) or Mu sat
Mu
and

ρ−1
ŝ = ki ξˆi + ξˆρ
i=1
The properties of the output feedback controller are stated in the following
theorems, whose proofs are given in Appendix D.
Theorem 12.5 Consider the system (12.42)–(12.45) and let k1 to kρ−1 be chosen
such that the polynomial (12.49) is Hurwitz. Suppose there exist V (z), (ξ), and κ0 ,
which satisfy (12.50), (12.51), and (12.52), and β is chosen to satisfy (12.53). Let
Ω and Ωμ be defined by (12.55) and (12.56), respectively. Consider the high-gain
observer (12.57)–(12.58), where α1 to αρ are chosen such that the polynomial (12.59)
is Hurwitz, and the output feedback controller (12.60). Let Ω0 be a compact set in
the interior of Ω and X be a compact subset of Rρ . Suppose the initial states satisfy
ˆ ∈ X. Then, there exists ε∗ , dependent on μ, such that for
(η(0), ξ(0)) ∈ Ω0 and ξ(0)
∗ ˆ
all ε ∈ (0, ε ) the states (η(t), ξ(t), ξ(t)) of the closed-loop system are bounded for
all t ≥ 0 and there is a finite time T , dependent on μ, such that (η(t), ξ(t)) ∈ Ωμ
for all t ≥ T . Moreover, if (ηr (t), ξr (t)) is the state of the closed-loop system
under the state feedback controller (12.54) with initial conditions ηr (0) = η(0) and
ξr (0) = ξ(0), then given any λ > 0, there exists ε∗∗ > 0, dependent on μ and λ,
such that for all ε ∈ (0, ε∗∗ ),
η(t) − ηr (t) ≤ λ and ξ(t) − ξr (t) ≤ λ, ∀ t ∈ [0, T ] (12.61)
3
Theorem 12.6 Suppose all the assumptions of Theorems 12.5 are satisfied with
(0) = 0 and the origin of η̇ = f0 (η, 0) is exponentially stable. Then, there exists
μ∗ > 0 and for each μ ∈ (0, μ∗ ) there exists ε∗ > 0, dependent on μ, such that for all
ε ∈ (0, ε∗ ), the origin of the closed-loop system under the output feedback controller
(12.60) is exponentially stable and Ω0 × X is a subset of its region of attraction. 3
Theorems 12.5 and 12.6 show that the output feedback controller (12.60) recovers
the stabilization or practical stabilization properties of the state feedback controller
(12.54) for sufficiently small ε. It also recovers its transient behavior, as shown by
(12.61).
12.6. EXERCISES 303

Example 12.4 In Example 10.1, we designed the state feedback controller

 
s
u = −2(|x1 | + |x2 | + 1) sat
μ

where x1 = θ − π, x2 = θ̇, s = x1 + x2 , and μ < 0.308 to stabilize the pendulum

θ̈ + sin θ + bθ̇ = cu

at (θ = π, θ̇ = 0) under the assumption that 0 ≤ b ≤ 0.2 and 0.5 ≤ c ≤ 2. Let us

take μ = 0.1. The controller achieves global stabilization. Suppose now that we
only measure θ. In preparation for using a high-gain observer, we will saturate the
state feedback control outside a compact set. With V1 (x) = 12 x21 , the set Ω is given
by Ω = {|x1 | ≤ c/θ1 } × {|s| ≤ c}, where c > 0 and 0 < θ1 < 1. Suppose that with
c = 2π and θ = 0.8, Ω includes all the initial states of interest. Over Ω, we have
|x1 | ≤ 2.5π and |x2 | ≤ 4.5π. Therefore, the output feedback controller is taken as
       
|x̂1 | |x̂2 | ŝ
u = −2 2.5π sat + 4.5π sat + 1 sat
2.5π 4.5π μ

where ŝ = x̂1 + x̂2 and x̂ = col(x̂1 , x̂2 ) is determined by the high-gain observer
2 1
x̂˙ 1 = x̂2 + (x1 − x̂1 ), x̂˙ 2 = φ0 (x̂, u) + 2 (x1 − x̂1 )
ε ε
with sufficiently small ε. We will consider two choices of φ0 . The first choice φ0 = 0
yields a linear observer. The second choice φ0 (x̂) = − sin(x̂1 + π) − 0.1x̂2 + 1.25u
duplicates the right-hand side of the state equation with 0.1 and 1.25 as nominal
values of b and c, respectively. In Figure 12.7, we compare the performance of the
state and output feedback controllers for ε = 0.05 and ε = 0.01. The pendulum
parameters are b = 0.01 and c = 0.5, and the initial conditions are θ(0) = θ̇(0) =
x̂1 (0) = x̂2 (0) = 0. Peaking is induced because x1 (0) = −π = x̂1 (0). Irrespective
of the choice of φ0 , the simulation results show that the response under output
feedback approaches the one under state feedback as ε decreases. For ε = 0.01, the
inclusion of φ0 in the observer has little effect on the response. However, for the
larger value ε = 0.05, there is an advantage for including φ0 in the observer.6

12.6 Exercises
12.1 For each of the following systems, design a globally stabilizing output feed-
back controller.
(1) ẋ1 = x2 , ẋ2 = −x31 + tanh(u), y = x2
6 Othersimulation results, as in [74, Example 14.19], show that the advantage of including φ0
may not be realized if it is not a good model of the state equation.
304 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

(a) (b)
3.5 4
3
2.5 3

2
2

ω
θ

1.5
SF
1 1
OF ε = 0.05
0.5 OF ε = 0.01
0 0
0 1 2 3 0 1 2 3

(c) (d)
3.5 4
3
2.5 3

2
2
ω
θ

1.5
1 1
0.5
0 0
0 1 2 3 0 1 2 3
Time Time

Figure 12.7: Comparison of state feedback (SF) and output feedback (OF) for Exam-
ple 12.4. Figures (a) and (b) show θ and ω = θ̇ for a linear high-gain observer with
φ0 = 0. Figures (c) and (d) show θ and ω for a nonlinear high-gain observer with
φ0 = − sin(x̂1 + π) − 0.1x̂2 + 1.25u.

(2) ẋ1 = x2 , ẋ2 = −x31 + tanh(u), y = x1

(3) ẋ1 = β1 (x1 ) + x2 , ẋ2 = β2 (x1 ) + β3 (x1 )x2 + u, y = x1

(4) ẋ1 = sin x1 + x2 , ẋ2 = u, y = x1

12.2 For each of the following systems, design a stabilizing output feedback con-
troller such that the set {|x1 | ≤ 1, |x2 | ≤ 1} is included in the region of attraction.

(1) ẋ1 = x1 + x2 /(1 + x21 ), ẋ2 = −x2 + u, y = x1

(2) ẋ1 = ax21 + x2 , ẋ2 = u, y = x1 , a is unknown with |a| ≤ 1

(3) ẋ1 = x2 , ẋ2 = −x2 |x2 | + u, y = x1

12.6. EXERCISES 305

(4) ẋ1 = x31 + x2 , ẋ2 = x1 x2 + u, ẋ3 = x1 x2 − x3 , y = x1

12.3 Consider the inverted pendulum on a cart (A.41)–(A.44), where it is required

to stabilize the pendulum at (θ, y) = (0, 0) and the measured outputs are θ and y.

(a) Using linearization with the (A.45) data, design a stabilizing output feedback
controller.

(b) Using simulation, find the range of x1 (0), when all other initial conditions are
zero, for which the pendulum is stabilized.

12.4 Consider the inverted pendulum (A.47) with the state feedback controller of
Exercise 9.5(a). Suppose you can only measure θ.

(a) With a = 1, design an observer with linear error dynamics and use it to im-
plement the controller. Use simulation to find the range of x1 (0), when all
other initial conditions are zero, for which the pendulum is stabilized and
|x1 (t)| ≤ π/2 for all t. Compare the performance of the state and output
feedback controllers.

(b) Repeat (a) using a high-gain observer with ε = 0.01 and saturated control.

(c) Compare the performance of the two observers when a = 1 in the controller
but it is perturbed in the pendulum with a ∈ [0.5, 1.5].

12.5 Repeat Exercise 12.4 for the the state feedback controller of Exercise 9.5(b).

12.6 Repeat Exercise 12.4 for the the state feedback controller of Exercise 10.10.

12.7 Consider the state feedback controller of Exercise 9.8(b) for the magnetic
levitation system.

(a) Implement the controller, using a high-gain observer and saturated control,
when the measured output is y = col(x1 , x3 ).

(b) Repeat (a) when y = x1 . Hint: Transform the system into the form (11.26)–
(11.28).

(c) Use simulation to compare the performance of the state and output feedback
controllers.

12.8 Repeat Exercise 12.7 for the the state feedback controller of Exercises 9.8(d).

12.9 Repeat Exercise 12.7 for the the state feedback controller of Exercises 9.8(e).

12.10 Repeat Exercise 12.7 for the the state feedback controller of Exercises 10.7.
Take into consideration the uncertainty of b and c.
306 CHAPTER 12. OUTPUT FEEDBACK STABILIZATION

12.11 Consider the electrostatic microactuator (A.33).

(a) Assuming you can only measure x1 and x3 , design an observer with linear
error dynamics to implement the state feedback controller of Exercise 9.9(b).
Use simulation to compare the performance
√ of the state and output feedback
controllers when x(0) = col(0.1, 0, 0.3) and x̂(0) = 0.

(b) If you can only measure x1 , repeat (a) using a high-gain observer.

12.12 Repeat Exercise 12.11 for the controller of Exercise 9.9(d).

12.13 Repeat Exercise 12.11 for the controller of Exercise 9.9(e).

12.14 Repeat Exercise 12.11 for the controller of Exercise 9.9(f).

12.15 Consider the electrostatic microactuator (A.33) where you can only measure
x3 . Design an output feedback controller to stabilize the plate at x1 = r. Assume
that ζ ∈ [0.1, 0.5], T ∈ [0.1, 0.5], and r ∈ [0.1, 0.9]. Simulate
√ the closed-loop system
with r = 0.5, ζ = 0.1, T = 0.2, and x(0) = col(0.1, 0, 0.3).

12.16 Consider the two-link manipulator defined by (A.36) and (A.37) with the
(A.38) data. Assume that the measured output is y = col(q1 , q2 ). It is desired to
regulate the manipulator to qr = (π/2, π/2) under the control constraints |u1 | ≤
6000 and |u2 | ≤ 5000. Design a passivity-based controller that meets the control
constraints. Hint: See Example 12.2 and note that you will need to replace Kp e
and Kd z by nonlinear functions.

12.17 Consider the TORA system of (A.49)–(A.52) where the measured output
is y = x1 = θ. Design a globally stabilizing output feedback controller. Hint: See
Exercises 5.10 and 9.13.

12.18 Consider the sliding mode controller of the TORA system in Exercise 10.13.
Design a high-gain observer to implement this controller with constant β when the
measured output is y = col(x1 , x3 ). Compare the performance of the state and
output feedback controllers using the (A.53) data and the controller parameters of
Exercise 10.14(a).

12.19 Repeat Problems (1) to (5) of Exercise 10.1 with the goal of achieving
semiglobal stabilization when you only measure x1 .
Chapter 13

Tracking and Regulation

In this chapter we study tracking and regulation problems for single-input–single-

output systems representable in the normal form:

η̇ = f0 (η, ξ) (13.1)
ξ˙i = ξi+1 , for 1 ≤ i ≤ ρ − 1 (13.2)
ξ˙ρ = a(η, ξ) + b(η, ξ)u (13.3)
y = ξ1 (13.4)

where η ∈ Dη ⊂ Rn−ρ , ξ = col(ξ1 , . . . , ξρ ) ∈ Dξ ⊂ Rn , for some domains Dη and Dξ .

The functions f0 , a, and b are locally Lipschitz and b = 0 over Dη × Dξ . When the
relative degree ρ = n, the system has no zero dynamics. In this case, the η variable
and its equation are dropped, but the rest of the development remains the same.
The goal of the tracking problem is to design a feedback controller such that the
output y asymptotically tracks a reference signal r; that is, limt→∞ [y(t) − r(t)] = 0,
while ensuring boundedness of all state variables. The regulation problem is a
special case of the tracking problem when r is constant. We assume that the system
(13.1)–(13.4) and the reference signal satisfy the following assumptions.

Assumption 13.1

b(η, ξ) ≥ b0 > 0, ∀ η ∈ Dη , ξ ∈ Dξ

Assumption 13.2 The system η̇ = f0 (η, ξ) is bounded-input–bounded-state stable

over Dη × Dξ ; that is, for every bounded ξ(t) in Dξ , the solution η(t) is bounded
and contained in Dη .

Assumption 13.3 r(t) and its derivatives up to r(ρ) (t) are bounded for all t ≥ 0
and the ρth derivative r(ρ) (t) is a piecewise continuous function of t. Moreover,
R = col(r, ṙ, . . . , r(ρ−1) ) ∈ Dξ for all t ≥ 0.

307
308 CHAPTER 13. TRACKING AND REGULATION

For minimum phase systems, where the origin of η̇ = f0 (η, 0) is asymptotically

stable, Assumption 13.2 holds locally in view of Lemma 4.7. It will hold globally or
regionally if the system is input-to-state stable or regionally input-to-state stable,
respectively. Notice, however, that the assumption might hold while the origin of
η̇ = f0 (η, 0) is stable but not asymptotically stable, as in the system

|ξ|
η̇ = − η+ξ
|ξ| + 1
When ξ = 0, the origin of η̇ = 0 is stable but not asymptotically stable. However,
Assumption 13.2 is satisfied because the derivative of V0 = 12 η 2 satisfies V̇0 ≤ 0 for
|η| ≥ |ξ| + 1.
The reference signal r(t) could be specified, together with its derivatives, as
given functions of time, or it could be the output of a reference model driven by
a command input uc (t). In the latter case, the assumptions on r can be met by
appropriately choosing the reference model. For example, for a relative degree two
system, a reference model could be a second-order linear time-invariant system
represented by the transfer function

ωn2
s2 + 2ζωn s + ωn2

where the positive constants ζ and ωn are chosen to shape the reference signal r(t)
for a given input uc (t). The signal r(t) can be generated by using the state model

ẏ1 = y2 , ẏ2 = −ωn2 y1 − 2ζωn y2 + ωn2 uc , r = y1

from which we can compute ṙ = y2 , and r̈ = ẏ2 . If uc (t) is piecewise continuous

and bounded, then r(t), ṙ(t), and r̈(t) will satisfy Assumption 13.3.
The change of variables

e1 = ξ1 − r, e2 = ξ2 − r(1) , ..., eρ = ξρ − r(ρ−1) (13.5)

transforms the system (13.1)–(13.4) into the form

η̇ = f0 (η, ξ) (13.6)
ėi = ei+1 , for 1 ≤ i ≤ ρ − 1 (13.7)
ėρ = a(η, ξ) + b(η, ξ)u − r(ρ) (13.8)

In these coordinates, the goal of the design is to ensure that the vector e =
col(e1 , . . . , eρ ) = ξ − R is bounded for all t ≥ 0 and converges to zero as t tends to
infinity. Boundedness of e implies bounded of ξ since R is bounded, which in turn
implies bounded of η by assumption 13.2. A state feedback controller that achieves
this goal will use the error vector e, and for that we need the following assumption.
Assumption 13.4 The signals r, r(1) , . . . , r(ρ) are available to the controller.
 309

In many control problems, the designer has some freedom in choosing the ref-
erence signal r. For example, one of the typical problems in controlling robot
manipulators is moving the manipulator from an initial to a final point within some
time interval. The first task in approaching this problem is planning the path be-
tween the two points, which has to comply with any physical constraints due to the
presence of obstacles. Then, the motion trajectory is planned by specifying veloc-
ities and accelerations of the moving parts as functions of time. The outcome of
this trajectory planning process is the reference signal that the output variable has
to track.1 The freedom in choosing the reference signal can be used to improve the
performance of the system, especially in the presence of constraints on the control
signal.

Feedback controllers for tracking and regulation are classified in the same way as
stabilizing controllers. We speak of state feedback if x can be measured; otherwise,
we speak of output feedback. Also, the feedback controller can be static or dynamic.
It may achieve local, regional, semiglobal, or global tracking. The new element here
is that these phrases refer not only to the size of the initial state, but to the size
of the reference and disturbance signals as well. For example, in a typical problem,
local tracking means tracking is achieved for sufficiently small initial states and
sufficiently small R, while global tracking means tracking is achieved for any
initial state and any bounded R. When we achieve ultimate boundedness and the
ultimate bound can be made arbitrarily small by choice of design parameters, we
say that we achieve practical tracking, which can be local, regional, semiglobal, or
global, depending on the size of the initial states and the reference and disturbance
signals.

The first three sections present state feedback controllers for the tracking prob-
lem. Section 13.1 uses feedback linearization, while Section 13.2 uses sliding mode
control to deal with model uncertainty and matched time-varying disturbance. Sec-
tion 13.3 deals with the special case of moving the state of the system from an initial
equilibrium point to a final equilibrium point either asymptotically or in finite time.
The design divides the control task between a feedforward component that shapes
the transient response by shaping the reference input, and a feedback component
that ensures stability in the presence of model uncertainty and disturbance. In
Section 13.4 we discuss the use of integral action to achieve robust regulation in
the presence parametric uncertainty and constant reference and disturbance sig-
nals. In Section 13.5 we consider output feedback control, where the only measured
signal is y. We use high-gain observers to implement the sliding mode controllers
of Sections 13.2 and 13.4.

1 To read about trajectory planning in robot manipulators, see [119, 135].

310 CHAPTER 13. TRACKING AND REGULATION

13.1 Tracking
Rewrite the system (13.6)–(13.8) as
 
η̇ = f0 (η, ξ), ė = Ac e + Bc a(η, ξ) + b(η, ξ)u − r(ρ)

where ξ = e + R and the pair (Ac , Bc ) represents a chain of ρ integrators. By

feedback linearization,
 
u = −a(η, ξ) + r(ρ) + v /b(η, ξ)

reduces the system to the cascade connection

η̇ = f0 (η, ξ), ė = Ac e + Bc v

The ė-equation is stabilized by v = −Ke, where Ac −Bc K is Hurwitz. The complete

state feedback control is given by
 
u = −a(η, ξ) + r(ρ) − Ke /b(η, ξ) (13.9)

and the closed-loop system is

η̇ = f0 (η, ξ), ė = (Ac − Bc K)e (13.10)

Since Ac − Bc K is Hurwitz, e(t) is bounded and limt→∞ e(t) = 0. Consequently,

ξ = e + R is bounded and, by Assumption 13.2, η is bounded.

Example 13.1 Consider the pendulum equation (A.2):

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + cu, y = x1

The system has relative degree two in R2 and is in the normal form. It has no zero
dynamics. We want the output y to track a reference signal r(t), with bounded
derivatives ṙ(t) and r̈(t). Taking

e1 = x1 − r, e2 = x2 − ṙ

we obtain
ė1 = e2 , ė2 = − sin x1 − bx2 + cu − r̈
The state feedback control (13.9) is given by
1
u= [sin x1 + bx2 + r̈ − k1 e1 − k2 e2 ]
c
where K = [k1 , k2 ] assigns the eigenvalues of Ac − Bc K at desired locations in
the open left-half complex plane. Because all the assumptions hold globally, this
13.1. TRACKING 311

(a) (b)
2 2

1.5 1.5
Output

Output
1 1

0.5 0.5

0 0

−0.5 −0.5
0 2 4 6 8 10 0 2 4 6 8 10
Time Time
(c) (d)
2
5
1.5 Control
Output

1 0

0.5
−5
0
−10
−0.5
0 2 4 6 8 10 0 2 4 6 8 10
Time Time

Figure 13.1: Simulation of the tracking control of Example 13.1.

control achieves global tracking. Figure 13.1 shows the response of the system for the
nominal parameters b = 0.03 and c = 1 when r = sin(t/3) and x(0) = col(π/2, 0).
To illustrate the impact
 of the feedback gain K, we consider two different designs.√
The first design K = 1 1 assigns the eigenvalues of Ac − Bc K at −0.5 ± j0.5 3,
  √
and the second K = 9 3 assigns the eigenvalues at −1.5±j1.5 3. Figure 13.1(a)
shows the response of the first design while 13.1(b) shows the second one. The
dashed curve is the reference signal r and the solid curve is the output y. We can
see that the higher the feedback gain, the shorter the transient period. A more
significant effect is shown in Figure 13.1(c) where we plot the response when b and
c are perturbed to b = 0.015 and c = 0.5 due to doubling the mass. The solid
curve is the output under the first design and the dash-dot curve is the output
under the second design. The response of the first design deteriorates significantly
under perturbations. The higher feedback gain makes the response more robust
to perturbations. This comes at the expense of a larger control signal, as shown
in Figure 13.1(d), where here again solid and dash-dot curves are for the first and
second designs, respectively.
312 CHAPTER 13. TRACKING AND REGULATION

13.2 Robust Tracking

The system

η̇ = f0 (η, ξ)
ėi = ei+1 , 1≤i≤ρ−1
ėρ = a(η, ξ) + b(η, ξ)u + δ(t, η, ξ, u) − r(ρ) (t)

is a perturbation of (13.6)–(13.8) that contains the matched time-varying distur-

bance δ, which is assumed to be piecewise continuous in t and locally Lipschitz in
(η, ξ, u). We may also allow uncertainty in the functions f0 , a and b as long as As-
sumptions 13.1 and 13.2 are satisfied. We design a state feedback controller using
sliding mode control. The ė-equation is in the regular form (10.4). To design the
sliding manifold, we start with the system

ėi = ei+1 , 1≤i≤ρ−1

where eρ is viewed as the control input, and design eρ to stabilize the origin. For
this linear (controllable canonical form) system, we can achieve this task by the
linear control
eρ = −(k1 e1 + · · · + kρ−1 eρ−1 )
where k1 to kρ−1 are chosen such that the polynomial

λρ−1 + kρ−1 λρ−2 + · · · + k1

is Hurwitz. Then, the sliding manifold is taken as

s = (k1 e1 + · · · + kρ−1 eρ−1 ) + eρ = 0

and

ρ−1
ṡ = ki ei+1 + a(η, ξ) + b(η, ξ)u + δ(t, η, ξ, u) − r(ρ) (t)
i=1

The control u can be taken as

'ρ−1 (
1 
u=v or u=− ki ei+1 + â(η, ξ) − r (ρ)
(t) + v (13.11)
b̂(η, ξ) i=1

where in the second case we cancel the known terms on the right-hand side of the
ṡ-equation. The functions â and b̂ are nominal models of a and b, respectively.
When â = a and b̂ = b, the term
 
− a(η, ξ) − r(ρ) (t) /b(η, ξ)
13.2. ROBUST TRACKING 313

is the feedback linearizing term we used in the previous section. In either case, the
ṡ-equation can be written as
ṡ = b(η, ξ)v + Δ(t, η, ξ, v)
Suppose  
 Δ(t, η, ξ, v) 
 
 b(η, ξ)  ≤ (η, ξ) + κ0 |v|, 0 ≤ κ0 < 1

for all t ≥ 0, v ∈ R, and (η, ξ) ∈ Dη × Dξ , where  and κ0 are known. Then,

v = −β(η, ξ) sat(s/μ) (13.12)
where β(η, ξ) is a locally Lipschitz function that satisfies
β(η, ξ) ≥ (η, ξ)/(1 − κ0 ) + β0 , with β0 > 0
ensures that sṡ ≤ −β0 b0 (1 − κ0 )|s| for |s| ≥ μ. Due to Assumption 13.2, we focus
our attention on the behavior of e(t). Setting ζ = col(e1 , . . . , eρ−1 ), it can be seen
that ζ satisfies the equation
ζ̇ = (Ac − Bc K)ζ + Bc s
 
where K = k1 k2 . . . kρ−1 and the pair (Ac , Bc ) represents a chain of (ρ − 1)
integrators. The matrix Ac − Bc K is Hurwitz by design. Let P be the solution of
the Lyapunov equation P (Ac − Bc K) + (Ac − Bc K)T P = −I. The derivative of
V0 = ζ T P ζ satisfies the inequality
V̇0 = −ζ T ζ + 2ζ T P Bc s ≤ −(1 − θ)ζ2 , ∀ ζ ≥ 2P Bc  |s|/θ
where 0 < θ < 1. Since ζ T P ζ ≤ λmax (P )ζ2 , for any σ ≥ μ
{ζ ≤ 2P Bc  σ/θ} ⊂ {ζ T P ζ ≤ λmax (P )(2P Bc /θ)2 σ 2 }
Set ρ1 = λmax (P )(2P Bc /θ)2 and choose c > μ such that Ω = {ζ T P ζ ≤ ρ1 c2 } ×
{|s| ≤ c} is in the domain of interest. The set Ω is positively invariant because sṡ < 0
on the boundary {|s| = c} and V̇0 < 0 on the boundary {V0 = ρ1 c2 }. Let e(0) ∈ Ω.
Within finite time, e(t) enters the set Ωμ = {ζ T P ζ ≤ ρ1 μ2 }×{|s| ≤ μ}, which is also
positively invariant. Inside Ωμ , we can use (B.4) to show that maxζ T P ζ≤ρ1 μ2 |e1 | =
   √
ρ1 μ2 LP −1/2 , where L = 1 0 . . . 0 . Setting k = LP −1/2  ρ1 shows
that the ultimate bound on |e1 | is kμ.
Example 13.2 Reconsider the pendulum tracking problem from Example 13.1:
ė1 = e2 , ė2 = − sin x1 − bx2 + cu − r̈
where r(t) = sin(t/3), and suppose that b and c are uncertain parameters that
satisfy 0 ≤ b ≤ 0.1 and 0.5 ≤ c ≤ 2. Taking s = e1 + e2 , we have
ṡ = e2 − sin x1 − bx2 + cu − r̈ = (1 − b)e2 − sin x1 − bṙ − r̈
314 CHAPTER 13. TRACKING AND REGULATION

(a) (b)
2
1.2
1.5
1
0.8
Output

1
0.6

s
0.5 0.4
0.2
0
0
−0.5 −0.2
0 2 4 6 8 10 0 0.2 0.4 0.6 0.8 1
Time Time

Figure 13.2: Simulation of the tracking control of Example 13.2.

It can be verified that

 
 (1 − b)e2 − sin x1 − bṙ − r̈  |e2 | + 1 + 0.1/3 + 1/9
 ≤ ≤ 2|e2 | + 2.3
 c  0.5

We take β = 2|e2 | + 3, which yields the control

 
e1 + e2
u = −(2|e2 | + 3) sat
μ

Simulation results with μ = 0.1 and x(0) = col(π/2, 0) are shown in Figure 13.2 for
b = 0.03, c = 1 (solid curve) and b = 0.015, c = 0.5 (dash-dot curve). The dashed
curve is the reference signal. The output responses in Figure 13.2(a) are almost
indistinguishable. Figure 13.2(b) shows that the trajectories enter the boundary
layer {|s| ≤ μ} within the time interval [0, 0.3]. The parameter values used in this
figure are the same as in Figure 13.1(c), which illustrates the robustness of the
sliding mode controller.

13.3 Transition Between Set Points

Consider the system (13.1)–(13.4):

η̇ = f0 (η, ξ)
˙ξi = ξi+1 , 1≤i≤ρ−1
ξ˙ρ = a(η, ξ) + b(η, ξ)u
y = ξ1
13.3. TRANSITION BETWEEN SET POINTS 315

where Assumptions 13.1 and 13.2 are satisfied over the domain of interest. For a
¯ satisfies the equations
constant input ū, the equilibrium point (η̄, ξ)
¯
0 = f0 (η̄, ξ)
¯
0 = ξi+1 , 1≤i≤ρ−1
¯ ¯
0 = b(η̄, ξ) + a(η̄, ξ)ū
ȳ = ξ¯1

Therefore, ξ¯ = col(ȳ, 0, · · · , 0). Assume that

¯
0 = f0 (η̄, ξ)

has a unique solution η̄ in the domain of interest; denote it by η̄ = φη (ȳ). Moreover,

assume that η̄ is an asymptotically stable equilibrium point of η̇ = f0 (η, ξ) ¯ (a
minimum phase property). Because b(η, ξ) = 0, ū is given by
¯ def
a(η̄, ξ)
ū = − ¯ = φu (ȳ)
b(η̄, ξ)

Thus, to maintain the output at a constant value ȳ, we need to maintain the system
at the equilibrium point η̄ = φη (ȳ) and ξ¯ = col(ȳ, 0, · · · , 0), using the constant
control ū = φu (ȳ). For each ȳ there is a unique triple (η̄, ξ,¯ ū). Without loss of
generality we assume that φη (0) = 0 and φu (0) = 0.
Suppose now we want to move the system from equilibrium at y = 0 to equilib-
rium at y = y ∗ . This task can be cast as a tracking problem by taking the reference
signal as
r = y ∗ and r(i) = 0 for i ≥ 1
for which the initial state e(0) = col(−y ∗ , 0, . . . , 0). Shaping the transient response
of e(t) is straightforward when a and b are known and the linearizing feedback
control
−a(η, ξ) − Ke
u=
b(η, ξ)
is used because e(t) satisfies the equation ė = (Ac −Bc K)e, whose transient response
can be shaped by assignment of the eigenvalues of Ac − Bc K. When a and b are
perturbed, the transient response will deviate from the solution of ė = (Ac −Bc K)e.
To cope with uncertainty, we can use the sliding mode controller of the previous
section, but shaping the transient response will be more involved because during
the reaching phase s satisfies a nonlinear equation.
An alternative approach that enables us to shape the transient response is to
choose the reference r(t) as a smooth signal with r and its derivatives r(1) to r(ρ−1)
all starting from zero at t = 0. Then, r(t) approaches y ∗ , with its derivatives
approaching zero, either asymptotically or over a finite period time. Because r
and its derivatives are zero at t = 0, the initial vector e(0) = 0. In the case of
316 CHAPTER 13. TRACKING AND REGULATION

feedback linearization with no uncertainty, e(0) = 0 implies that e(t) ≡ 0 because

e satisfies the equation ė = (Ac − Bc K)e. Therefore y(t) ≡ r(t) for all t ≥ 0.
In the case of sliding mode control, it is shown in the previous section that the set
Ωμ = {ζ T P ζ ≤ ρ1 μ2 }×{|s| ≤ μ} is positively invariant, where ζ = col(e1 , . . . , eρ−1 )
and s = k1 e1 + · · · + kρ−1 eρ−1 + eρ . Because 0 ∈ Ωμ ,

e(0) = 0 ⇒ e(0) ∈ Ωμ ⇒ e(t) ∈ Ωμ , ∀t≥0

Hence, y(t) will be within O(μ) neighborhood of r(t) for all t ≥ 0.

The reference r(t) can be shaped to achieve the desired transient response. For
example, r(t) can be taken as the zero-state response of the Hurwitz transfer func-
tion
aρ
sρ + a1 sρ−1 + · · · + aρ−1 s + aρ
with a step input of amplitude y ∗ . The positive constants a1 to aρ are chosen to
shape the response of r. A state-space realization of this transfer function is
⎡ ⎤ ⎡ ⎤
1
⎢ .. ⎥ ⎢ ⎥ ∗
⎢ ⎥
ż = ⎢ . ⎥z + ⎢ ⎣
⎥y
⎦
⎣ 1 ⎦
−aρ −a1 aρ
 
r = 1 z

By taking z(0) = 0 we enforce the condition that the initial values of r and its
derivatives up to r(ρ−1) are all zero. As time approaches infinity, r(t) approaches
y ∗ while its derivatives approach zero.
Another approach is to plan the trajectory (r(t), ṙ(t), . . . , r(ρ−1) (t)) to move
from (0, 0, . . . , 0) to (y ∗ , 0, . . . , 0) in finite time T . Figure 13.3 shows an example
when ρ = 2. For this example,
⎧ 2
⎪
⎨ 2
at
for 0 ≤ t ≤ T2
2 2
r(t) = − 4 + aT t − 2
aT at
for T2 ≤ t ≤ T
⎪
⎩ aT 2
4 for t ≥ T

Taking a = 4y ∗ /T 2 ensures that r(t) = y ∗ for t ≥ T . In this case, there is an

important difference between the case with no zero dynamics and the one with
asymptotically stable zero dynamics. When there are no zero dynamics, the input
u reaches φu (y ∗ ) during the finite period T . When there are zero dynamics, the η
variables will continue to move beyond the period T . In this case η(t) and u(t) will
asymptotically approach their equilibrium values φη (y ∗ ) and φu (y ∗ ).
The freedom in choosing the reference signal r to move y from zero to y ∗ can
be used to deal with control constraints.2 It is clear that aggressively moving the
2 See [24, 50, 64] for techniques that take control constraints into consideration.
13.3. TRANSITION BETWEEN SET POINTS 317

r(2)
0 T/2 T
−a

r(1) at a(T−t)

aT 2/4

r
at2/2 −aT 2/4 + aT t − at2/2

Figure 13.3: Trajectory planning for ρ = 2.

output over a short period of time will require large control effort. By appropriately
choosing the reference signal we can avoid control saturation that might deteriorate
the response. The next example illustrates this point.

Example 13.3 Reconsider the pendulum equation of Example 13.1 with the nom-
inal parameters b = 0.03 and c = 1. Suppose the pendulum is resting at the
open-loop equilibrium point x = 0 and we want to move it to a new equilibrium
point at x = col(π/2, 0). We take the reference signal r as the output of the second-
order transfer function 1/(τ s + 1)2 driven by a step input of amplitude π/2. The
tracking controller is

u = sin x1 + 0.03x2 + r̈ − 9e1 − 3e2

Taking the initial conditions of the reference model to be zero, we find that the
tracking error e(t) = x(t) − R(t) will be identically zero and the motion of the
pendulum will track the desired reference signal for all t. The choice of the time
constant τ determines the speed of motion from the initial to the final position. If
there were no constraint on the magnitude of the control u, we could have chosen τ
arbitrarily small and achieved arbitrarily fast transition from x1 = 0 to x1 = π/2.
However, the control input u is the torque of a motor and there is a maximum
torque that the motor can supply. This constraint puts a limit on how quick we can
move the pendulum. By choosing τ to be compatible with the torque constraint, we
can avoid control saturation. Figure 13.4 shows two different choices of τ when the
318 CHAPTER 13. TRACKING AND REGULATION

τ = 0.2 τ = 0.8
2 2

1.5 1.5
Output

Output
1 1
output
output
0.5 reference 0.5
reference

0 0
0 1 2 3 4 5 0 1 2 3 4 5

τ = 0.2 τ = 0.8

2 2

1 1
Control

Control

0 0

−1 −1
−2 −2
0 1 2 3 4 5 0 1 2 3 4 5
Time Time

Figure 13.4: Simulation of the tracking control of Example 13.3.

control is constrained to |u| ≤ 2. For τ = 0.2, the control saturates during the initial
transient causing the output y(t) to deviate from the reference r(t), which reflects
the fact that the reference signal demands a control effort that cannot be delivered
by the motor. On the other hand, with τ = 0.8, the output signal achieves good
tracking of the reference signal. In both cases, we could not achieve a settling time
better than about 4, but by choosing τ = 0.8, we were able to avoid the overshoot
that took place when τ = 0.2.

13.4 Robust Regulation via Integral Action

The system

η̇ = f0 (η, ξ, w) (13.13)
˙ξi = ξi+1 , 1≤i≤ρ−1 (13.14)
ξ˙ρ = a(η, ξ, w) + b(η, ξ, w)u (13.15)
y = ξ1 (13.16)
13.4. ROBUST REGULATION VIA INTEGRAL ACTION 319

is a perturbation of (13.1)–(13.4) where w ∈ Dw ⊂ R is a vector of unknown

constant parameters and disturbances defined over some domain Dw . We assume
that Assumptions 13.1 and 13.2 hold for all w ∈ Dw with b0 independent of w. Our
goal is to design a state feedback controller such that all state variables are bounded
and the output y is asymptotically regulated to a constant reference r ∈ Dr ⊂ R,
for some domain Dr . This is a special case of the tracking problem of Section 13.2,
where the reference is constant and the uncertainty is parameterized by w. The
sliding mode controller of Section 13.2 will ensure that the regulation error y − r
will be ultimately bounded by kμ for some k > 0. By using integral action, we can
ensure that the error will converge to zero as time tends to infinity. The controller
¯ that satisfies the equations
will drive the trajectory to an equilibrium point (η̄, ξ)
0 = ¯ w)
f0 (η̄, ξ,
0 = ξ¯i+1 , 1≤i≤ρ−1
0 ¯ ¯ w)ū
= a(η̄, ξ, w) + b(η̄, ξ,
r = ξ¯1
Hence, ξ¯ = col(r, 0, . . . , 0).
¯ w) has a
Assumption 13.5 For all (r, w) ∈ Dr × Dw , the equation 0 = f0 (η̄, ξ,
unique solution η̄ ∈ Dη , which we denote by η̄ = φη (r, w).
Because b = 0, the steady-state control needed to maintain this equilibrium point
is given by
¯ w) def
a(η̄, ξ,
ū = − ¯ w) = φu (r, w)
b(η̄, ξ,
Augmenting the integrator
ė0 = y − r (13.17)
with the system and applying the change of variables
⎡ ⎤ ⎡ ⎤
e1 ξ1 − r
⎢ e2 ⎥ ⎢ ξ2 ⎥
⎢ ⎥ ⎢ ⎥
z = η − η̄, e=⎢ . ⎥=⎢ .. ⎥
⎣ .. ⎦ ⎣ . ⎦
eρ ξρ
we obtain the augmented system
def
ż = f0 (z + η̄, ξ, w) = f˜0 (z, e, r, w)
ėi = ei+1 , for 0 ≤ i ≤ ρ − 1
ėρ = a(η, ξ, w) + b(η, ξ, w)u
which preserves the normal form structure with a chain of ρ + 1 integrators. There-
fore, the design of sliding mode control can proceed as in Section 13.2. We take
s = k0 e0 + k1 e1 + · · · + kρ−1 eρ−1 + eρ
320 CHAPTER 13. TRACKING AND REGULATION

where k0 to kρ−1 are chosen such that the polynomial

λρ + kρ−1 λρ−1 + · · · + k1 λ + k0

is Hurwitz. Then,


ρ−1
ṡ = ki ei+1 + a(η, ξ, w) + b(η, ξ, w)u
i=0

The control u can be taken as

'ρ−1 (
1 
u=v or u=− ki ei+1 + â(η, ξ) + v (13.18)
b̂(η, ξ) i=0

where â and b̂ are nominal models of a and b, respectively. Then,

ṡ = b(η, ξ, w)v + Δ(η, ξ, r, w)

If  
 Δ(η, ξ, r, w) 
 
 b(η, ξ, w)  ≤ (η, ξ)

for all (η, ξ, r, w) ∈ Dη × Dξ × Dr × Dw , with a known locally Lipschitz function ,

we take  
s
v = −β(η, ξ) sat (13.19)
μ
where β(η, ξ) is a locally Lipschitz function that satisfies β(η, ξ) ≥ (η, ξ) + β0 with
β0 > 0. The closed-loop system has an equilibrium point at (z, e0 , e) = (0, ē0 , 0).
Showing convergence to this equilibrium point can be done by analysis similar to
the analysis of Section 10.1, under the following two assumptions.

Assumption 13.6 For all (r, w) ∈ Dr ×Dw , there is a Lyapunov function V1 (z, r, w)
for the system ż = f˜0 (z, e, r, w) that satisfies the inequalities

α1 (z) ≤ V1 (z, r, w) ≤ α2 (z)

∂V1 ˜
f0 (z, e, r, w) ≤ −α3 (z), ∀ z ≥ α4 (e)
∂z
for some class K functions α1 to α4 .

Assumption 13.7 z = 0 is an exponentially stable equilibrium point of the system

ż = f˜0 (z, 0, r, w).

The convergence of the error to zero is stated in the following theorem whose
proof is given in Appendix D.
13.4. ROBUST REGULATION VIA INTEGRAL ACTION 321

Theorem 13.1 Suppose Assumptions 13.1, 13.2, 13.5, 13.6 and 13.7 are satisfied
for the system (13.13)–(13.16) and consider the controller (13.17)–(13.19). Then,
there are positive constants c, ρ1 and ρ2 and a positive definite matrix P such that
the set
Ω = {V1 (z) ≤ α2 (α4 (cρ2 )} × {ζ T P ζ ≤ ρ1 c2 } × {|s| ≤ c}
where ζ = col(e0 , e1 , . . . , eρ−1 ), is compact and positively invariant, and for all
initial states in Ω, limt→∞ |y(t) − r| = 0. 3

In the special case when β = k (a constant) and u = v, the controller is given

by  
k0 e0 + k1 e1 + · · · + kρ−1 eρ−1 + eρ
u = −k sat (13.20)
μ
When ρ = 1 the controller (13.20) is a classical PI (Proportional-Integral) controller
followed by saturation and when ρ = 2 it is a classical PID (Proportional-Integral-
Derivative) controller followed by saturation.

Example 13.4 Consider the pendulum equation (A.2) and suppose that the sus-
pension point is subjected to a constant horizontal acceleration. The state model
is given by

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + cu + d cos x1 , y = x1

It is desired to regulate y to a constant reference r. Suppose b, c, d are uncertain

parameters that satisfy 0 ≤ b ≤ 0.1, 0.5 ≤ c ≤ 2, and 0 ≤ d ≤ 0.5. Define
e1 = x1 − r, e2 = x2 , and augment the integrator ė0 = e1 with the system to obtain

ė0 = e1 , ė1 = e2 , ė2 = − sin x1 − bx2 + cu + d cos x1

Taking s = e0 + 2e1 + e2 , assigns the eigenvalues of λ2 + 2λ + 1 at −1, −1. The

derivative ṡ is given by

ṡ = e1 + (2 − b)e2 − sin x1 + cu + d cos x1

It can be verified that

 
 e1 + (2 − b)e2 − sin x1 + d cos x1  |e1 | + 2|e2 | + 1 + 0.5
 ≤ = 2|e1 | + 4|e2 | + 3
 c  0.5
We take β = 2|e1 | + 4|e2 | + 4, which yields the control
 
e0 + 2e1 + e2
u = −(2|e1 | + 4|e2 | + 4) sat
μ
For comparison, we design also a sliding mode controller without integral action,
where s = e1 + e2 and

ṡ = (1 − b)e2 − sin x1 + cu + d cos x1

322 CHAPTER 13. TRACKING AND REGULATION

2 1.6

1.58
1.5
Output

Output
1.56
1
1.54
0.5
1.52

0 1.5
0 2 4 6 8 10 9 9.2 9.4 9.6 9.8 10
Time Time

Figure 13.5: Simulation of the regulating controllers of Example 13.4 with (dashed)
and without (solid) integral action.

Using  
 (1 − b)e2 − sin x1 + d cos x1  |e2 | + 1 + 0.5
 ≤ = 2|e2 | + 3
 c  0.5
we choose β = 2|e2 | + 4, which yields the control
 
e1 + e2
u = −(2|e2 | + 4) sat
μ
Simulation results with μ = 0.1 and x(0) = 0, r = π/2, b = 0.03, c = 1, and d = 0.3
are shown in Figure 13.5. The solid curve is the response without integral action,
while the dashed one is the response under integral action. The controller without
integral action results in a steady-state error due to the non-vanishing disturbance
d cos x1 , while the one with integral action regulates the error to zero. The inclusion
of integral action comes at the expense of the transient response, which shows an
overshoot that is not present in the case without integral action.

13.5 Output Feedback

The tracking and regulation controllers of Sections 13.1 to 13.4 require measurement
of the whole state vector. In this section we derive output feedback controllers
when we only measure the output y. We use high-gain observers as in Section 12.4
and 12.5. The starting point is to design partial state feedback controllers that
use only the vector ξ, which is formed of the output y and its derivatives y (1) to
y (ρ−1) . Because feedback linearization requires full state, we limit our discussion to
the sliding mode controllers of Sections 13.2 and 13.4. The tracking sliding mode
controller (13.11) can be taken as
 
k1 e1 + · · · + kρ−1 eρ−1 + eρ
u = −β(ξ) sat (13.21)
μ
13.5. OUTPUT FEEDBACK 323

The regulation sliding mode controller (13.18) can be taken as

 
k0 e0 + k1 e1 + · · · + kρ−1 eρ−1 + eρ
u = −β(ξ) sat (13.22)
μ

where ė0 = e1 = y − r. The main restriction here is that β is allowed to depend

only on ξ rather than the full state vector. Because β is chosen to dominate a
function (η, ξ), this restriction may force us to limit the analysis to a compact set
over which the η-dependent part of (η, ξ) can be bounded by a constant. In the
output feedback case, we use the high-gain observer
αi
ê˙ i = êi+1 + (y − r − ê1 ), 1≤i≤ρ−1 (13.23)
εi
αρ
ê˙ ρ = (y − r − ê1 ) (13.24)
ερ
to estimate e by ê, where ε is a sufficiently small positive constant and α1 to αρ are
chosen such that the polynomial

λρ + α1 λρ−1 + · · · + αρ−1 λ + αρ (13.25)

is Hurwitz. The estimate ξˆ is given by ξˆ = ê + R. To overcome the peaking phe-

nomenon of high-gain observers, the control should be a globally bounded function
of ê. Because the saturation function is bounded, we only need to saturate β(ξ) ˆ
ˆ Then,
outside a compact set of interest. Denote the saturated function by βs (ξ).
the output feedback tracking controller is given by
 
ˆ k1 ê1 + · · · + kρ−1 êρ−1 + êρ
u = −βs (ξ) sat (13.26)
μ

while the output feedback regulation controller is given by

 
ˆ sat k0 e0 + k1 ê1 + · · · + kρ−1 êρ−1 + êρ
u = −βs (ξ) (13.27)
μ

Note that the integrator state e0 is available because it is obtained by integrating

the regulation error e1 = y − r. In both (13.26) and (13.27), we can replace ê1
by the measured signal e1 . In the special case when βs of (13.26) is constant or
function of ê rather than ξ,ˆ we do not need the derivatives of r, as required by
Assumption 13.4, because ê is provided by the observer (13.23)–(13.24), which is
driven by y − r; so only r is needed.
Similar to the results of Sections 12.4 and 12.5, it can be shown that the output
feedback controllers (13.26) and (13.27) recover the performance of the partial state
feedback controllers (13.21) and (13.22), respectively, for sufficiently small ε. In the
regulation case, it can be shown that the regulation error converges to zero.3
3 The proof for a slightly modified problem formulation can be found in [73].
324 CHAPTER 13. TRACKING AND REGULATION

For relative degree one systems, the controllers (13.21) and (13.22) are given by
   
y−r k0 e0 + y − r
u = −β(y) sat and u = −β(y) sat
μ μ

respectively, where ė0 = y − r. These controllers depend only on the measured

output y; hence, no observer is needed.

Example 13.5 We use the high-gain observer

2 1
ê˙ 1 = ê2 + (e1 − ê1 ), ê˙ 2 = 2 (e1 − ê1 )
ε ε
to implement the tracking controller
 
e1 + e2
u = −(2|e2 | + 3) sat
μ

of Example 13.2 and the regulating controller

 
e0 + 2e1 + e2
u = −(2|e1 | + 4|e2 | + 4) sat
μ

of Example 13.4. We use ê2 from the observer to replace e2 but keep e1 since it
is a measured signal. To overcome the observer peaking, we saturate |ê2 | in the
β function over a compact set of interest. There is no need to saturate ê2 inside
the saturation because the saturation function is globally bounded. We use the
analysis of the state feedback controller to determine the compact set of interest.
For the tracking controller of Example 13.2, the analysis of Section 13.2 shows that
Ω = {|e1 | ≤ c/θ} × {|s| ≤ c}, with c > 0 and 0 < θ < 1, is positively invariant.
Taking c = 2 and 1/θ = 1.1, we obtain Ω = {|e1 | ≤ 2.2} × {|s| ≤ 2}. Over Ω,
|e2 | ≤ |e1 | + |s| ≤ 4.2. We saturate |ê2 | at 4.5, which results in the output feedback
controller      
|ê2 | e1 + ê2
u = − 2 × 4.5 sat + 3 sat
4.5 μ
For the regulating controller of Example 13.4, ζ = col(e0 , e1 ) satisfies the equation

0 1 0
ζ̇ = Aζ + Bs, where A = and B =
−1 −2 1

From the proof of Theorem 13.1 in Appendix D, it can be seen that Ω = {ζ T P ζ ≤

ρ1 c2 } × {|s| ≤ c} is positively invariant, where c > 0, P is the solution of the
Lyaunov equation P A + AT P = −I, ρ1 = λmax (P )(2P B/θ)2 , and 0 < θ < 1.
Taking c = 4 and 1/θ = 1.003, we obtain Ω = {ζ T√ P ζ ≤ 55} × {|s| ≤ 4}. Using
(B.4), it can be shown that over Ω, |e0 + 2e1 | ≤ 55  1 2 P −1/2  = 22.25;
13.6. EXERCISES 325

(a) (b)
1.6 2

1.4
1.5
1.2
Output

Output
1
1

0.8 0.5
0.6
0
0 1 2 3 4 0 2 4 6
Time Time

Figure 13.6: Simulation of the output feedback controllers of Example 13.5, comparing
the response under state feedback (solid) with the one under output feedback when
ε = 0.05 (dashed) and ε = 0.01 (dash-dot).

consequently, |e2 | ≤ |e0 + 2e1 | + |s| ≤ 26.25. We saturate |ê2 | at 27, which results
in the output feedback controller
     
|ê2 | e0 + 2e1 + ê2
u = − 2|e1 | + 4 × 27 sat + 4 sat
27 μ

The simulation results of Figure 13.6 use the same data of Examples 13.2 and 13.4
with two different values of ε, 0.05 and 0.01. Figure 13.6(a) is for the tracking
controller of Example 13.2 and Figure 13.6(b) for the regulating controller of Ex-
ample 13.4. In both cases, the solid curve is for state feedback and the other two
curves are for output feedback with ε = 0.05 (dashed curve) and ε = 0.01 (dash-dot
curve). As we have seen before with high-gain observers, the response under output
feedback approaches the one under state feedback as the ε decreases.

13.6 Exercises
13.1 Consider a tracking problem for the magnetic levitation system (A.30)–
(A.32) where y = x1 , r(t) = 0.5 sin ωt, α = 1.5, β = 2, b = 0, and c = 1.

(a) Design a state feedback controller using feedback linearization. Using simula-
tion with ω = 1, choose the controller parameters to have settling time less
than 20 time units.

(b) Design an observer to implement the controller of part (a) assuming you mea-
sure x1 and x3 . Use simulation, with ω = 1, x(0) = col(1, 0, 1), and x̂(0) = 0,
to compare the performance of the state and output feedback controllers.
326 CHAPTER 13. TRACKING AND REGULATION

(c) Repeat (b) assuming you can only measure x1 .

13.2 Repeat the previous exercise when the reference signal is taken as r(t) =
r0 + (r1 − r0 )q(t), where q (3) (t) is defined by
⎧
⎪
⎪ 1/2 for 0 ≤ t < 1
⎨
(3) −1/2 for 1 ≤ t < 3
q (t) =
⎪
⎪ 1/2 for 3 ≤ t < 4
⎩
0 for t ≥ 4

and q(0) = q̇(0) = q̈(0) = 0, which is chosen to steer the system from equilibrium
at x1 = r0 to equilibrium at x1 = r1 over four time units. In the simulation take
r0 = 1, r1 = 2, and x(0) = col(1, 0, 1).

13.3 Reconsider the previous exercise when c ∈ [0.5, 1.5] is uncertain and the
only available signal is the tracking error x1 − r. Using sliding mode control and
a high-gain observer, design a locally Lipschitz feedback controller. What is the
steady-state tracking error? Simulate the response for different values of c. Hint:
The controller can take the form u = K sat((K1 ê1 + K2 ê2 + ê3 )/μ) where ê1 , ê2 ,
and ê3 are estimates of the tracking error e1 = x1 − r and its derivatives e2 = ė1
and e3 = ë1 , respectively.

13.4 Repeat the previous exercise using integral action.

13.5 Consider a tracking problem for the electrostatic microactuator (A.33) where
r(t) = r0 + (r1 − r0 )q(t) and q(t) is the output of the transfer function 1/(s + 1)3
with a unit step input. The reference r(t) is chosen to steer x1 from equilibrium at
r0 > 0 to equilibrium at r1 > 0.
(a) Design a locally Lipschitz state feedback controller using feedback linearization.
√ system with ζ = 0.1, T = 0.2, r0 = 0.1, r1 = 0.5,
Simulate the closed-loop
and x(0) = col(0.1, 0, 0.3).
(b) Design an observer to implement the controller of part (a) assuming you mea-
sure x1 and x3 . Use simulation to compare the performance of the state and
output feedback controllers.
(c) Repeat (b) assuming you can only measure x1 .

13.6 Repeat the previous exercise using sliding mode control when ζ ∈ (0, 0.5] and
T ∈ [0.1, 0.5] are uncertain parameters. What is the steady-state tracking error?

13.7 Repeat the previous exercise using integral action.

13.8 Consider the robot manipulator (A.34) and let M̂ , Ĉ and ĝ be nominal
models of M , C, and g, respectively. We want to design a feedback controller such
that q(t) asymptotically tracks a reference trajectory qr (t), where qr (t), q̇r (t), and
q̈r (t) are continuous and bounded.
13.6. EXERCISES 327

(a) Design a locally Lipschitz state feedback controller to achieve global practical
tracking.
(b) Assuming you measure q but not q̇, design a locally Lipschitz output feedback
controller to achieve semiglobal practical tracking.
Hint: Assume appropriate bounds on the model uncertainty.

13.9 The tank system of Exercise 1.9 is modeled by

  
ḣ = u − k ρgh /A(h)
√
Assume that A(h) ≤ c1 and k ρg ≤ c2 for known positive constants c1 and c2 .
Using sliding mode control, design a locally Lipschitz output feedback controller
that regulates h to r with zero steady-state error.

13.10 Consider the DC motor(A.24) with ϑ = 0, nonlinear friction η(ω) = k1 ω +

k2 sgn(ω), and unknown positive constants a, k1 , and k2 . Assume that a ∈ [0.5, 1.5],
k1 ≤ 0.1 and k2 ≤ 0.2. Design a locally Lipschitz output feedback controller to
regulate ω to ω ∗ with zero steady-state error.

13.11 A normalized model of a pendulum whose suspension point is subjected to

a constant horizontal acceleration is given by

ẋ1 = x2 , ẋ2 = − sin x1 − bx2 + cu + h

where 0 ≤ b ≤ 0.4, 0.5 ≤ c ≤ 1.5, and |h| ≤ 1.

(a) Design a locally Lipschitz state feedback controller to regulate x1 to zero with
zero steady-state error.
(b) Repeat (a) if you can only measure x1 .

13.12 Consider the magnetic levitation system (A.29) with uncertain parameters
b ∈ [0, 0.1] and c ∈ [0.8, 1.2]. It is required to regulate x1 to r ∈ [0.5, 1.5] with zero
steady-state error, while meeting the constraint −2 ≤ u ≤ 0.
(a) Design a locally Lipschitz state feedback controller.
(b) Repeat (a) if you can only measure x1 .

13.13 Consider the system ÿ = f (y, ẏ) + u, where f is unknown, locally Lipschitz
function with f (0, 0) = 0, and the control u is constrained by |u| ≤ U for a given
constant U . Suppose the system is initially at the equilibrium point (0, 0) and we
want y to track a reference signal r(t) with the properties: (1) r(t) has bounded
continuous derivatives up to the second order; (2) maxt≥0 |r̈ − f (r, ṙ)| < U ; (3)
r(0) = ṙ(0) = 0.
328 CHAPTER 13. TRACKING AND REGULATION

(a) If the measurements of y, ẏ, r, ṙ, and r̈ are available, design a locally Lipschitz
feedback controller such that |y(t) − r(t)| ≤ δ ∀ t ≥ 0 for a given δ.

(b) Repeat (a) if only the measurements of y and r are available.

13.14 A simplified model of an underwater vehicle in yaw is given by

ψ̈ + aψ̇|ψ̇| = u

where ψ is the heading angle, u is a normalized torque, and a is a positive parameter.

To move ψ from set point 0 to set point ψ ∗ , a reference signal is taken as ψr (t) =
ψ ∗ [1 − e−t (1 + t)].

(a) With a = 1, use feedback linearization to design a state feedback controller

such that ψ(t) ≡ ψr (t) for all t ≥ 0.
(b) Suppose a = 1. Find an upper bound on |a − 1| such that the tracking error
under the controller of part (a) converges to zero as t tends to infinity.

(c) Assuming |a − 1| ≤ 0.5, design a locally Lipschitz state feedback controller to

achieve global asymptotic tracking.

(d) Assuming |a − 1| ≤ 0.5 and you can only measure ψ, design a locally Lipschitz
output feedback controller to achieve semiglobal asymptotic tracking.
Appendix A

Examples

A.1 Pendulum
Consider the simple pendulum shown in Figure A.1, where l denotes the length of
the rod and m the mass of the bob. Assume the rod is rigid and has zero mass.
Let θ denote the angle subtended by the rod and the vertical axis through the
pivot point. The pendulum swings in the vertical plane. The bob of the pendulum
moves in a circle of radius l. To write the equation of motion of the pendulum, let
us identify the forces acting on it. There is a downward gravitational force equal
to mg, where g is the acceleration due to gravity. There is also a frictional force
resisting the motion, which we assume to be proportional to the speed of the bob
with a coefficient of friction k. Suppose also that a torque T is applied at the pivot
point in the direction of θ. By taking moments about the pivot point, we obtain

d2 θ dθ
ml2 2
+ mgl sin θ + kl2 =T
dt dt

mg

Figure A.1: Pendulum.

329
330 APPENDIX A. EXAMPLES

When T = 0, the equilibrium points are obtained by setting dθ/dt = d2 θ/dt2 = 0,

which shows that the pendulum has infinitely many equilibrium points at (θ, dθ/dt) =
(nπ, 0), for n = 0, ±1, ±2, . . .. From the physical description of the pendulum, it
is clear that the pendulum has only two equilibrium positions corresponding to
the equilibrium points (0, 0) and (π, 0). Other equilibrium points are repetitions of
these two positions, which correspond to the number of full swings the pendulum
would make before it rests at one of the two equilibrium positions. For example, if
the pendulum makes m complete 360◦ revolutions before it rests at the downward
vertical position, then, mathematically, we say that the pendulum approaches the
equilibrium point (2mπ, 0). In our investigation of the pendulum, we will limit our
attention to the two “nontrivial” equilibrium points at (0, 0) and (π, 0). Physically,
we can see that these two equilibrium points are quite distinct from each other.
While the pendulum can indeed rest at (0, 0), it can hardly maintain rest at (π, 0)
because infinitesimally small disturbance from that equilibrium will take the pen-
dulum away. The difference between the two equilibrium points is in their stability
properties.
We normalize the pendulum equation so that all variables are dimensionless.
By linearization at (0, 0) it can be seen that when there is no friction (k = 0) and
no applied torque (T = 0) the pendulum oscillates around the downward vertical
position with frequency Ω where Ω2 = g/l. We define the dimensionless time τ = Ωt.
Then
dθ dθ d2 θ d2 θ
=Ω and 2
= Ω2 2
dt dτ dt dτ
Defining the dimensionless torque u = T /(mo gl) for some nominal mass mo and
˙ and (·),
denoting the first and second derivatives with respect to τ by (·) ¨ respectively,
we obtain the normalized equation
θ̈ + sin θ + b θ̇ = c u (A.1)
where b = k/(mΩ) and c = mo /m. In our study of the pendulum equation we
sometimes deal with the mass m as uncertain parameter, which results in uncer-
tainty in the coefficient b and c of (A.1). To obtain a state model of the pendulum,
we take the state variables as x1 = θ and x2 = θ̇, which yields
ẋ1 = x2 , ẋ2 = − sin x1 − b x2 + c u (A.2)
When u = 0 we have the unforced state equation
ẋ1 = x2 , ẋ2 = − sin x1 − b x2 (A.3)
Neglecting friction by setting b = 0 results in the system
ẋ1 = x2 , ẋ2 = − sin x1 (A.4)
which is conservative in the sense that if the pendulum is given an initial push, it
will keep oscillating forever with a nondissipative energy exchange between kinetic
and potential energies. This, of course, is not realistic, but gives insight into the
behavior of the pendulum.
A.2. MASS–SPRING SYSTEM 331

Fsp Ff

F
m
. y

Figure A.2: Mass-spring mechanical system

A.2 Mass–Spring System

In the mass–spring mechanical system, shown in Figure A.2, we consider a mass m
sliding on a horizontal surface and attached to a vertical surface through a spring.
The mass is subjected to an external force F . We define y as the displacement from
a reference position and write Newton’s law of motion

mÿ + Ff + Fsp = F

where Ff is a resistive force due to friction and Fsp is the restoring force of the
spring. We assume that Fsp is a function only of the displacement y and write it as
Fsp = g(y). We assume also that the reference position has been chosen such that
g(0) = 0. The external force F is at our disposal. Depending upon F , Ff , and g,
several interesting time-invariant and time-varying models arise.
For a relatively small displacement, the restoring force of the spring can be
modeled as a linear function g(y) = ky, where k is the spring constant. For a
large displacement, however, the restoring force may depend nonlinearly on y. For
example, the function

g(y) = k(1 − a2 y 2 )y, |ay| < 1

models the so-called softening spring, where, beyond a certain displacement, a large
displacement increment produces a small force increment. On the other hand, the
function
g(y) = k(1 + a2 y 2 )y
models the so-called hardening spring, where, beyond a certain displacement, a
small displacement increment produces a large force increment.
The resistive force Ff may have components due to static, Coulomb, and viscous
friction. When the mass is at rest, there is a static friction force Fs that acts parallel
to the surface and is limited to ±μs mg, where 0 < μs < 1 is the static friction
coefficient. This force takes whatever value, between its limits, to keep the mass at
rest. For motion to begin, there must be a force acting on the mass to overcome
the static friction. In the absence of an external force, F = 0, the static friction
332 APPENDIX A. EXAMPLES

Ff Ff

v v

(a) (b)

Ff Ff

v v

(c) (d)

Figure A.3: Examples of friction models. (a) Coulomb friction; (b) Coulomb plus linear
viscous friction; (c) static, Coulomb, and linear viscous friction; (d) static, Coulomb,
and linear viscous friction—Stribeck effect.

force will balance the restoring force of the spring and maintain equilibrium for
|g(y)| ≤ μs mg. Once motion has started, the resistive force Ff , which acts in the
direction opposite to motion, is modeled as a function of the sliding velocity v = ẏ.
The resistive force due to Coulomb friction Fc has a constant magnitude μk mg,
where μk is the kinetic friction coefficient, that is,

−μk mg, for v < 0
Fc =
μk mg, for v > 0

As the mass moves in a viscous medium, such as air or lubricant, there will be
a frictional force due to viscosity. This force is usually modeled as a nonlinear
function of the velocity; that is, Fv = h(v), where h(0) = 0. For small velocity, we
can assume that Fv = cv. Figure A.3 shows various examples of friction models. In
Figure A.3(c), the static friction is higher than the level of Coulomb friction, while
Figure A.3(d) shows a similar situation, but with the force decreasing continuously
with increasing velocity, the so-called Stribeck effect.
The combination of a hardening spring, linear viscous friction, and a periodic
external force F = A cos ωt results in the Duffing’s equation

mÿ + cẏ + ky + ka2 y 3 = A cos ωt (A.5)

A.3. TUNNEL-DIODE CIRCUIT 333

which is a classical example in the study of periodic excitation of nonlinear systems.

The combination of a linear spring, static friction, Coulomb friction, linear vis-
cous friction, and external force F results in

mÿ + ky + cẏ + η(y, ẏ) = F

where
⎧
⎨ μk mg sign(ẏ), for |ẏ| > 0
η(y, ẏ) = −ky, for ẏ = 0 and |y| ≤ μs mg/k
⎩
−μs mg sign(y), for ẏ = 0 and |y| > μs mg/k

The value of η(y, ẏ) for ẏ = 0 and |y| ≤ μs mg/k is obtained from the equilibrium
condition ÿ = ẏ = 0. With x1 = y, x2 = ẏ, and u = F , the state model is

ẋ1 = x2 , ẋ2 = [−kx1 − cx2 − η(x1 , x2 ) + u]/m (A.6)

Let us note two features of this state model. First, with u = 0 it has an equilibrium
set, rather than isolated equilibrium points. Second, the right-hand side is a discon-
tinuous function of the state. The discontinuity is a consequence of the idealization
we adopted in modeling friction. One would expect the transition from static to
sliding friction to take place in a smooth way, not abruptly as our idealization sug-
gests.1 The discontinuous idealization, however, allows us to carry out piecewise
linear analysis since in each of the regions {x2 > 0} and {x2 < 0}, we can use the
model
ẋ1 = x2 , ẋ2 = [−kx1 − cx2 − μk mg sign(x2 ) + u]/m
to predict the behavior of the system via linear analysis.

A.3 Tunnel-Diode Circuit

Consider the tunnel-diode circuit shown in Figure A.4, where the diode is char-
acterized by iR = h(vR ) [25]. The energy-storing elements in this circuit are the
capacitor C and the inductor L. Assuming they are linear and time invariant, we
can model them by the equations

dvC diL
iC = C and vL = L
dt dt
where i and v are the current through and the voltage across an element, with
the subscript specifying the element. We take x1 = vC and x2 = iL as the state
variables and u = E as a constant input. To write the state equation for x1 , we
1 The smooth transition from static to sliding friction can be captured by dynamic friction
models; see, for example, [6] and [100].
334 APPENDIX A. EXAMPLES

iL L C
i,mA
+ vL −
1
iC iR
i = h(v)
R 0.5

+ +
vR
vC C 0
−
−
E
−0.5
0 0.5 1 v,V
(a) (b)

Figure A.4: (a) Tunnel-diode circuit; (b) Tunnel-diode vR –iR characteristic.

need to express iC as a function of the state variables x1 , x2 and the input u. Using
Kirchhoff’s current law at node , c we obtain

iC + iR − iL = 0

Therefore, iC = −h(x1 ) + x2 . Similarly, we need to express vL as a function of x1 ,

x2 , and u. Using Kirchhoff’s voltage law in the left loop, we obtain

vC − E + RiL + vL = 0

Hence, vL = −x1 − Rx2 + u. We can now write the state model for the circuit as
1 1
ẋ1 = [−h(x1 ) + x2 ] , ẋ2 = [−x1 − Rx2 + u] (A.7)
C L
The equilibrium points of the system are determined by setting ẋ1 = ẋ2 = 0 and
solving for x1 and x2 :

0 = −h(x1 ) + x2 , 0 = −x1 − Rx2 + u

Therefore, the equilibrium points correspond to the solutions of the equation

h(x1 ) = (E − x1 )/R

Figure A.5 shows graphically that, for certain values of E and R, this equation has
three isolated solutions which correspond to three isolated equilibrium points of the
system. The number of equilibrium points might change as the values of E and R
change. For example, if we increase E for the same R, we will reach a point beyond
which only the point Q3 will exist. On the other hand, if we decrease E for the
same R, we will end up with the point Q1 as the only equilibrium.
A.4. NEGATIVE-RESISTANCE OSCILLATOR 335

iR

1.2
1
0.8
Q1 Q2
0.6
0.4
Q3
0.2
0
0 0.5 1 vR

Figure A.5: Equilibrium points of the tunnel-diode circuit.

A.4 Negative-Resistance Oscillator

Figure A.6 shows the basic circuit of an important class of electronic oscillators [25].
The inductor and capacitor are assumed to be linear, time invariant and passive,
that is, L > 0 and C > 0. The resistive element is an active circuit characterized
by the v–i characteristic i = h(v), shown in the figure. The function h(·) satisfies
the conditions

h(0) = 0, h (0) < 0, h(v) sign(v) → ∞ as |v| → ∞

where h (v) is the derivative of h(v) with respect to v. By Kirchhoff’s current law,

iC + iL + i = 0

Hence,  t
dv 1
C + v(s) ds + h(v) = 0
dt L −∞

Differentiating once with respect to t and multiplying through by L, we obtain

d2 v dv
CL 2
+ v + Lh (v) =0
dt dt
The foregoing equation can be written in a form that coincides with some well-
known equations
√ in nonlinear systems theory by changing the time variable from t
to τ = t/ CL. The derivatives of v with respect to t and τ are related by

dv √ dv d2 v d2 v
= CL and = CL
dτ dt dτ 2 dt2
Denoting the derivative of v with respect to τ by v̇, we can rewrite the circuit
equation as
v̈ + εh (v)v̇ + v = 0
336 APPENDIX A. EXAMPLES

i i = h(v)
+

C L
Resistive
v Element v
iC iL

(a) (b)

Figure A.6: (a) Basic oscillator circuit; (b) Typical driving-point characteristic.


where ε = L/C. This equation is a special case of Liénard’s equation

v̈ + f (v)v̇ + g(v) = 0 (A.8)

When h(v) = −v + 13 v 3 , the circuit equation takes the form

v̈ − ε(1 − v 2 )v̇ + v = 0 (A.9)

which is known as the Van der Pol equation. This equation, which was used by
Van der Pol to study oscillations in vacuum tube circuits, is a fundamental example
in nonlinear oscillation theory. It possesses a periodic solution that attracts every
other solution except the zero solution at the unique equilibrium point v = v̇ = 0.
To write a state model for the circuit, let us take x1 = v and x2 = v̇ to obtain

ẋ1 = x2 , ẋ2 = −x1 − εh (x1 )x2 (A.10)

Alternatively, if we take the state variables as z1 = iL and z2 = vc , we obtain the

state model
ż1 = z2 /ε, ż2 = ε[−z1 − h(z2 )] (A.11)
The state models in the x and z coordinates look different, but they are equivalent
representations of the system. This equivalence can be seen by noting that these
models can be obtained from each other by a change of coordinates z = T (x). Since
we have chosen both x and z in terms of physical variables, it is not hard to find
the map T . We have

x1 = v = z2

dv √ dv L
x2 = = CL = [−iL − h(vC )] = ε[−z1 − h(z2 )]
dτ dt C
A.5. DC-TO-DC POWER CONVERTER 337

Thus,

−h(x1 ) − x2 /ε −1 z2
z = T (x) = and x=T (z) =
x1 −εz1 − εh(z2 )

If a current source with current u is connected in parallel with the circuit, we arrive
at the forced equation

ż1 = z2 /ε, ż2 = ε[−z1 − h(z2 ) + u] (A.12)

a special case of which is the forced van der Pol equation

 
ż1 = z2 /ε, ż2 = ε −z1 + z2 − 13 z23 + u (A.13)

A.5 DC-to-DC Power Converter

Figure A.7 shows the circuit of a DC-to-DC switched power converter, known as the
boost converter, with ideal switch and passive linear components having capacitance
C, inductance L, and resistance R [128]. When the switch is in the s = 1 position,
Kirchoff’s laws yield the equations

diL dvC vC
L = E, C =−
dt dt R
while in the s = 0 position the equations are

diL dvC vC
L = −vC + E, C = iL −
dt dt R
where iL is the current through the inductor, vC the voltage across the capacitor,
and E the constant voltage of the voltage source. The foregoing equations can be
written as
diL dvC vC
L = −(1 − s)vC + E, C = (1 − s)iL − (A.14)
dt dt R
where s is a discrete variable that takes the values 0 or 1. When the switching
variable s(t) is a high-frequency square waveform of period T and duty ratio μ ∈
(0, 1); that is, ⎧
⎨ 1 for tk ≤ t < tk + μT
s(t) =
⎩
0 for tk + μT ≤ t < tk + T
averaging can be used to approximate (A.14) by the equations [11, Chapter 2]

diL dvC vC
L = −(1 − μ)vC + E, C = (1 − μ)iL − (A.15)
dt dt R
338 APPENDIX A. EXAMPLES

iL L 0

1 s
+
E vC C R
−

Figure A.7: Boost converter

where μ is a dimensionless continuous control input that satisfies 0 < μ(t) < 1. The
control problem is to regulate vc to a desired voltage Vd with DC gain k = Vd /E > 1.
This is achieved at the equilibrium point

iL = īL = Vd2 /(ER), vc = Vd , μ = μ̄ = 1 − E/Vd

1

Set α = R L/C. With the dimensionless variables

1 L vC − V d t
x1 = (iL − īL ), x2 = , u = μ − μ̄, τ=√
E C E LC
the state model is given by
1 1
ẋ1 = − x2 + (x2 + k)u, ẋ2 = x1 − αx2 − (x1 + αk 2 )u (A.16)
k k
˙ denotes derivative with respect to τ .
where (·)

A.6 Biochemical Reactor

Consider a biochemical reactor with two components—biomass and substrate—
where the biomass cells consume the substrate [16, 57]; a schematic is shown in
Figure A.8. Assume the reactor is perfectly mixed and the volume V is constant. Let
X and S be the concentrations (mass/volume) of the biomass cells and substrate,
respectively, Sf the concentration of the substrate in the feed stream, and F the flow
rate (volume/time). Assume there is no biomass in the feed stream. Let μ(S)X be
the rate of biomass cell generation (mass/volume/time) and μ(S)X/Y the rate of
the substrate consumption, where μ(S) is the specific growth rate and Y the yield.
The dynamic model is developed by writing material balances on the biomass and
substrate; that is,

rate of biomass accumulation = generation − out by flow

rate of substrate accumulation = in by flow − out by flow − consumption

A.6. BIOCHEMICAL REACTOR 339

F
SF

X
S F
X
S
V

Figure A.8: Biochemical reactor.

Since V is constant, the foregoing equations result in the model

dX XF dS μ(S)X (SF − S)F

= μ(S)X − , =− + (A.17)
dt V dt Y V

The control variable is the dilution rate F/V while SF is constant. A typical
model of the specific growth rate is μ(S) = μmax S/(km + S + k1 S 2 ), where μmax ,
km , and k1 are positive constants. When F is constant, equation (A.17) has a trivial
equilibrium point at (X = 0, S = SF ). If F ≤ maxS≥0 μ(s), there will be two other
(nontrivial) equilibrium points, which satisfy the steady-state conditions

Xs = (SF − Ss )Y, μ(Ss ) = Fs /V (A.18)

It can be seen by linearization that the equilibrium point at which μ (Ss ) > 0 is
asymptotically stable while the one with μ (Ss ) < 0 is unstable. The two equilibrium
points coincide when μ (Ss ) = 0. The system is regulated to operate at one of these
nontrivial equilibrium points.
Define the dimensionless state, control, and time variables by

X S F tF0
x1 = , x2 = , u= , τ=
X0 S0 F0 V

where X0 , S0 , and F0 are nominal steady-state quantities that satisfy (A.18) for a
nominal specific growth rate μ0 (S). The normalized state model is given by

ẋ1 = x1 [ν(x2 ) − u], ẋ2 = −(α − 1)x1 ν(x2 ) + (α − x2 )u (A.19)

where α = SF /S0 > 1 and ν(x2 ) = μ(x2 S0 )/μ0 (S0 ). In the nominal case where
μmax , km and k1 are known, ν(x2 ) is given by
x2
ν(x2 ) = (A.20)
β + γx2 + (1 − β − γ)x22

where β = km /(km + S0 + k1 S02 ) and γ = S0 /(km + S0 + k1 S02 ).

340 APPENDIX A. EXAMPLES

A.7 DC Motor
The equations of motion of the separately excited DC motor are given by [84]

dia
Ta = −ia + va − φω (A.21)
dt
dφ
Tf = −fe (φ) + vf (A.22)
dt
dω
Tm = ia φ − f (ω) − δ(t) (A.23)
dt

where ia , va , φ, vf , and ω are normalized (dimensionless) armature current, ar-

mature voltage, magnetic flux, field voltage, and speed, respectively, fe (φ) is the
inverse normalized magnetization curve, f (ω) + δ(t) is a normalized load torque,
which consists of a speed-dependent term f (ω) and a disturbance term δ(t), and
Ta , Tf and Tm are time constants.
The most common form to control a DC motor is the armature control method
where φ is constant and va is the control input. In this case, the model reduces to
equations (A.21) and (A.23). Further simplification of this model results from the
fact that Ta is typically much smaller than Tm . Neglecting the armature circuit
dynamics by setting Ta = 0 yields ia = va − φω, which upon substitution in (A.23)
results in the equation

dω
Tm = −φ2 ω + φva − f (ω) − δ(t)
dt

Setting u = va /φ, η(ω) = f (ω)/φ2 , ϑ(t) = δ(t)/φ2 , and defining the dimensionless
time τ = tφ2 /Tmo where Tmo is a nominal value of Tm , we obtain the normalized
model
ω̇ = a[−ω + u − η(ω) − ϑ(t)] (A.24)

where a = Tmo /Tm , and ω̇ = dω/dτ .

In a field-controlled DC motor, vf is the control input while va is held constant
at va = Va . Taking x1 = ia , x2 = φ, and x3 = ω as the state variables, u = vf as
the control input, and τ = t/Tmo as dimensionless time, the state equation is given
by
ẋ = f (x) + gu + hδ (A.25)

where ẋ = dx/dτ , f = Dα, g = Dβ, h = Dγ,

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
Tmo /Ta 0 0 −x1 − x2 x3 + Va 0 0
D=⎣ 0 Tmo /Tf 0 ⎦, α = ⎣ −fe (x2 ) ⎦ , β = ⎣1⎦ , γ = ⎣0⎦
0 0 Tmo /Tm x1 x2 − f (x3 ) 0 1
A.8. MAGNETIC LEVITATION 341

Controller

m
Ligh
source

Figure A.9: Magnetic levitation system.

A.8 Magnetic Levitation

Figure A.9 shows a schematic diagram of a magnetic levitation system, where a ball
of magnetic material is suspended by means of an electromagnet whose current is
controlled by feedback from the, optically measured, ball position [149, pp. 192–
200]. This system has the basic ingredients of systems constructed to levitate mass,
used in gyroscopes, accelerometers, and fast trains. The equation of motion of the
ball is
d2 y dy
m 2 = −k + mg + F (y, i) (A.26)
dt dt
where m is the mass of the ball, y ≥ 0 its vertical downward position (y = 0 when
the ball is next to the coil), k a viscous friction coefficient, g the acceleration due to
gravity, F (y, i) the force generated by the electromagnet, and i its electric current.
The inductance of the electromagnet depends on the position of the ball and can
be modeled as  
1
L(y) = L0 α +
1 + y/a
where a, α, and L0 are positive constants. This model represents the case that the
inductance has its highest value when the ball is next to the coil and decreases to
a constant value as the ball is removed to y = ∞. With E(y, i) = 12 L(y)i2 as the
energy stored in the electromagnet, the force F (y, i) is given by

∂E L0 i2
F (y, i) = =− (A.27)
∂y 2a(1 + y/a)2

From Kirchhoff’s voltage law, the electric circuit of the coil is modeled by
dφ
= −Ri + v (A.28)
dt
where φ = L(y)i is the magnetic flux linkage, v the applied voltage, and R the series
resistance of the circuit.
342 APPENDIX A. EXAMPLES

In what follows we develop two normalized models of the system: a two dimen-
sional model in which the current i is treated as the control input and a three-
dimensional model in which the voltage v is the control input. Treating the current
as the control input is valid when the time constant of the electric circuit is much
smaller than the time constant of the mechanical motion, or when an inner high-
gain feedback loop is closed around the circuit with feedback from the current. In
current control, we can substitute F from (A.27) into (A.26). However, the result-
ing equation will depend nonlinearly on i. It is better to treat F as the control
input and then solve (A.27) for i.
For the current-controlled system, we take the dimensionless time, state, and
control variables as
y 1 dy F
τ = λt, x1 = , x2 = , u=
a aλ dt mo g
where λ2 = g/a and mo is a nominal mass. The state model is given by
ẋ1 = x2 , ẋ2 = −bx2 + 1 + cu (A.29)
where ẋi denotes the derivative of xi with respect to τ , b = k/(λm), and c = mo /m.
By definition, x1 ≥ 0 and u ≤ 0. In some of our investigations we will also consider
the constraint |F | ≤ Fmax , which will put a lower bound on u. For convenience, we
will take Fmax = 2mo g so that −2 ≤ u ≤ 0.
For the voltage-controlled system, the current i is a state variable that satisfies
the equation
di L0 i dy
L(y) = −Ri + +v
dt a(+y/a)2 dt
which is obtained from (A.28) upon differentiating φ. With τ , x1 , x2 , b, and c as
defined earlier, let x3 = i/I and u = v/V , where the base current and voltage I
and V satisfy 8amo g = L0 I 2 and V = RI. The state model is given by
ẋ1 = x2 (A.30)
4cx23
ẋ2 = −bx2 + 1 − (A.31)
(1 + x1 )2

1 βx2 x3
ẋ3 = −x3 + u + (A.32)
T (x1 ) (1 + x1 )2
where β = λL0 /R and T (x1 ) = β[α + 1/(1 + x1 )].

A.9 Electrostatic Microactuator

Figure A.10 shows a schematic diagram of of a parallel-plate electrostatic actuator
driven by a voltage source [120]. The equation of motion of the movable upper
electrode is
d2 y dy
m 2 +b + k(y − y0 ) = F
dt dt
A.9. ELECTROSTATIC MICROACTUATOR 343

b k

R
m
+
+ is
vs va y y0
−
−

Figure A.10: Electrostatic microactuator

where m is the mass of the electrode, b the damping coefficient, k the elastic
constant, y the air gap, y0 the zero-voltage gap, and F the electrostatic force.
To prevent contact between the electrodes, an insulating layer of thickness δy0 is
mounted on the fixed electrode, with 0 < δ  1. Consequently, y is constrained to
δy0 ≤ y ≤ y0 . The capacitance of the structure is C = εA/y, where A is the area
of the electrode and ε the permittivity of the gap. The electric energy stored in the
capacitance is E = 12 Cva2 where va is the applied voltage, which is related to the
source voltage vs by the equation vs = va + Ris , in which is is the source current
and R its resistance. The force F is given by

∂E v 2 ∂C εAva2
F = = a =−
∂y 2 ∂y 2y 2
Letting q = Cva be the electric charge on the capacitance, it follows that is = dq/dt.
Therefore, the system is represented by the equations

d2 y dy q2 dq qy
m + b + k(y − y 0 ) = − , R = vs −
dt2 dt 2εA dt εA
For a constant input vs = Vs , the equilibrium points satisfy the equation
 3
q C0 q
Vs = − 2
C0 2ky0 C0

where C0 = εA/y0 is the zero-voltage capacitance. By differentiation, it can be

shown that the right-hand side of the foregoing equation
 increases monotonically
with q until it reaches the maximum value Vp = 8ky02 /(27C0 ), then it decreases
monotonically. Thus, for Vs < Vp , there are two equilibrium points, which merge
into one point at Vs = Vp , while no equilibrium points exist for Vs > Vp . Open-
loop control by applying a constant voltage Vs can stabilize the system only in a
limited range of the air gap. Moreover, the transient response cannot be shaped as
344 APPENDIX A. EXAMPLES

it is determined by the open-loop dynamics. Both drawbacks can be addressed by

feedback control. It is typical to assume that the position y and the charge q are
available for feedback control. The actual measurements are the voltage va and the
capacitance C,from which y and q are calculated using y = εA/C and q = Cva .
Let ω0 = k/m be the frequency of undamped free oscillation. Take the di-
mensionless time, state and control variables as
y dx1 q vs
τ = ω0 t, x1 = 1 − , x2 = , x3 = , u=
y0 dτ Qp Vp
where Qp = (3/2)C0 Vp . By definition, x1 is constrained to 0 ≤ x1 ≤ 1 − δ. The
normalized state model is given by

1 1 2
ẋ1 = x2 , ẋ2 = −x1 − 2ζx2 + x23 , ẋ3 = −(1 − x1 )x3 + u (A.33)
3 T 3
˙ denotes derivative with respect to τ .
where ζ = b/(2mω0 ), T = ω0 C0 R, and (·)

A.10 Robot Manipulator

The nonlinear dynamic equations for an m-link robot [119, 135] take the form

M (q)q̈ + C(q, q̇)q̇ + Dq̇ + g(q) = u (A.34)

where q is an m-dimensional vector of generalized coordinates representing joint po-

sitions, u an m-dimensional control (torque) input, and M (q) a symmetric positive
definite inertia matrix. The terms C(q, q̇)q̇, Dq̇, and g(q) account for centrifu-
gal/Coriolis forces, viscous damping, and gravity, respectively. For all q, q̇ ∈ Rm ,
Ṁ − 2C is a skew-symmetric matrix, where Ṁ is the total derivative of M (q) with
respect to t, D a positive semidefinite symmetric matrix, and g(q) = [∂P (q)/∂q]T ,
where P (q) is the total potential energy of the links due to gravity.
Taking x = col(x1 , x2 ), where x1 = q and x2 = q̇, yields the state equation

ẋ1 = x2 , ẋ2 = M −1 (x1 )[u − C(x1 , x2 )x2 − Dx2 − g(x1 )] (A.35)

The special case of a two-link robot with neglected damping, shown in Fig-
ure A.11, can be modeled [119] by equation (A.34) with

a1 + 2a4 cos q2 a2 + a4 cos q2

M= , D=0 (A.36)
a2 + a4 cos q2 a3

−q̇2 −(q̇1 + q̇2 ) b1 cos q1 + b2 cos(q1 + q2 )

C = a4 sin q2 , g= (A.37)
q̇1 0 b2 cos(q1 + q2 )
where a1 through a4 , b1 , and b2 are positive constants with nominal values

a1 = 200.01, a2 = 23.5, a3 = 122.5, a4 = 25, b1 = 784.8, b2 = 245.25 (A.38)

A.11. INVERTED PENDULUM ON A CART 345

Load
Pendulum

V
q2 θ
mg
y
Cart F H
q1

Figure A.11: Two-link robot. Figure A.12: Inverted pendulum

on a cart.

Consider also the case when, due to an unknown load, the actual system parameters
are perturbed to

a1 = 259.7, a2 = 58.19, a3 = 157.19, a4 = 56.25, b1 = 1030.1, b2 = 551.8125

(A.39)

A.11 Inverted Pendulum on a Cart

Consider the inverted pendulum on a cart of Figure A.12 [82]. The pivot of the
pendulum is mounted on a cart that can move in a horizontal direction. The cart is
driven by a motor that exerts a horizontal force F . The figure shows also the forces
acting on the pendulum, which are the force mg at the center of gravity, a horizontal
reaction force H, and a vertical reaction force V at the pivot. Writing horizontal
and vertical Newton’s laws at the center of gravity of the pendulum yields

d2 d2
m (y + L sin θ) = H and m (L cos θ) = V − mg
dt2 dt2
Taking moments about the center of gravity yields the torque equation

J θ̈ = V L sin θ − HL cos θ

while a horizontal Newton’s law for the cart yields

M ÿ = F − H − k ẏ

Here m is the mass of the pendulum, M the mass of the cart, L the distance from the
center of gravity to the pivot, J the moment of inertia of the pendulum with respect
to the center of gravity, k a friction coefficient, y the displacement of the pivot, θ
the angular rotation of the pendulum (measured clockwise), and g the acceleration
346 APPENDIX A. EXAMPLES

due to gravity. We will derive two models of this system, a fourth-order model that
describes the motion of the pendulum and cart when F is viewed as the control
input, and a second-order model that describes the motion of the pendulum when
the cart’s acceleration is viewed as the input.
Carrying out the indicated differentiation to eliminate H and V , we obtain

d2 d
H=m (y + L sin θ) = m (ẏ + Lθ̇ cos θ) = m(ÿ + Lθ̈ cos θ − Lθ̇2 sin θ)
dt2 dt
d2 d
V =m (L cos θ) + mg = m (−Lθ̇ sin θ) + mg = −mLθ̈ sin θ − mLθ̇2 cos θ + mg
dt2 dt
Substituting H and V in the θ̈- and ÿ-equations yields

J θ̈ = −mL2 θ̈(sin θ)2 − mL2 θ̇2 sin θ cos θ + mgL sin θ

−mLÿ cos θ − mL2 θ̈(cos θ)2 + mL2 θ̇2 sin θ cos θ
= −mL2 θ̈ + mgL sin θ − mLÿ cos θ

M ÿ = F − m(ÿ + Lθ̈ cos θ − Lθ̇2 sin θ) − k ẏ

The foregoing two equations can be rewritten as

J + mL2 mL cos θ θ̈ mgL sin θ

=
mL cos θ m + M ÿ F + mLθ̇2 sin θ − k ẏ
Hence,

θ̈ 1 m+M −mL cos θ mgL sin θ

= (A.40)
ÿ Δ(θ) −mL cos θ J + mL2 F + mLθ̇2 sin θ − k ẏ

where

Δ(θ) = (J + mL2 )(m + M ) − m2 L2 cos2 θ ≥ (J + mL2 )M + mJ > 0

Using x1 = θ, x2 = θ̇, x3 = y, and x4 = ẏ as the state variables and u = F as the

control input, the state equation is given by

ẋ1 = x2 (A.41)
1  
ẋ2 = (m + M )mgL sin x1 − mL cos x1 (u + mLx2 sin x1 − kx4 ) (A.42)
2
Δ(x1 )
ẋ3 = x4 (A.43)
1  
ẋ4 = −m L g sin x1 cos x1 + (J + mL )(u + mLx2 sin x1 − kx4 ) (A.44)
2 2 2 2
Δ(x1 )

In simulation, we shall use the numerical data:

m = 0.1, M = 1, k = 0.1, J = 0.008, g = 9.81, L = 0.5 (A.45)

A.12. TRANSLATIONAL OSCILLATOR WITH ROTATING ACTUATOR 347

Viewing the cart acceleration ÿ as the control input to the pendulum, we can
model the motion of the pendulum by the second-order equation

(J + mL2 )θ̈ = mgL sin θ − mLÿ cos θ

With ÿ = 0, the pendulum oscillates freely about θ = π with frequency ω where

ω 2 = mgL/(J + mL2 ). Introducing the dimensionless time τ = Ωt, where Ω is a
nominal value of ω, and dimensionless input u = −ÿ/g, the normalized equation of
motion is given by
θ̈ = a(sin θ + u cos θ) (A.46)
˙ now denotes differentiation with respect to τ and a = (ω/Ω)2 . With
where (.)
x1 = θ and x2 = θ̇ as the state variables, the state model is given by

ẋ1 = x2 , ẋ2 = a(sin x1 + u cos x1 ) (A.47)

We shall refer to the model (A.40) (or (A.41)–(A.44)) as the inverted pendulum on
a cart and the model (A.46 (or (A.47)) as the inverted pendulum.

A.12 Translational Oscillator with Rotating Actuator

Figure A.13 shows a schematic diagram of a Translational Oscillator with Rotating
Actuator (TORA) system [145]. It consists of a platform of mass M connected to
a fixed frame of reference by a linear spring, with spring constant k. The platform
can only move in the horizontal plane, parallel to the spring axis. On the platform,
a rotating proof mass is actuated by a DC motor. It has mass m and moment of
inertial J around its center of mass, located at a distance L from its rotational axis.
The control torque applied to the proof mass is denoted by u. The rotating proof
mass creates a force that can be controlled to dampen the translational motion of
the platform. We will derive a model for the system, neglecting friction. Figure A.13
shows that the proof mass is subject to forces Fx and Fy and a torque u. Writing
Newton’s law at the center of mass and taking moments about it yield the equations
d2 d2
m 2
(xc +L sin θ) = Fx , m 2 (L cos θ) = Fy , and J θ̈ = u+Fy L sin θ−Fx L cos θ
dt dt
where θ is the angular position of the proof mass (measured counter clockwise).
The platform is subject to the forces Fx and Fy , in the opposite directions, as well
as the restoring force of the spring. Newton’s law for the platform yields

M ẍc = −Fx − kxc

where xc is the translational position of the platform.

From the foregoing equations, Fx and Fy are given by

d2 d
Fx = m 2
(xc + L sin θ) = m (ẋc + Lθ̇ cos θ) = m(ẍc + Lθ̈ cos θ − Lθ̇2 sin θ)
dt dt
348 APPENDIX A. EXAMPLES

M
k
xc u Fx
L
θ
m
Fy

Figure A.13: Translational Oscillator with Rotating Actuator (TORA) system.

d2 d
Fy = m 2
(L cos θ) = m (−Lθ̇ sin θ) = −mLθ̈ sin θ − mLθ̇2 cos θ
dt dt
Eliminating Fx and Fy from the θ̈- and ẍc -equations yields

J + mL2 mL cos θ θ̈ u
=
mL cos θ m + M ẍc mLθ̇2 sin θ − kxc

Thus,

θ̈ 1 m+M −mL cos θ u

= (A.48)
ẍc Δ(θ) −mL cos θ J + mL2 mLθ̇2 sin θ − kxc

where

Δ(θ) = (J + mL2 )(m + M ) − m2 L2 cos2 θ ≥ (J + mL2 )M + mJ > 0

Using x1 = θ, x2 = θ̇, x3 = xc , and x4 = ẋc as the state variables and u as the

control input, the state equation is given by

ẋ1 = x2 (A.49)
1  
ẋ2 = (m + M )u − mL cos x1 (mLx22 sin x1 − kx3 ) (A.50)
Δ(x1 )
ẋ3 = x4 (A.51)
1  
ẋ4 = −mLu cos x1 + (J + mL2 )(mLx22 sin x1 − kx3 ) (A.52)
Δ(x1 )

The following data are used in simulation:

M = 1.3608, m = 0.096, , J = 0.0002175, k = 186.3 (A.53)

Appendix B

Mathematical Review

Euclidean Space1
The set of all n-dimensional vectors x = col(x1 , . . . , xn ), where x1 , . . . , xn are real
numbers, defines the n-dimensional Euclidean space denoted by Rn . The one-
dimensional Euclidean space consists of all real numbers and is denoted by R.
Vectors in Rn can be added by adding their corresponding components. They can
be multiplied by a scalar by multiplying )
each component by the scalar. The inner
n
product of two vectors x and y is xT y = i=1 xi yi .

Vector and Matrix Norms

The norm x of a vector x is a real-valued function with the properties
• x ≥ 0 for all x ∈ Rn , with x = 0 if and only if x = 0.
• x + y ≤ x + y, for all x, y ∈ Rn .
• αx = |α| x, for all α ∈ R and x ∈ Rn .
The second property is the triangle inequality. We use the Euclidean (2-) norm,
defined by
 1/2  T 1/2
x = |x1 |2 + · · · + |xn |2 = x x
An m × n matrix A of real elements defines a linear mapping y = Ax from Rn into
Rm . The induced 2-norm of A is defined by
 1/2
A = max Ax = max Ax = λmax (AT A)
x=1 x≤1

where λmax (AT A) is the maximum eigenvalue of AT A. For real matrices A and B
of dimensions m × n and n × , respectively, AB ≤ A B.

1 The mathematical review is patterned after similar reviews in [17] and [88]. For complete
coverage of the reviewed topics, the reader may consult [3], [112], or [113].

349
350 APPENDIX B. MATHEMATICAL REVIEW

Quadratic Forms
The quadratic form of a real symmetric n × n matrix P is defined by

n 
n
V (x) = xT P x = pij xi xj
i=1 j=1

It is positive semidefinite if V (x) ≥ 0 for all x and positive definite if V (x) > 0 for
all x = 0. It can be shown that V (x) is positive definite (positive semidefinite) if
and only if all the eigenvalues of P are positive (nonnegative), which is true if and
only if all the leading principal minors of P are positive (all principal minors of P
are nonnegative).2 If V (x) = xT P x is positive definite (positive semidefinite), we
say that the matrix P is positive definite (positive semidefinite) and write P > 0
(P ≥ 0). The following relations of positive definite matrices are used in the text:

λmin (P )x2 ≤ xT P x ≤ λmax (P )x2 (B.1)

where λmin (P ) and λmax (P ) are the minimum and maximum eigenvalues of P ,
respectively.
min xT P x = λmin (P )r2 (B.2)
x=r

For b ∈ R ,n 3

r2
min xT P x = (B.3)
|bT x|=r bT P −1 b
For a symmetric positive definite matrix P , the square root matrix P 1/2 is a sym-
metric positive definite matrix such that P 1/2 P 1/2 = P . For b ∈ Rn ,
√ √
max |bT x| = max  c bT P −1/2 y = c bT P −1/2  (B.4)
xT P x≤c y T y≤1

where P −1/2 is the inverse of P 1/2 .

Topological Concepts in Rn
Convergence of Sequences: A sequence of vectors x0 , x1 , . . ., xk , . . . in Rn ,
denoted by {xk }, converges to a limit vector x if

xk − x → 0 as k → ∞

which is equivalent to saying that, given any ε > 0, there is an integer N such that

xk − x < ε, ∀k≥N

2 This is a well-known fact in matrix theory. Its proof can be found in [15] or [44].
3 Following [89, Section 10.3], the Lagrangian associated with the constrained optimization
problem is L(x, λ) = xT P x + λ[(bT x)2 − r 2 ]. The first-order necessary conditions are 2P x +
2λ(bT x)b = 0 and (bT x)2 − r 2 = 0. It can be verified that the solutions λ = −1/(bT P −1 b) and
x = ±rP −1 b/(bT P −1 b) yield the minimal value r 2 /(bT P −1 b).
APPENDIX B. MATHEMATICAL REVIEW 351

The symbol “∀” reads “for all.” A vector x is an accumulation point of a sequence
{xk } if there is a subsequence of {xk } that converges to x; that is, if there is an
infinite subset K of the nonnegative integers such that {xk }k∈K converges to x. A
bounded sequence {xk } in Rn has at least one accumulation point in Rn . A sequence
of real numbers {rk } is increasing (monotonically increasing or nondecreasing) if
rk ≤ rk+1 ∀ k. If rk < rk+1 , it is strictly increasing. Decreasing (monotonically
decreasing or nonincreasing) and strictly decreasing sequences are defined similarly
with rk ≥ rk+1 . An increasing sequence of real numbers that is bounded from above
converges to a real number. Similarly, a decreasing sequence of real numbers that
is bounded from below converges to a real number.
Sets: A subset S ⊂ Rn is open if, for every vector x ∈ S, one can find an
ε-neighborhood of x

N (x, ε) = {z ∈ Rn | z − x < ε}

such that N (x, ε) ⊂ S. A set S is closed if and only if its complement in Rn is

open. Equivalently, S is closed if and only if every convergent sequence {xk } with
elements in S converges to a point in S. A set S is bounded if there is r > 0 such
that x ≤ r for all x ∈ S. A set S is compact if it is closed and bounded. A point
p is a boundary point of a set S if every neighborhood of p contains at least one
point of S and one point not belonging to S. The set of all boundary points of S,
denoted by ∂S, is called the boundary of S. A closed set contains all its boundary
points. An open set contains none of its boundary points. The interior of a set S is
S − ∂S. An open set is equal to its interior. The closure of a set S, denoted by S̄,
is the union of S and its boundary. A closed set is equal to its closure. An open set
S is connected if every pair of points in S can be joined by an arc lying in S. A set
S is called a region if it is the union of an open connected set with some, none, or
all of its boundary points. If none of the boundary points are included, the region
is called an open region or domain. A set S is convex if, for every x, y ∈ S and
every real number θ, 0 < θ < 1, the point θx + (1 − θ)y ∈ S. If x ∈ X ⊂ Rn and
y ∈ Y ⊂ Rm , we say that (x, y) belongs to the product set X × Y ⊂ Rn × Rm .
Continuous Functions: A function f mapping a set S1 into a set S2 is denoted
by f : S1 → S2 . A function f : Rn → Rm is continuous at a point x if f (xk ) → f (x)
whenever xk → x. Equivalently, f is continuous at x if, given ε > 0, there is δ > 0
such that
x − y < δ ⇒ f (x) − f (y) < ε
The symbol “⇒” reads “implies.” A function f is continuous on a set S if it is
continuous at every point of S, and it is uniformly continuous on S if, given ε > 0
there is δ > 0 (dependent only on ε) such that the inequality holds for all x, y ∈ S.
Note that uniform continuity is defined on a set, while continuity is defined at a
point. For uniform continuity, the same constant δ works for all points in the set.
Clearly, if f is uniformly continuous on a set S, then it is continuous on S. The
opposite statement is not true in general. However, if S is a compact set, then
352 APPENDIX B. MATHEMATICAL REVIEW

continuity and uniform continuity on S are equivalent. The function

(a1 f1 + a2 f2 )(·) = a1 f1 (·) + a2 f2 (·)

is continuous for any two scalars a1 and a2 and any two continuous functions f1 and
f2 . If S1 , S2 , and S3 are any sets and f1 : S1 → S2 and f2 : S2 → S3 are functions,
then the function f2 ◦ f1 : S1 → S3 , defined by

(f2 ◦ f1 )(·) = f2 (f1 (·))

is called the composition of f1 and f2 . The composition of two continuous functions

is continuous. If S ⊂ Rn and f : S → Rm , then the set of f (x) such that x ∈ S is
called the image of S under f and is denoted by f (S). If f is a continuous function
defined on a compact set S, then f (S) is compact; hence, continuous functions
on compact sets are bounded. Moreover, if f is real valued, that is, f : S → R,
then there are points p and q in the compact set S such that f (x) ≤ f (p) and
f (x) ≥ f (q) for all x ∈ S. If f is a continuous function defined on a connected set
S, then f (S) is connected. A function f defined on a set S is one to one on S if
whenever x, y ∈ S, and x = y, then f (x) = f (y). If f : S → Rm is a continuous,
one-to-one function on a compact set S ⊂ Rn , then f has a continuous inverse
f −1 on f (S). The composition of f and f −1 is identity; that is, f −1 (f (x)) = x.
A function f : R → Rn is piecewise continuous on an interval J ⊂ R if for every
bounded subinterval J0 ⊂ J, f is continuous for all x ∈ J0 , except, possibly, at a
finite number of points where f may have discontinuities. Moreover, at each point
of discontinuity x0 , the right-side limit limh→0 f (x0 + h) and the left-side limit
limh→0 f (x0 − h) exist; that is, the function has a finite jump at x0 .
Differentiable functions: A function f : R → R is differentiable at x if the
limit
f (x + h) − f (x)
f  (x) = lim
h→0 h
exists. The limit f  (x) is called the derivative of f at x. A function f : Rn → Rm is
continuously differentiable at a point x0 if the partial derivatives ∂fi /∂xj exist and
are continuous at x0 for 1 ≤ i ≤ m, 1 ≤ j ≤ n. It is continuously differentiable on
a set S if it is continuously differentiable at every point of S. It is of class C  , for
 ≥ 1, on S if each fi has continuous partial derivatives up to order . It is smooth
if it is of class C ∞ . For a continuously differentiable function f : Rn → R, the row
vector ∂f /∂x is defined by

∂f ∂f ∂f
= , ...,
∂x ∂x1 ∂xn

The gradient vector, denoted by ∇f (x), is

T
∂f
∇f (x) =
∂x
APPENDIX B. MATHEMATICAL REVIEW 353

For a continuously differentiable function f : Rn → Rm , the Jacobian matrix

[∂f /∂x] is an m×n matrix whose element in the ith row and jth column is ∂fi /∂xj .
Suppose S ⊂ Rn is open, f maps S into Rm , f is continuously differentiable at
x0 ∈ S, g maps an open set containing f (S) into Rk , and g is continuously differ-
entiable at f (x0 ). Then the mapping h of S into Rk , defined by h(x) = g(f (x)), is
continuously differentiable at x0 and its Jacobian matrix is given by the chain rule
  
∂h  ∂g  ∂f 
=
∂x x=x0 ∂f  ∂x 
f =f (x0 ) x=x0

Mean Value and Implicit Function Theorems

If x and y are two distinct points in Rn , then the line segment L(x, y) joining x and
y is
L(x, y) = {z | z = θx + (1 − θ)y, 0 < θ < 1}
Mean Value Theorem
Assume that f : Rn → R is continuously differentiable at each point x of an
open set S ⊂ Rn . Let x and y be two points of S such that the line segment
L(x, y) ⊂ S. Then there is a point z of L(x, y) such that

∂f 
f (y) − f (x) = (y − x) (B.5)
∂x x=z
When f : Rn → Rn , a multidimensional version of the Mean Value Theorem is
given by  1
∂f
f (y) − f (x) = (x + σ(y − x)) dσ (y − x) (B.6)
0 ∂x
which can be seen by setting J(x) = [∂f /∂x](x), and h(σ) = f (x + σ(y − x)) for
0 ≤ σ ≤ 1. By the chain rule, h (σ) = J(x + σ(y − x))(y − x). Using
 1
f (y) − f (x) = h(1) − h(0) = h (σ) dσ
0

we arrive at (B.6).
Implicit Function Theorem
Assume that f : Rn ×Rm → Rn is continuously differentiable at each point (x, y)
of an open set S ⊂ Rn × Rm . Let (x0 , y0 ) be a point in S for which f (x0 , y0 ) = 0
and for which the Jacobian matrix [∂f /∂x](x0 , y0 ) is nonsingular. Then there exist
neighborhoods U ⊂ Rn of x0 and V ⊂ Rm of y0 such that for each y ∈ V the
equation f (x, y) = 0 has a unique solution x ∈ U . Moreover, this solution can be
given as x = g(y), where g is continuously differentiable at y = y0 .

The proofs of the mean-value and implicit-function theorems, as well as the

other facts stated earlier in this appendix, can be found in any textbook on ad-
vanced calculus or mathematical analysis.4

4 See, for example, [3].

354 APPENDIX B. MATHEMATICAL REVIEW

Lemma B.1 (Gronwall–Bellman Inequality)5 Let λ : [a, b] → R be continuous

and μ : [a, b] → R be continuous and nonnegative. If a continuous function y :
[a, b] → R satisfies
 t
y(t) ≤ λ(t) + μ(s)y(s) ds
a
for a ≤ t ≤ b, then on the same interval
 t  t

y(t) ≤ λ(t) + λ(s)μ(s) exp μ(τ ) dτ ds

a s

In particular, if λ(t) ≡ λ is a constant, then

 t

y(t) ≤ λ exp μ(τ ) dτ

a

If, in addition, μ(t) ≡ μ ≥ 0 is a constant, then

y(t) ≤ λ exp[μ(t − a)]

Lemma B.2 (Comparison Lemma)6 Consider the scalar differential equation

u̇ = f (t, u), u(t0 ) = u0

where f (t, u) is continuous in t and locally Lipschitz in u, for all t ≥ 0 and all
u ∈ J ⊂ R. Let [t0 , T ) (T could be infinity) be the maximal interval of existence
of the solution u(t), and suppose u(t) ∈ J for all t ∈ [t0 , T ). Let v(t) be a con-
tinuous function whose upper right-hand derivative D+ v(t) satisfies the differential
inequality
D+ v(t) ≤ f (t, v(t)), v(t0 ) ≤ u0
with v(t) ∈ J for all t ∈ [t0 , T ). Then, v(t) ≤ u(t) for all t ∈ [t0 , T ). 3

5 See [74, Lemma A.1] for the proof.

6 See [74, Lemma 3.4] for the proof. If v(t) is differentiable at t, then D+ v(t) = v̇(t). If
|v(t + h) − v(t)| /h ≤ g(t, h), ∀ h ∈ (0, b], and limh→0+ g(t, h) = g0 (t), then D+ v(t) ≤ g0 (t).
Appendix C

Composite Lyapunov
Functions

The main challenge in Lyapunov theory is the search for a Lyapunov function. A
useful tool that helps this search is to represent the system as interconnection of
lower-order components, find a Lyapunov function for each component, and use
them to form a composite Lyapunov function for the whole system. We illustrate
this tool by examining cascade systems, interconnected systems with limited inter-
connections [4, 93, 126] and singularly perturbed systems [76]. For convenience we
present the idea for time-invariant systems and make use of quadratic-type Lya-
punov functions, which were introduced in Section 4.2.

C.1 Cascade Systems

Consider the cascade connection
η̇ = f1 (η, ξ), ξ˙ = f2 (ξ) (C.1)
where f1 and f2 are locally Lipschitz and f1 (0, 0) = 0, f2 (0) = 0. In Lemma 4.6
it is shown that if η̇ = f1 (η, ξ) is input-to-state stable and the origin of ξ˙ = f2 (ξ)
is globally asymptotically stable, then the origin of (C.1) is globally asymptotically
stable. Here we show how to construct a Lyapunov function for the cascade connec-
tion. For convenience, we assume that the origin of ξ˙ = f2 (ξ) is exponentially stable
and there is a continuously differentiable Lyapunov function V2 (ξ) that satisfies the
inequalities

∂V2 ∂V2
c1 ξ ≤ V2 (ξ) ≤ c2 ξ ,
2 2
f2 (ξ) ≤ −c3 ξ ,
2 ≤ c4 ξ (C.2)
∂ξ ∂ξ
in some neighborhood of ξ = 0, where c1 to c4 are positive constants. Suppose the
origin of η̇ = f1 (η, 0) is asymptotically stable and there is a continuously differen-

355
356 APPENDIX C. COMPOSITE LYAPUNOV FUNCTIONS

tiable Lyapunov function V1 (η) that satisfies the inequality

∂V1
f1 (η, 0) ≤ −W1 (η)
∂η

in some neighborhood of η = 0, where W1 (η) is positive definite and continuous.

Construct a composite Lyapunov function for the system (C.1) as

V (η, ξ) = bV1 (η) + V2 (ξ)

where b is a positive constant to be chosen. The derivative V̇ is given by

∂V1 1 ∂V2
V̇ (η, ξ) = b f1 (η, ξ) +  f2 (ξ)
∂η 2 V2 (ξ) ∂ξ
∂V1 ∂V1 1 ∂V2
= b f1 (η, 0) + b [f1 (η, ξ) − f1 (η, 0)] +  f2 (ξ)
∂η ∂η 2 V2 (ξ) ∂ξ

On any bounded neighborhood of the origin, we can use continuous differentiability

of V1 and the Lipschitz property of f1 to show that

∂V1

∂η [f1 (η, ξ) − f1 (η, 0)] ≤ kξ

for some positive constant k. Therefore

c3
V̇ (η, ξ) ≤ −bW1 (η) + bkξ − √ ξ
2 c2
√
Choosing b < c3 /(2k c2 ) ensures that V̇ is negative definite. The function V (η, ξ)
is continuously differentiable everywhere around the origin, except on the manifold
ξ = 0. Both V (η, ξ) and V̇ (η, ξ) are defined and continuous around the origin. It
can be easily seen that the statement of Theorem 3.3 is still valid and the origin of
(C.1) is asymptotically stable.
The foregoing analysis is limited to a bounded neighborhood of the origin. We
can remove this restriction and obtain a continuously differentiable composite Lya-
punov function if V1 (η) is a quadratic-type Lyapunov function. In particular, sup-
pose V1 (η) satisfies the inequalities

∂V1 ∂V1
f1 (η, 0) ≤ −c φ (η),
∂η ≤ k φ(η)
2
(C.3)
∂η

in some neighborhood of η = 0, where c and k are positive constants and φ is a

positive definite continuous function, and take the composite Lyapunov function as

V (η, ξ) = bV1 (η) + V2 (ξ)

C.2. INTERCONNECTED SYSTEMS 357

in which b is a positive constant to be chosen. The derivative V̇ satisfies

∂V1 ∂V1 ∂V2

V̇ (η, ξ) = b f1 (η, 0) + b [f1 (η, ξ) − f1 (η, 0)] + f2 (ξ)
∂η ∂η ∂ξ
≤ −bc φ2 (η) + bkL φ(η)ξ − c3 ξ2

where L is a Lipschitz constant of f1 with respect to ξ. The right-hand-side of the

foregoing inequality can be written as a quadratic form in (φ, ξ), to obtain

T


T

φ(η) bc −bkL/2 φ(η) def φ(η) φ(η)

V̇ ≤ − = − Q
ξ −bkL/2 c3 ξ ξ ξ

The choice b < 4cc3 /(kL)2 ensures that Q is positive definite, which shows that V̇
is negative definite; hence, the origin of (C.1) is asymptotically stable. If the as-
sumptions hold globally and V1 (η) is radially unbounded, the origin will be globally
asymptotically stable. If the origin of η̇ = f1 (η, 0) is exponentially stable and V1 (η)
satisfies inequalities similar to (C.2), then (C.3) is satisfied with φ(η) = η. In this
case the foregoing analysis shows that the origin of (C.1) is exponentially stable.
It is worthwhile to note that the same composite Lyapunov function can be
constructed for a system in the form

η̇ = f1 (η, ξ), ξ˙ = f2 (η, ξ)

if f2 (η, 0) = 0 for all η and there is a Lyapunov function V2 (ξ) that satisfies the
inequalities

∂V2 ∂V2
c1 ξ ≤ V2 (ξ) ≤ c2 ξ ,
2 2
f2 (η, ξ) ≤ −c3 ξ ,
2 ≤ c4 ξ
∂ξ ∂ξ

These inequalities show that the origin ξ = 0 is an exponentially stable equilibrium

point of ξ˙ = f2 (η, ξ) uniformly in η.

C.2 Interconnected Systems

Consider the interconnected system

ẋi = fi (xi ) + gi (x), i = 1, 2, . . . , m (C.4)

where xi ∈ Rni , n1 + · · · + nm = n, x = col(x1 , . . . , xm ), and fi and gi are locally

Lipschitz functions. Suppose fi (0) = 0 and gi (0) = 0 for all i so that the origin
x = 0 is an equilibrium point of (C.4). Ignoring the interconnection terms gi , the
system decomposes into m isolated subsystems:

ẋi = fi (xi ) (C.5)

358 APPENDIX C. COMPOSITE LYAPUNOV FUNCTIONS

with each one having an equilibrium point at its origin xi = 0. We start by searching
for Lyapunov functions that establish asymptotic stability of the origin for each
isolated subsystem. Suppose this search has been successful and that, for each
subsystem, we have a continuously differentiable Lyapunov function Vi (xi ) whose
derivative along the trajectories of the isolated subsystem (C.5) is negative definite.
The function
m
V (x) = bi Vi (xi ), bi > 0
i=1
is a composite Lyapunov function for the collection of the m isolated subsystems
for all values of the positive constants bi . Viewing the interconnected system (C.4)
as a perturbation of the isolated subsystems (C.5), it is reasonable to try V (x) as a
Lyapunov function candidate for (C.4). The derivative of V (x) along the trajectories
of (C.4) is given by

m
∂Vi m
∂Vi
V̇ (x) = bi fi (xi ) + bi gi (x)
i=1
∂xi i=1
∂xi

The first term on the right-hand side is negative definite by virtue of the fact that
Vi is a Lyapunov function for the ith isolated subsystem, but the second term is, in
general, indefinite. The situation is similar to our investigation of perturbed systems
in Section 4.2. Therefore, we may approach the problem by performing worst case
analysis where the term [∂Vi /∂xi ]gi is bounded by a nonnegative upper bound. Let
us illustrate the idea by using quadratic-type Lyapunov functions. Suppose that,
for i = 1, 2, . . . , m, Vi (xi ) satisfies

∂Vi ∂Vi
fi (xi ) ≤ −ci φ2i (xi ),
∂xi ∂xi ≤ ki φi (xi )

in some neighborhood of x = 0, for some positive constants ci and ki , and positive

definite continuous functions φi . Furthermore, suppose that in the same neighbor-
hood of x = 0, the interconnection terms gi (x) satisfy the bound

m
gi (x) ≤ γij φj (xj )
j=1
)m
for some nonnegative constants γij . Then, the derivative of V (x) = i=1 bi Vi (xi )
along the trajectories of the interconnected system (C.4) satisfies the inequality
⎡ ⎤
m 
m
V̇ (x) ≤ bi ⎣−ci φ2i (xi ) + ki γij φi (xi )φj (xj )⎦
i=1 j=1

The right-hand side is a quadratic form in φ1 , . . . , φm , which we rewrite as

V̇ (x) ≤ − 12 φT (DS + S T D)φ

C.3. SINGULARLY PERTURBED SYSTEMS 359

where φ = col(φ1 , . . . , φm ), D = diag(b1 , . . . , bm ), and S is an m × m matrix whose

elements are defined by
⎧
⎨ ci − ki γii , i=j
sij =
⎩
−ki γij , i = j

If there is a positive diagonal matrix D such that DS + S T D > 0, then V̇ (x) is

negative definite since φ(x) = 0 if and only if x = 0. Thus, a sufficient condition
for asymptotic stability of the origin as an equilibrium point of the interconnected
system is the existence of a positive diagonal matrix D such that DS + S T D is
positive definite. This is the case if and only if S is an M -matrix;1 that is, the leading
principal minors of S are positive. The M -matrix condition can be interpreted
as a requirement that the diagonal elements of S be “larger as a whole” than
the off-diagonal elements. It can be seen that diagonally dominant matrices with
nonpositive off-diagonal elements are M -matrices. The diagonal elements of S are
measures of the “degree of stability” for the isolated subsystems in the sense that the
constant ci gives a lower bound on the rate of decrease of the Lyapunov function Vi
with respect to φ2i (xi ). The off-diagonal elements of S represent the “strength of the
interconnections” in the sense that they give an upper bound on gi (x) with respect
to φj (xj ) for j = 1, . . . , m. Thus, the M -matrix condition says that if the degrees
of stability for the isolated subsystems are larger as a whole than the strength of
the interconnections, then the origin of the interconnected system is asymptotically
stable.

C.3 Singularly Perturbed Systems

Consider the singularly perturbed system

ẋ = f (x, z), εż = g(x, z) (C.6)

where f and g are locally Lipschitz in a domain that contains the origin, ε is a small
positive constant, f (0, 0) = 0, and g(0, 0) = 0. The system has a two-time-scale
structure because whenever |gi | ≥ k > 0, zi (t) moves faster in time than x(t). A
reduced model that captures the slow motion of x can be obtained by setting ε = 0
in the ż-equation and solving for z in terms of x. Suppose z = h(x) is the unique
solution of
0 = g(x, z)
in the domain of interest of x, h(x) is continuously differentiable with locally Lips-
chitz partial derivatives, and h(0) = 0. Substitution of z = h(x) in the ẋ-equation
results in the slow model
ẋ = f (x, h(x)) (C.7)
1 See [93] for the proof of this fact.
360 APPENDIX C. COMPOSITE LYAPUNOV FUNCTIONS

The fast dynamics are captured by the ż-equation when x is treated as a constant
parameter. This can be seen by changing the time variable from t to τ = (t − t0 )/ε
and setting ε = 0 in x(t0 + ετ ), which freezes x at x(t0 ). The resulting fast model
is given by
dz
= g(x, z) (C.8)
dτ
Our goal is to construct a composite Lyapunov function for the singularly perturbed
system (C.6) as a weighted sum of Lyapunov functions for the slow and fast models.
Because the fast model has an equilibrium point at z = h(x), it is more convenient
to work in the (x, y)-coordinates, where

y = z − h(x)

This change of variables shifts the equilibrium point of the fast model to the origin.
In the new coordinates, the singularly perturbed system is given by

∂h
ẋ = f (x, y + h(x)), εẏ = g(x, y + h(x)) − ε f (x, y + h(x)) (C.9)
∂x
Setting ε = 0 in the ẏ-equation yields

0 = g(x, y + h(x))

whose unique solution is y = 0. Substitution of y = 0 in the ẋ-equation results

in the slow model (C.7). Changing the time variable in the ẏ-equation from t to
τ = (t − t0 )/ε, and then setting ε = 0 on the right-hand side, results in the fast
model
dy
= g(x, y + h(x)) (C.10)
dτ
in which x is treated as a constant parameter. The slow model (C.7) has an equi-
librium point at x = 0. Suppose this equilibrium point is asymptotically stable
and there is a continuously differentiable Lyapunov function V1 (x) that satisfies the
inequality
∂V1
f (x, h(x)) ≤ −a1 φ21 (x) (C.11)
∂x
in some neighborhood of x = 0, where a1 is a positive constant and φ1 is a continuous
positive definite function. The fast model has an equilibrium point at y = 0.
Suppose this equilibrium point is asymptotically stable and there is a continuously
differentiable Lyapunov function V2 (x, y) such that

∂V2
g(x, y + h(x)) ≤ −a2 φ22 (y) (C.12)
∂y

in some neighborhood of (x = 0, y = 0), where a2 is a positive constant and φ2 is

a continuous positive definite function. We allow V2 to depend on x because x is
C.3. SINGULARLY PERTURBED SYSTEMS 361

a parameter of the fast model and Lyapunov functions may, in general, depend on
the system’s parameters, but we require V2 to satisfy the inequalities

W1 (y) ≤ V2 (x, y) ≤ W2 (y)

for some positive definite continuous functions W1 and W2 . Now consider the com-
posite Lyapunov function candidate

V (x, y) = bV1 (x) + V2 (x, y)

where b is a positive constant to be chosen. Calculating the derivative of V along

the trajectories of (C.9), we obtain
∂V1 1 ∂V2
V̇ (x, y) = b f (x, y + h(x)) + g(x, y + h(x))
∂x ε ∂y
∂V2 ∂h ∂V2
− f (x, y + h(x)) + f (x, y + h(x))
∂y ∂x ∂x
∂V1 1 ∂V2
= b f (x, h(x)) + g(x, y + h(x))
∂x ε ∂y

∂V1 ∂V2 ∂V2 ∂h

+b [f (x, y + h(x)) − f (x, h(x))] + − f (x, y + h(x))
∂x ∂x ∂y ∂x

We have represented the derivative V̇ as the sum of four terms. The first two terms
are the derivatives of V1 and V2 along the trajectories of the slow and fast models.
These two terms are negative definite in x and y, respectively, by inequalities (C.11)
and (C.12). The other two terms represent the effect of the interconnection between
the slow and fast dynamics, which is neglected at ε = 0. Suppose that these
interconnection terms satisfy the inequalities
∂V1
[f (x, y + h(x)) − f (x, h(x))] ≤ k1 φ1 (x)φ2 (y) (C.13)
∂x

∂V2 ∂V2 ∂h
− f (x, y + h(x)) ≤ k2 φ1 (x)φ2 (y) + γφ22 (y) (C.14)
∂x ∂y ∂x
for some nonnegative constants k1 , k2 , and γ. Using inequalities (C.11) to (C.14),
we obtain
1
V̇ (x, y) ≤ −ba1 φ21 (x) − a2 φ22 (y) + bk1 φ1 (x)φ2 (y) + k2 φ1 (x)φ2 (y) + γφ22 (y)
ε
= −φ (x, y)Qφ(x, y)
T

where φ = col(φ1 , φ2 ) and

⎡ ⎤
ba1 − 12 bk1 − 12 k2
Q=⎣ ⎦
− 12 bk1 − 1
2 k2 a2 /ε − γ
362 APPENDIX C. COMPOSITE LYAPUNOV FUNCTIONS

The right-hand side of the last inequality is negative definite when

a 
2
ba1 − γ > 14 (bk1 + k2 )2
ε
Hence, for all
4a1 a2 b
ε<
4a1 bγ + (bk1 + k2 )2
the origin of (C.6) is asymptotically stable.
Appendix D

Proofs

Proof of Lemma 9.5: Let V1 (η) be a Lyapunov function for η̇ = f0 (η, 0) that
satisfies

∂V1 ∂V1
c1 η ≤ V1 (η) ≤ c2 η , f0 (η, 0) ≤ −c3 η ,
∂η ≤ c4 η
2 2 2
∂η
in a domain Dη ⊂ Rn−ρ that contains the origin. The existence of V1 (η) is guaran-
teed by (the converse Lyapunov) Theorem 3.8. Let P = P T > 0 be the solution of
the Lyapunov equation P (A − BK) + (A − BK)T P = −I. Let Dξ ⊂ Rρ be a do-
main containing the origin such that Dη × Dξ ⊂ Dz . As in Section C.1, a Lyapunov
function for the nominal system (9.9) can be constructed as V (z) = bV1 (η) + ξ T P ξ,
with a sufficiently small b > 0. The derivative of V with respect to the perturbed
system (9.11) is given by
∂V1 ∂V1
V̇ = b f0 (η, 0) + b [f0 (η, ξ) − f0 (η, 0)]
∂η ∂η
+ ξ T [P (A − BK) + (A − BK)T P ]ξ + 2ξ T P BΔ(z)
≤ −bc3 η2 + bc4 Lη ξ − ξ2 + 2P B ξ Δ(z)

where L is a Lipschitz constant of f0 with respect to ξ over Dz . Using the bound

Δ(z) ≤ kz + δ ≤ kη + kξ + δ, we arrive at

V̇ ≤ −bc3 η2 + bc4 Lη ξ − ξ2 + 2kP B ξ2 + 2kP B ξ η
+ 2δP B ξ

T

η bc3 −(kP B + bc4 L/2) η

= −
ξ −(kP B + bc4 L/2) 1 − 2kP B ξ
+ 2δP B ξ

T

def η η

= − (Q + kQ1 ) + 2δP B ξ
ξ ξ

363
364 APPENDIX D. PROOFS

Choosing b < 4c3 /(c4 L)2 ensures that Q is positive definite. Then, there is k ∗ > 0,
dependent on b, such that for all k < k∗ the matrix Q + kQ1 is positive definite.
Hence, if δ = 0 we can conclude that the origin of the perturbed system (9.11)
is exponentially stable. When δ > 0 we continue the analysis to show ultimate
boundedness. Let λm > 0 be the minimum eigenvalue of Q + kQ1 . Then
2δP B
V̇ ≤ −λm z2 + 2δP B z ≤ −(1 − θ)λm z2 , for z ≥
λm θ
for 0 < θ < 1. Application of Theorem 4.5 shows that there are positive constants
δ ∗ and c such that z(t) will be ultimately bounded by δc if δ < δ ∗ . 2

Proof of Theorem 12.5: Write the trajectory under state feedback as χr =

(ηr , ξr ) and the one under output feedback as χ = (η, ξ). As in the proof of
Lemma 11.3, represent the dynamics of the observer using the scaled estimation
errors
ξi − ξˆi
ζi = ρ−i , 1≤i≤ρ (D.1)
ε
to obtain
ˆ
εζ̇ = A0 ζ + εB0 Δ(t, χ, ξ)
where ⎡ ⎤ ⎡ ⎤
−α1 1 0 ··· 0 0
⎢ −α2 0 1 ··· 0⎥ ⎢0⎥
⎢ ⎥ ⎢ ⎥
⎢ .. .. ⎥ , ⎢ ⎥
A0 = ⎢ .
..
. .⎥ B0 = ⎢ ... ⎥
⎢ ⎥ ⎢ ⎥
⎣−αρ−1 0 1⎦ ⎣0⎦
−αρ 0 ··· ··· 0 1
and
ˆ = a(η, ξ) − a0 (ξ)
Δ(t, χ, ξ) ˆ + [b(η, ξ) − b0 (ξ)]γ
ˆ s (ξ)
ˆ + δ(t, η, ξ, γs (ξ))
ˆ

ˆ b0 (ξ),
The matrix A0 is Hurwitz by design. Because a0 (ξ), ˆ and γs (ξ)
ˆ are globally
bounded,
ˆ ≤ L1 ,
|Δ(t, χ, ξ)| ∀ χ ∈ Ω, ξˆ ∈ Rρ , t ≥ 0 (D.2)
where throughout the proof Li , for i = 1, 2, . . ., denote positive constants inde-
pendent of ε. Let P be the positive definite solution of the Lyapunov equation
P A0 + AT0 P = −I, W (ζ) = ζ T P ζ, and Σ = {W (ζ) ≤ kε2 }. The first step of the
proof is to show that the constant k > 0 in the definition of Σ can be chosen such
that, for sufficiently small ε, the set Ω × Σ is positively invariant; that is, χ(t0 ) ∈ Ω
and ζ(t0 ) ∈ Σ imply that χ(t) ∈ Ω and ζ(t) ∈ Σ for all t ≥ t0 . Using (D.2), it can
be shown that, for all χ ∈ Ω,

εẆ = −ζ T ζ + 2εζ T P B0 Δ ≤ −ζ2 + 2εL1 P B0  ζ

≤ − 12 ζ2 , ∀ ζ ≥ 4εL1 P B0 
APPENDIX D. PROOFS 365

Using the inequalities

λmin (P )ζ2 ≤ ζ T P ζ ≤ λmax (P )ζ2

we arrive at
εẆ ≤ −σW, ∀ W ≥ ε2 W0 (D.3)
where σ = 1/(2λmax (P )) and W0 = λmax (P )(4L1 P B0 )2 . Taking k = W0 shows
that ζ(t) cannot leave Σ because Ẇ is negative on its boundary. On the other hand,
ˆ ≤ L2 ε. Using the Lipschitz property of β, ψ, and sat(·), it can
for ζ ∈ Σ, ξ − ξ
ˆ ≤ L3 ε/μ, where the μ factor appears because of the
be shown that γs (ξ) − γs (ξ)
function sat(s/μ). Since γs (ξ) = γ(ξ) in Ω, we have

ˆ ≤ εL3
γ(ξ) − γs (ξ) (D.4)
μ
Inspection of equation (12.47) shows that the control u appears only in the ṡ equa-
tion. Using (D.4) and recalling the analysis of sliding mode control from Sec-
tion 10.1, it can be shown that

εL3
sṡ ≤ b(η, ξ) −β0 |s| + |s|
μ

Taking ε ≤ μβ0 /(2L3 ) yields

sṡ ≤ −(b0 β0 /2)|s|

With this inequality, the analysis of Section 10.1 carries over to show that the
trajectory (η(t), ξ(t)) cannot leave Ω and enters Ωμ in finite time.
ˆ ∈ X, the
The second step of the proof is to show that for all χ(0) ∈ Ω0 and ξ(0)
trajectory (χ(t), ζ(t)) enters Ω × Σ within a time interval [0, τ (ε)], where τ (ε) → 0
as ε → 0. Notice that, due to the scaling (D.1), the initial condition ζ(0) could be
of the order of 1/ερ−1 . Because Ω0 is in the interior of Ω and the control γs (ξ) ˆ is
globally bounded, there is time T1 > 0, independent of ε, such that χ(t) ∈ Ω for
t ∈ [0, T1 ]. During this time interval, (D.2) and consequently (D.3) hold. It follows
from Theorem 4.5 that
! "  −σt/ε 
e L4 2
W (t) ≤ max e−σt/ε W (0), ε2 W0 ≤ max , ε W 0
ε2(ρ−1)

which shows that ζ(t) enters Σ within the time interval [0, τ (ε)] where
 
ε L4
τ (ε) = ln
σ W0 ε2ρ

By l’Hôpital’s rule it can be verified that limε→0 τ (ε) = 0. By choosing ε small

enough, we can ensure that τ (ε) < T1 . Thus, the trajectory (χ(t), ζ(t)) is bounded
366 APPENDIX D. PROOFS

ˆ is globally
for all t ≥ 0 and enters the set Ωμ ×Σ within finite time T . Because γs (ξ)
bounded and χ(0) = χr (0),

χ(t) − χr (t) ≤ L5 τ (ε) ∀ t ∈ [0, τ (ε)]

Over the interval [τ (ε), T ], the χ̇-equation under output feedback is O(ε) perturba-
tion of the corresponding equation under state feedback. Therefore, (12.61) follows
from continuous dependence of the solutions of differential equations on parame-
ters.1 2

Proof of Theorem 12.6: Write the closed-loop system under state feedback
as
χ̇ = f (χ, γ(ξ)) (D.5)
Inside Ωμ × Σ, the closed-loop system under output feedback is given by

χ̇ = f (χ, γ(ξ − Dζ)), εζ̇ = A0 ζ + εB0 Δ(t, χ, ξ − Dζ) (D.6)

where D is a diagonal matrix whose ith diagonal element is ερ−i and Δ(t, 0, 0) = 0.
Because the origin of (D.5) is exponentially stable, by (the converse Lyapunov)
Theorem 3.8 there is a Lyapunov function V (χ) that satisfies

∂V ∂V
c1 χ ≤ V (χ) ≤ c2 χ ,
2 2
f (χ, γ(ξ)) ≤ −c3 χ ,
2 ≤ c4 χ
∂χ ∂χ

in some neighborhood N of the origin, where c1 to c4 are positive constants. Take

μ small enough so that Ωμ ⊂ N . As in section C.3, we take Vc (χ, ζ) = V (χ) + W (ζ)
as a Lyapunov function candidate for (D.6). It can be shown that
1
V̇c ≤ −c3 χ2 + c5 χ ζ − ζ2 + c6 ζ2
ε
for some positive constant c5 and c6 . The foregoing inequality can be written as

T

χ c3 −c5 χ

V̇c ≤ − 12 c3 (χ2 + ζ2 ) − 1
2 ζ −c5 (2/ε − c3 − 2c6 ) ζ

The matrix of the quadratic form can be made positive definite by choosing ε small
enough. Then, V̇c ≤ − 12 c3 (χ2 + ζ2 ), which shows that the origin of (D.6) is
exponentially stable and all trajectories in Ωμ × Σ converge to the origin. Since all
ˆ ∈ X enter Ωμ × Σ, we conclude that Ω0 × X
trajectories with χ(0) ∈ Ω0 and ξ(0)
is a subset of the region of attraction. 2

Proof of Theorem 13.1: Using Assumption 13.6 we can show that there are
two compact positively invariant sets Ω and Ωμ such that every trajectory starting
1 See [74, Theorem 3.4].
APPENDIX D. PROOFS 367

in Ω enters Ωμ in finite time. The construction of Ω and Ωμ is done in three steps.

We write the closed-loop system in the form

ż = f˜0 (z, e, r, w)
ζ̇ = Aζ + Bs
 
s
ṡ = −b(η, ξ, w)β(η, ξ) sat + Δ(η, ξ, r, w)
μ

where ζ = col(e0 , . . . , eρ−1 ), e = Aζ + Bs, and A is Hurwitz. For |s| ≥ μ, we have

|Δ|
sṡ ≤ b −β + |s| ≤ b[−β + ]|s| ≤ −b0 β0 |s|
b

which shows that the set {|s| ≤ c}, with c > μ, is positively invariant. In the second
step, we use the Lyapunov function V2 (ζ) = ζ T P ζ, where P is the solution of the
Lyapunov equation P A + AT P = −I, and the inequality

V̇2 ≤ −ζ T ζ + 2ζ P B |s|

to show that the set {V2 ≤ c2 ρ1 } × {|s| ≤ c} is positively invariant, for ρ1 =

(2||P B/θ)2 λmax (P ), where 0 < θ < 1. Inside this set, we have e = Aζ + Bs ≤
 def
cA ρ1 /λmin (P ) + c = cρ2 . Finally, we use the inequality

V̇1 ≤ −α3 (z), ∀ z ≥ α4 (cρ2 )

to show that

Ω = {V1 ≤ α2 (α4 (cρ2 ))} × {V2 ≤ c2 ρ1 } × {|s| ≤ c}

is positively invariant. Similarly, it can be shown that

Ωμ = {V1 ≤ α2 (α4 (μρ2 ))} × {V2 ≤ μ2 ρ1 } × {|s| ≤ μ}

is positively invariant and every trajectory starting in Ω enters Ωμ in finite time.

Inside Ωμ , the system has an equilibrium point at (z = 0, ζ = ζ̄, s = s̄) where
ζ̄ = col(ē0 , 0, . . . , 0) and s̄ = k0 ē0 . Shifting the equilibrium point to the origin by
the change of variables ν = ζ − ζ̄ and τ = s − s̄, the system takes the singularly
perturbed form

ż = f˜0 (z, Aν + Bτ, r, w)

ν̇ = Aν + Bτ
μτ̇ = −b(η, ξ, w)β(η, ξ)τ + μΔ̃(z, ν, τ, r, w)

where Δ̃(0, 0, 0, r, w) = 0. Setting μ = 0 results in τ = 0 and yields the slow model

ż = f˜0 (z, Aν, r, w), ν̇ = Aν

368 APPENDIX D. PROOFS

which is a cascade connection of ż = f˜0 (z, Aν, r, w) with ν̇ = Aν. If z = 0 is

an exponentially stable equilibrium point of ż = f˜0 (z, 0, r, w) and V0 (z, r, w) is the
converse Lyapunov function provided by Theorem 3.8, we can construct a composite
Lyapunov function αV0 + ν T P ν with α > 0, as in Section C.1, to show that (z =
0, ν = 0) is an exponentially stable equilibrium point of the slow model. Then
we can construct a composite Lyapunov function V = αV0 + ν T P ν + 12 τ 2 , as in
Section C.3 to show that, for sufficiently small μ, (z = 0, ν = 0, τ = 0) is an
exponentially stable equilibrium point of the closed-loop system. It can be shown
that derivative of V satisfies the inequality
⎡ ⎤T ⎡ ⎤⎡ ⎤
z αk1 −αk2 −(αk3 + k4 ) z
V̇ ≤ − ⎣ ν ⎦ ⎣ −αk2 1 −k5 ⎦ ⎣ ν ⎦
|τ | −(αk3 + k4 ) −k5 (k6 /μ) − k7 |τ |
- ./ 0
Q

for some positive constants k1 to k7 . The derivative V̇ can be made negative definite
by choosing α < k1 /k22 to make (q11 q22 − q12
2
) positive, and then choosing μ small
enough to make the determinant of Q positive. Hence, every trajectory in Ωμ
converges to the equilibrium point (z = 0, ν = 0, τ = 0) as t → ∞. Because e = 0
at this point, we conclude that the error converges to zero. 2
Bibliography

[1] B.D.O. Anderson and S. Vongpanitlerd. Network Analysis and Synthesis: A

Modern Systems Theory Approach. Prentice-Hall, Englewood Cliffs, NJ, 1973.

[2] P.J. Anstaklis and A.N. Michel. Linear Systems. McGraw-Hill, New York,
1997.

[3] T.M. Apostol. Mathematical Analysis. Addison-Wesley, Reading, MA, 1957.

[4] M. Araki. Stability of large-scale nonlinear systems–quadratic-order theory of

composite-system method using M -matrices. IEEE Trans. Automat. Contr.,
AC-23:129–141, 1978.

[5] M. Arcak and P. Kokotovic. Nonlinear observers: a circle criterion design and
robustness analysis. Automatica, 37:1923–1930, 2001.

[6] B. Armstrong and C. Canudas de Wit. Friction modeling and compensation.

In W. Levine, editor, The Control Handbook, pages 1369–1382. CRC Press,
1996.

[7] R.B. Ash. Real Analysis and Probability. Academic Press, New York, 1972.

[8] A. Astolfi, D. Karagiannis, and R. Ortega. Nonlinear and Adaptive Control

with Applications. Springer-Verlag, London, 2008.

[9] A.N. Atassi and H.K. Khalil. A separation principle for the stabilization of
a class of nonlinear systems. IEEE Trans. Automat. Contr., 44:1672–1687,
1999.

[10] A. Bacciotti and L. Rosier. Lyapunov Functions and Stability in Control

Theory. Springer-Verlag, Berlin, 2005.

[11] S. Banerjee and G.C. Verghese, editors. Nonlinear Phenomena in Power

Electronics: Bifurcation, Chaos, Control, and Applications. IEEE Press, New
York, 2001.

369
370 BIBLIOGRAPHY

[12] J.S. Baras, A. Bensoussan, and M.R. James. Dynamic observers as asymptotic
limits of recursive filters: speial cases. SIAM J. Applied Math, 48:1147–1158,
1988.
[13] B.R. Barmish, M. Corless, and G. Leitmann. A new class of stabilizing con-
trollers for uncertain dynamical systems. SIAM J. Control & Optimization,
21:246–255, 1983.
[14] T. Basar and P. Bernhard. H∞ -Optimal Control and Related Minimax Design
Problems. Birkhäuser, Boston, second edition, 1995.
[15] R. Bellman. Introduction to Matrix Analysis. McGraw-Hill, New York, second
edition, 1970.
[16] B.W. Bequette. Process Dynamics: Modeling, Analysis, and Simulation.
Prentice Hall, Upper Saddle River, NJ, 1998.
[17] D.P. Bertsekas. Dynamic Programming. Prentice-Hall, Englewood Cliffs, NJ,
1987.
[18] F. Blanchini. Set invariance in control–a survey. Automatica, 35:1747–1767,
1999.
[19] C.I. Byrnes and A. Isidori. Asymptotic stabilization of minimum phase non-
linear systems. IEEE Trans. Automat. Contr., 36:1122–1137, 1991.
[20] C.I. Byrnes, A. Isidori, and J.C. Willems. Passivity, feedback equivalence, and
the global stabilization of minimum phase nonlinear systems. IEEE Trans.
Automat. Contr., 36:1228–1240, 1991.
[21] F.M. Callier and C.A. Desoer. Multivariable Feedback Systems. Springer-
Verlag, New York, 1982.
[22] C. T. Chen. Linear System Theory and Design. Holt, Rinehart and Winston,
New York, 1984.
[23] G. Chesi. Domain of Attraction: Analysis and Control via SOS Programming.
Springer-Verlag, London, 2011.
[24] S. Ching, Y. Eun, C. Gokcek, P.T. Kabamba, and S.M. Meerkov. Quasilinear
Control: Performance Analysis and Design of Feedback Systems with Non-
linear Sensors and Actuators. Cambridge University Press, Cambridge, UK,
2011.
[25] L.O. Chua, C.A. Desoer, and E.S. Kuh. Linear and Nonlinear Circuits.
McGraw-Hill, New York, 1987.
[26] L.O. Chua and Y. S. Tang. Nonlinear oscillation via volterra series. IEEE
Trans. Circuits Syst., CAS-29:150–168, 1982.
BIBLIOGRAPHY 371

[27] C.M. Close and D.K. Frederick. Modeling and Analysis of Dynamic Systems.
Houghton Mifflin, Boston, second edition, 1993.

[28] M. Corless. Control of uncertain nonlinear systems. J. Dyn. Sys. Measurement

and Control, 115:362–372, 1993.

[29] M. Corless and G. Leitmann. Continuous state feedback guaranteeing uniform

ultimate boundedness for uncertain dynamic systems. IEEE Trans. Automat.
Contr., AC-26:1139–1144, 1981.

[30] M. Corless and R. Shorten. On the characterization of strict positive realness

for general matrix transfer functions. IEEE Trans. Automat. Contr., 55:1899–
1904, 2010.

[31] J-M. Coron. Control and Nonlinearity. American Mathematical Society, Prov-
idence, RI, 2007.

[32] J-M. Coron, L. Praly, and A. Teel. Feedback stabilization of nonlinear sys-
tems: sufficient conditions and Lyapunov and input–output techniques. In
A. Isidori, editor, Trends in Control, pages 293–347. Springer-Verlag, New
York, 1995.

[33] Z. Deng. Nonlinear and Adaptive Control. IET, London, 2013.

[34] C. Desoer and Y-T. Wang. Foundation of feedback theory for nonlinear dy-
namical systems. IEEE Trans. Circuits Syst., 27:104–123, 1980.

[35] C.A. Desoer and M. Vidyasagar. Feedback Systems: Input–Output Properties.

Academic Press, New York, 1975.

[36] A.B. Dixon, D.M. Dawson, and S. Nagarkatti. Nonlinear Control Engineering
Systems: A Lyapunov-Based Approach. Birkhauser, Boston, MA, 2003.

[37] J.C. Doyle, K. Glover, P.P. Khargonekar, and B.A. Francis. State-space so-
lutions to standard H2 and H∞ control problems. IEEE Trans. Automat.
Contr., 34:831–847, 1989.

[38] A.F. Filippov. Differential equations with discontinuous right-hand side.

Amer. Math. Soc. Translations, 42:199–231, 1964.

[39] T.I. Fossen. Guidance and Control of Ocean Vehicles. John Wiley & Sons,
New York, 1994.

[40] A.L. Fradkov, I.V. Miroshnik, and V.O. Nikiforov. Nonlinear and Adaptive
Control of Complex Systems. Springer, Netherlands, 1999.

[41] B.A. Francis. A course in H∞ control theory, volume 88 of Lect. Notes Contr.
Inf Sci. Springer-Verlag, New York, 1987.
372 BIBLIOGRAPHY

[42] R.A. Freeman and P.V. Kokotovic. Optimal nonlinear controllers for feedback
linearizable systems. In Proc. American Control Conf., pages 2722–2726,
Seattle, WA, 1995.

[43] R.A. Freeman and P.V. Kokotovic. Robust Nonlinear Control Design: State
Space and Lyapunov Techniques. Birkhauser, Boston, MA, 1996.

[44] F.R. Gantmacher. Theory of Matrices. Chelsea Publ., Bronx, NY, 1959.

[45] F.M. Gardner. Phaselock Techniques. Wiley-Interscience, New York, 1979.

[46] J-P. Gauthier and I. Kupka. Observability and Observers for Nonlinear Sys-
tems. Cambridge University Press, Cambridge, UK, 2001.

[47] A. Gelb. Applied Optimal Estimation. MIT Press, Cambridge, MA, 1974.

[48] R. Genesio, M. Tartaglia, and A. Vicino. On the estimation of asymptotic

stability regions: state of the rrt and new proposals. IEEE Trans. Automat.
Contr., AC-30:747–755, 1985.

[49] W. Gerstner and W. Kistler. Spiking Neuron Models. Cambridge University

Press, Cambridge, UK, 2002.

[50] E. Gilbert and I. Kolmanovsky. Nonlinear tracking control in the presence of

state and control constraints: a generalized reference governor. Automatica,
38:2063–2073, 2002.

[51] G.H. Golub and C.F. Van Loan. Matrix Computations. The John Hopkins
University Press, Baltimore, 1983.

[52] J. Guckenheimer and P. Holmes. Nonlinear Oscillations, Dynamical Systems,

and Bifurcations of Vector Fields. Springer-Verlag, New York, 1983.

[53] W.M. Haddad and V. Chellabonia. Nonlinear Dynamical Systems and Con-
trol. Princeton University Press, Princeton, NJ, 2008.

[54] W. Hahn. Stability of Motion. Springer-Verlag, New York, 1967.

[55] J. Hale and H. Kocak. Dynamics and Bifurcations. Springer-Verlag, New

York, 1991.

[56] J.K. Hale. Ordinary Differential Equations. Wiley-Interscience, New York,

1969.

[57] K.M. Hangos, J. Bokor, and G. Szederkenyi. Analysis and Control of Nonlin-
ear Process Systems. Springer, London, 2004.

[58] P. Hartman. Ordinary Differential Equations. Wiley, New York, 1964.

BIBLIOGRAPHY 373

[59] J. Hespanha. Linear Systems Theory. Princeton University Press, Princeton,

NJ, 2009.

[60] D. Hill and P. Moylan. The stability of nonlinear dissipative systems. IEEE
Trans. Automat. Contr., AC-21:708–711, 1976.

[61] D.J. Hill and P.J. Moylan. Stability results for nonlinear feedback systems.
Automatica, 13:377–382, 1977.

[62] M.W. Hirsch and S. Smale. Differential Equations, Dynamical Systems, and
Linear Algebra. Academic Press, New York, 1974.

[63] J.C. Hsu and A.U. Meyer. Modern Control Principles and Applications.
McGraw-Hill, New York, 1968.

[64] T. Hu and Z. Lin. Control Systems with Actuator Saturation: Analysis and
design. Birkhauser, Boston, MA, 2001.

[65] T. Hu and Z. Lin. Composite quadratic Lyapunov functions for constrained

control systems. IEEE Trans. Automat. Contr., 48:440–450, 2003.

[66] A. Isidori. Nonlinear Control Systems. Springer-Verlag, Berlin, third edition,

1995.

[67] A. Isidori. Nonlinear Control Systems II. Springer-Verlag, London, 1999.

[68] A. Isidori and A. Astolfi. Disturbance attenuation and H∞ control via mea-
surement feedback in nonlinear systems. IEEE Trans. Automat. Contr.,
37:1283–1293, 1992.

[69] Z.P. Jiang, A.R. Teel, and L. Praly. Small gain theorem for ISS systems and
applications. Mathematics of Control, Signals, and Systems, 7:95–120, 1994.

[70] T. Kailath. Linear Systems. Prentice-Hall, Englewood Cliffs, NJ, 1980.

[71] L. Karafyllis and Z-P. Jiang. Stability and Stabilization of Nonlinear Systems.
Springer, London, 2011.

[72] N. Kazantzis and C. Kravaris. Nonlinear observer design using Lyapunov

auxiliary theorem. Syst. Contr. Lett., 34:241–247, 1998.

[73] H.K. Khalil. Universal integral controllers for minimum-phase nonlinear sys-
tems. IEEE Trans. Automat. Contr., 45:490–494, 2000.

[74] H.K. Khalil. Nonlinear Systems. Prentice Hall, Upper Saddle River, New
Jersey, third edition, 2002.

[75] P. Kokotovic and M. Arcak. Constructive nonlinear control: a historical

perspective. Automatica, 37:637–662, 2001.
374 BIBLIOGRAPHY

[76] P.V. Kokotovic, H.K. Khalil, and J. O’Reilly. Singular Perturbations Methods
in Control: Analysis and Design. SIAM, Philadelphia, PA, 1999.

[77] P.V. Kokotović and R. Marino. On vanishing stability regions in nonlinear

systems with high-gain feedback. IEEE Trans. Automat. Contr., AC-31:967–
970, 1986.

[78] N.N. Krasovskii. Stability of Motion. Stanford University Press, Stanford,

1963.

[79] A.J. Krener. The convergence of the extended Kalman filter. In A. Rantzer
and C.I. Byrnes, editors, Directions in Mathematical Systems Theory and
Optimization, pages 173–182. Springer-Verlag, Berlin, 2003.

[80] M. Krstic and H. Deng. Stabilization of Uncertain Systems. Springer, London,

1998.

[81] M. Krstic, I. Kanellakopoulos, and P. Kokotovic. Nonlinear and Adaptive

Control Design. Wiley-Interscience, New York, 1995.

[82] H. Kwakernaak and R. Sivan. Linear Optimal Control Systems. Wiley-

Interscience, New York, 1972.

[83] J.P. LaSalle. An invariance principle in the theory of stability. In J.K. Hale and
J.P. LaSalle, editors, Differential Equations and Dynamical Systems, pages
277–286. Academic Press, New York, 1967.

[84] W. Leonard. Control of Electrical Drives. Springer, Berlin, second edition,

1996.

[85] A. Levant. Higher-order sliding modes, differentiation, and output-feedback

control. Int. J. Contr., 76:924–941, 2003.

[86] Z. Lin and A. Saberi. Robust semi-global stabilization of minimum-phase

input–output linearizable systems via partial state and output feedback. IEEE
Trans. Automat. Contr., 40:1029–1041, 1995.

[87] R. Lozano, B. Brogliato, O. Egeland, and B. Maschke. Dissipative Systems

Analysis and Control: Theory and Applications. Springer, London, 2000.

[88] D.G. Luenberger. Optimization by Vector Space Methods. Wiley, New York,
1969.

[89] D.G. Luenberger. Introduction to Linear and Nonlinear Programming.

Addison-Wesley, Reading, MA, 1973.

[90] I.M.Y. Mareels and D.J. Hill. Monotone stability of nonlinear feedback sys-
tems. J. Mathematical Systems, Estimation and Control, 2:275–291, 1992.
BIBLIOGRAPHY 375

[91] R. Marino and P. Tomei. Nonlinear Control Design: Geometric, Adaptive &
Robust. Prentice-Hall, London, 1995.

[92] H.J. Marquez. Nonlinear Control Systems: Analysis and design. Wiley, Hobo-
ken, FL, 2003.

[93] A.N. Michel and R.K. Miller. Qualitative Analysis of Large Scale Dynamical
Systems. Academic Press, New York, 1977.

[94] A.N. Michel, N.R. Sarabudla, and R.K. Miller. Stability analysis of complex
dynamical systems. Circuits Systems Signal Process, 1:171–202, 1982.

[95] R.K. Miller and A.N. Michel. Ordinary Differential Equations. Academic
Press, New York, 1982.

[96] J.B. Moore and B.D.O. Anderson. Applications of the multivariable Popov
criterion. Int.J. Control, 5:345–353, 1967.

[97] K.S. Narendra and J. Taylor. Frequency Domain Methods for Absolute Sta-
bility. Academic Press, New York, 1973.

[98] H. Nijmeijer and A.J. van der Schaft. Nonlinear Dynamic Control Systems.
Springer-Verlag, Berlin, 1990.

[99] E. Noldus and M. Loccufier. A new trajectory reversing method for the
estimation of asymptotic stability regions. Int. J. Contr., 61:917–932, 1995.

[100] H. Olsson. Control Systems with Friction. PhD thesis, Lund Institute of
Technology, Lund, Sweden, 1996.

[101] R. Ortega, A. Loria, P.J. Nicklasson, and H. Sira-Ramirez. Passivity-based

Control of Euler-Lagrange Systems. Springer, London, 1998.

[102] B.E. Paden and S.S. Sastry. A calculus for computing Filippov’s differential
inclusion with application to the variable structure control of robot manipu-
lators. IEEE Trans. Circuits Syst.–I, CAS-34:73–82, 1987.

[103] M.A. Pai. Power System Stability Analysis by the Direct Method of Lyapunov.
North-Holland, Amsterdam, 1981.

[104] T.S. Parker and L.O. Chua. Practical Numerical Algorithms for Chaotic Sys-
tems. Springer-Verlag, New York, 1989.

[105] W.R. Perkins and J.B. Cruz. Engineering of Dynamic Systems. John Wiley,
New York, 1969.

[106] W. Perruquetti and J-P. Barbot. Sliding Mode Control in Engineering. CRS
Press, Boca Raton, FL, 2002.
376 BIBLIOGRAPHY

[107] Z. Qu. Robust control of nonlinear uncertain systems under generalized

matching conditions. Automatica, 29:985–998, 1993.

[108] S. Raghavan and J.K. Hedrick. Observer design for a class of nonlinear sys-
tems. Int. J. Contr., 59:515–528, 1994.

[109] R. Rajamani. Observers for Lipschitz nonlinear systems. IEEE Trans. Au-
tomat. Contr., 43:397–401, 1998.

[110] K. Reif, F. Sonnemann, and R. Unbehauen. An EKF-based nonlinear observer

with a prescribed degree of stability. Automatica, 34:1119–1123, 1998.

[111] N. Rouche, P. Habets, and M. Laloy. Stability Theory by Lyapunov’s Direct

Method. Springer-Verlag, New York, 1977.

[112] H.L. Royden. Real Analysis. Macmillan, New York, 1963.

[113] W. Rudin. Principles of Mathematical Analysis. McGraw-Hill, New York,

third edition, 1976.

[114] W.J. Rugh. Linear System Theory. Prentice-Hall, Upper Saddle River, NJ,
second edition, 1996.

[115] A. Saberi, P.V. Kokotovic, and H.J. Sussmann. Global stabilization of par-
tially linear composite systems. SIAM J. Control & Optimization, 28:1491–
1503, 1990.

[116] M. Safonov. Stability and Robustness of Multivariable Feedback Systems. MIT

Press, Cambridge, MA, 1980.

[117] I.W. Sandberg. Global inverse function theorems. IEEE Trans. Circuits Syst.,
CAS-27:998–1004, 1980.

[118] S. Sastry. Nonlinear Systems: Analysis, Stability and Control. Springer, New
York, 1999.

[119] L. Sciavicco and B. Siciliano. Modeling and Control of Robot Manipulators.

Springer-Velag, London, second edition, 2000.

[120] S.D. Senturia. Microsystem Design. Kluwer Academic, Norwell, MA, 2001.

[121] R. Sepulchre, M. Jankovic, and P. Kokotovic. Constructive Nonlinear Control.

Springer, London, 1997.

[122] D. Shevitz and B. Paden. Lyapunov stability theory of nonsmooth systems.

IEEE Trans. Automat. Contr., 39:1910–1914, 1994.

[123] H. Shim, J.H. seo, and A.R. Teel. Nonlinear observer design via passivation
of error dynamics. Automatica, 39:885–892, 2003.
BIBLIOGRAPHY 377

[124] Y. Shtessel, C. Edwards, L. Fridman, and A. Levant. Sliding Mode Control

and Observation. Birkhauser, Boston, MA, 2013.
[125] D.D. Siljak. Nonlinear Systems. Wiley, New York, 1969.
[126] D.D. Siljak. Large Scale Dynamic Systems: Stability and Structure. North-
Holland, New York, 1978.
[127] D. Simon. Optimal State Estimation: Kalman, H-Infinity and Nonlinear Ap-
proximation. Wiley Interscience, New York, 2006.
[128] H. Sira-Ramirez and M. Rios-Bolfvar. Sliding mode control of dc-to-dc power
converters via extended linearization. IEEE Trans. Circuits Syst., 41:652–661,
1994.
[129] J-J.E. Slotine and W. Li. Applied Nonlinear Control. Prentice Hall, Engle-
wood Cliffs, NJ, 1991.
[130] E. Sontag and A. Teel. Changing supply functions in input/state stable sys-
tems. IEEE Trans. Automat. Contr., 40:1476–1478, 1995.
[131] E.D. Sontag. Smooth stabilization implies coprime factorization. IEEE Trans.
Automat. Contr., 34:435–443, 1989.
[132] E.D. Sontag. Mathematical Control Theory: Deterministic Finite Dimen-
sional Systems. Springer, New York, second edition, 1998.
[133] E.D. Sontag. Input-to-state stability: Basic concepts and results. In
A.A. Agrachev et.al., editor, Nonlinear and Optimal Control Theory. Springer-
Verlag, Berlin, 2008.
[134] E.D. Sontag and Y. Wang. On characterizations of the input-to-state stability
property. Syst. Contr. Lett., 24:351–359, 1995.
[135] M.W. Spong, S. Hutchinson, and M. Vidyasagar. Robot Modeling and Control.
Wiley, Hoboken, New Jersey, 2005.
[136] S.H. Strogatz. Nonlinear Dynamics and Chaos. Addison Wesley, Reading,
MA, 1994.
[137] H.J. Sussmann and P.V. Kokotovic. The peaking phenomenon and the global
stabilization of nonlinear systems. IEEE Trans. Automat. Contr., 36:424–440,
1991.
[138] A.R. Teel. A nonlinear small-gain theorem for the analysis of control systems
with saturation. IEEE Trans. Automat. Contr., 41:1256–1270, 1996.
[139] A.R. Teel, T.T. Georgiou, L. Praly, and E. Sontag. Input-output stability. In
W. Levine, editor, The Control Handbook. CRC Press, 1995.
378 BIBLIOGRAPHY

[140] U. Topcu, A.K. Packard, P. Seiler, and G.J. Balas. Robust region-of-attraction
estimation. IEEE Trans. Automat. Contr., 55:137–142, 2010.
[141] J. Tsinias. Partial-state global stabilization for general triangular systems.
Syst. Contr. Lett., 24:139–145, 1995.
[142] V. Utkin, J. Guldner, and J. Shi. Sliding Mode Control in Electromechanical
Systems. CRS Press, Boca Raton, FL, second edition, 2009.
[143] A. van der Schaft. L2 -Gain and Passivity Techniques in Nonlinear Control.
Springer, London, 2000.
[144] M. Vidyasagar. Nonlinear Systems Analysis. SIAM, Philadelphia, PA, second
edition, 2002.
[145] C.-J. Wan, D.S. Bernstein, and V.T. Coppola. Global stabilization of the
oscillating eccentric rotor. In Proc. IEEE Conf. on Decision and Control,
pages 4024–4029, Orlando, FL, 1994.
[146] S. Wiggins. Introduction to Applied Nonlinear Dynamical Systems and Chaos.
Springer-Verlag, New York, 1990.
[147] J.C. Willems. The Analysis of Feedback Systems. MIT Press, Cambridge,
MA, 1971.
[148] J.L. Willems. The computation of finite stability regions by means of open
Lyapunov surfaces. Int.J. Control, 10:537–544, 1969.
[149] H.H. Woodson and J.R. Melcher. Electromechanical Dynamics, Part I: Dis-
crete Systems. John Wiley, New York, 1968.
[150] F.F. Wu and C.A. Desoer. Global inverse function theorem. IEEE Trans.
Circuit Theory, CT-19:199–201, 1972.
[151] V.A. Yakubovich, G.A. Leonov, and A.Kh. Gelig. Stability of Stationary Sets
in Control Systems with Discontinuity. World Scientific, Singapore, 2004.
[152] T. Yoshizawa. Stability Theory By Liapunov’s Second Method. The Mathe-
matical Society of Japan, Tokyo, 1966.
[153] K.D. Young, V.I. Utkin, and U. Ozguner. A control engineer’s guide to sliding
mode control. IEEE Trans. Contr. Syst. Tech., 7:328–342, 1999.
[154] G. Zames. Feedback and optimal sensitivity: model reference transformations,
multiplicative seminorms, and approximate inverses. IEEE Trans. Automat.
Contr., AC-26:301–320, 1981.
[155] A. Zecevic and D.D. Siljak. Estimating the region of attraction for large scale
systems with uncertainties. Automatica, 46:445–451, 2010.
BIBLIOGRAPHY 379

[156] V.I. Zubov. Methods of A.M. Lyapunov and Their Application. Noordhoff,
Groningen, The Netherlands, 1964.
Symbols 1

≡ identically equal
≈ approximately equal
def
= defined as
< (>) less (greater) than
≤ (≥) less (greater) than or equal to
 () much less (greater) than
∀ for all
∈ belongs to
⊂ subset of
→ tends to
⇒ implies
⇔
) equivalent to, if and only if
1 summation
product
|a| the absolute value of a scalar a
x the norm of a vector x (349)
A the induced norm of a matrix A (349)
max maximum
min minimum
sup supremum, the least upper bound
inf infimum, the greatest lower bound
Rn the n-dimensional Euclidean space (349)
Br the ball {x ∈ Rn | x ≤ r}
M the closure of a set M
∂M the boundary of a set M
(x, y) ∈ X × Y x ∈ X and y ∈ Y
dist(p, M ) the distance from a point p to
a set M (55)
f : S1 → S2 a function f mapping a set S1 into
a set S2 (351)

1 The page where the symbol is defined is given in parentheses.

380
SYMBOLS 381

f2 ◦ f1 the composition of two functions (352)

f −1 (·) the inverse of a function f (352)
f  (·) the first derivative of a real-valued function f (352)
∇f the gradient vector (352)
∂f
∂x the Jacobian matrix (353)
ẏ the first derivative of y with respect to time
ÿ the second derivative of y with respect to time
y (i) the ith derivative of y with respect to time
Lf h the Lie derivative of h with respect to
the vector field f (172)
[f, g] the Lie bracket of the vector fields f and g (181)
adkf g [f, adk−1
f g] (181)
diag[a1 , . . . , an ] a diagonal matrix with diagonal
elements a1 to an
block diag[A1 , . . . , An ] a block diagonal matrix with diagonal
blocks A1 to An
AT (xT ) the transpose of a matrix A ( a vector x)
λmax (P ) (λmin (P )) the maximum (minimum) eigenvalue of
a symmetric matrix P
P >0 a positive definite matrix P (350)
P ≥0 a positive semidefinite matrix P (350)
Re[z] or Re z the real part of a complex variable z
Im[z] or Im z the imaginary part of a complex variable z
z̄ or z ∗ the conjugate of a complex variable z
Z∗ the conjugate of a complex matrix Z
sat(·) the saturation function (238)
Sat(·) the vector saturation function (243)
sgn(·) the signum function (233)
O(·) order of magnitude notation,
f (μ) = O(μ) ≡ |f (μ)| ≤ Kμ
3 designation of the end of theorems, lemmas,
and corollaries
 designation of the end of examples
2 designation of the end of proofs
[xx] see reference number xx in the bibliography
Index

Asymptotic stability, 38 Converse Lyapunov theorems, 68

global, 40 Corollary, see Corollary Index,
global uniform, 77 Coulomb friction, 322
region of, 39
uniform, 77
Autonomous system, 6 DC motor, 173, 179, 186, 195, 230, 327,
340
DC-to-DC power converter, see boost con-
Backstepping, 211 verter
Barbashin–Krasovskii theorem, 57 Decrescent function, 78
Basin, 40 Diffeomorphism, 7
Biochemical reactor, 33, 194, 195, 228, global, 7
260, 338 local, 7
Boost converter, 119, 194, 195, 230, 280, Distribution, 182
337 involutive, 182
Boundary layer, 238 Domain, 351
Boundedness, 85 Domain of attraction, 40
Duffing’s equation, 277
Dynamic feedback, 197
Cascade systems, 355
Causal mapping, 123
Electrostatic microactuator, 194, 195, 196,
Center, 20
229, 261, 278, 306, 326, 342,
Chaos, 9
Equilibrium points, 7
Chattering, 236, 239
continuum of, 7
Circle criterion, 157, 160
hyperbolic, 21
multivariable, 157
isolated, 7
Class K function, 76
multiple, 8
Class KL function, 76 Existence and uniqueness, 5
Comparison lemma, 354 Exponential stability, 43, 77
Control Lyapunov function, 223 global, 43, 77
global, 223 Extended Kalman Filter, 266
robust, 223
Control saturation, 301
Controller form, 179, 201 Fast model, 360
Convergence, 350 Feedback linearization, 201

382
INDEX 383

Feedback passivation, 219, 285 Invariant manifold, 242

Feedforward, 309 Invariant set, 55
Finite escape time, 8 positively, 55
Functions Inverted pendulum, 188, 195, 228, 261,
composition of, 352 305, 347
continuous, 351 Inverted pendulum on a cart, 194, 228,
piecewise, 352 305, 345
uniformly, 351 ISS, 95
continuously differentiable, 352
differentiable, 352
in C  , 352 Jacobian matrix, 7, 353
inverse of, 352 Josephson junction, 11
one-to-one, 352
smooth, 352 Kalman–Yakubovich–Popov lemma, 114
Krasovskii’s method, 71
Gain function, 123
Gain margin, 225 L stability, 123
Gradient vector, 352 connection with passivity, 134
Gronwall–Bellman inequality, 354 finite-gain, 123
of state models, 127
H∞ control, 132 small-signal, 127
H∞ norm, 133 L2 gain, 132
Hamilton–Jacobi–Bellman equation, 72 Lp space, 122
Hamilton–Jacobi inequality, 135 LaSalle’s theorem, 56
Hardening spring, 331 Lemma, see Lemma Index,
Harmonic oscillator, 27, 279 Level surface, 49
High-gain feedback, 248, 255, 257 Lie bracket, 181
High-gain observer , 271, 288, 301, 323 Lie derivative, 172
Hölder’s inequality Liénard’s equation, 336
Hurwitz Limit cycles, 8
matrix, 43 Linear growth bound, 80
transfer function, 112 Linear time-invariant system, 40
Hydraulic systems, 12, 327 Linearization, 21, 43, 199, 263, 282
Lipschitz
condition, 3
Implicit function theorem, 353 globally, 3
Inner product, 349 locally, 3
Input-to-state stability, 95, 98 Loop transformation, 148
regional, 98 Lossless system, 105, 109
local, 98 Lure’s problem, 156
Integral action, 318 Lyapunov equation, 60
Interconnected systems, 357 Lyapunov function, 49
Invariance principle, 54 composite, 355
384 INDEX

cascade systems, 355 Normal form, 171

interconnected systems, 357 Nyquist criterion, 159
singularly perturbed systems, 359 multivariable, 158
control, 223 Nyquist plot, 159
energy, 51
energy-like, 57
Lure-type, 165 Observer-based control, 282, 286
quadratic, 50, 60, 157 Observer form, 187
quadratic-type, 84, 356, 358 Observers
ISS, 95 global, 269
Lyapunov redesign, 251 high-gain, 271, 288
Lyapunov’s indirect method, 44 local, 264
Lyapunov’s stability theorem, 46 nonlinear, 263
with linear error dynamics, 270
Orbit, 15
M-matrix, 359 periodic (closed), 27
Magnetic levitation, 194, 195, 229, 245, Output equation, 2
260, 294, 305, 325, 327, 341 Output feedback
Mass–spring system, 12, 72, 119, 331 regulation, 322
Matched disturbance, 309 stabilization, 281
Matched uncertainties, 231, 248 tracking, 322
Matching condition, 205 Output feedback form, 193
Mean value theorem, 353
Measurement noise, 273
Min–max control, 252 Parseval’s theorem, 133
Minimum phase systems, 177, 296 Partial feedback linearization, 207
Passive system, 105, 109
input strictly, 105, 109
Negative definite function, 49, 78 output strictly, 105, 109
Negative resistance oscillator, 7, 28, 335 strictly, 109
Negative semidefinite function, 49, 78 Passivity, 103
Neuron (FitzHugh–Nagumo) model, 35 Passivity-based control, 217, 283
Nodal sink, 18 Passivity theorems, 142
Nodal source, 19 PD controller, 220
Nonautonomous system, 6 Peaking phenomenon, 210, 272, 289
Nonlinear phenomena, essentially, 8 Pendulum 26, 45, 51, 73, 111, 140, 200,
Nonlinearity 233, 236, 255, 260, 279, 303,
memoryless, 103 310, 313, 317, 321, 327, 329
saturation, 238 Perturbation
sector, 106 nonvanishing, 75
signum, 233 structured, 82
Norm unstructured, 82
induced matrix, 349 vanishing, 75
vector, 349 Perturbed systems, 80
INDEX 385

Phase-locked loop, 12 Robust regulation

Phase plane, 15 output feedback, 322
Phase portrait, 16 state feedback, 318
PI controller, 321 Robust stabilization
PID controller, 312 output feedback, 296
Popov criterion, 164 state feedback
multivariable, 165 Robustness, 91, 153, 154, 205
Popov plot, 167 Rotating rigid spacecraft, 119
Positive definite function, 49, 78
Positive definite matrix, 50, 350
Positive limit point (set), 55 Saddle, 19
Positive real lemma, 114 Saturation function, 238
Positive real transfer function, 112 vector, 243
strictly, 112 Scalar system, 39
Positive semidefinite function, 49, 78 Sector condition (nonlinearity), 107
Positive semidefinite matrix, 50 Semitrailer-like vehicle, 195
Practical stabilization, 231 Separation principle, 288, 296
global, 232 Separatrix, 25
local, 232 Set, 351
regional, 232 boundary of, 351
semiglobal, 232 bounded, 351
Proper map, 7 closed, 351
closure of, 351
compact, 351
Quadratic form, 50, 350 connected, 351
interior of, 35
open, 351
Radially unbounded function, 49, 78 Set point, 314
Reaching phase, 233 Signum function, 233
Reference model, 308 Sliding mode control, 232, 233, 312, 319,
Region, 351 322, 323
Region of attraction, 39 continuously implemented, 238
Regular form, 241 Sliding manifold, 233
Regulation, see robust regulation Sliding phase, 233
Relative degree Slow model, 359
higher than one, 298 Small-gain theorem, 152
linear systems, 112 Softening spring, 331
nonlinear systems, 172 Sontag’s formula, 223
one, 296 Spiral sink, 20
Relaxation oscillation, 31 Spiral source, 20
Riccati equation, 267 Stability
Robot manipulator, 11, 119, 180, 184, absolute, 155, 156
188, 219, 230, 261, 278, 285, with a finite domain, 156
306, 326, 344 asymptotic, 38
386 INDEX

exponential, 43, 77 Time-invariant system, 6

input–output, 121 Time varying system, 6, 75
input-to-state, 94 TORA system, 194, 230, 261, 262, 306,
L, 121 347
matrix, 43 Tracking, 307
of feedback systems, 141 output feedback, 322
of equilibrium point, 37 state feedback, 310
of linear system, 40 robust, 312
of perturbed system, 80 Trajectory, 15
uniform, 77 Translational oscillator with rotating ac-
Stabilization, tuator, see TORA
global, 198 Triangle inequality, 349
local, 198 Tunnel diode circuit, 24, 33, 119, 333
practical, 231
regional, 198 Ultimate boundedness, 85
robust, 231 Underwater vehicle, 261, 328
via high-gain feedback, 257 Unstable
via Lyapunov redesign, 251 equilibrium point, 38
via sliding mode control, 232 focus, 20
semiglobal, 198 limit cycle, 31
via backstepping, 211 node, 19
via feedback linearization, 201
via linearization, 199
via output feedback, 281 Van der Pol equation, 29, 173, 178, 196,
via partial feedback linearization, 207 228, 336
via passivity-based control, 217 Variable gradient method, 52
via state feedback, 197 Variable structure control, see sliding mode
Stable control
equilibrium point, 38 Vector field, 15
focus, 19 Vector field diagram, 15
limit cycle, 31
node, 18 Wien–bridge oscillator, 34, 278
structurally, 21
State equation, 2
unforced, 2 Zero dynamics, 177
State (–space) model, 2 Zero-state observable, 116
State plane, 15
Static feedback, 197, 282
Storage function, 109
Stribeck effect, 332
Strict feedback form, 214
Strictly passive, 109
Synchronous generator, 11
INDEX 387

Corollary Index1

6.1 (130) 8.1 (192)

Lemma Index

1.1 (5) 1.2 (5) 1.3 (5) 1.4 (7) 3.1 (56) 3.2 (61)
4.1 (77) 4.2 (77) 4.3 (91) 4.4 (92) 4.5 (96) 4.6 (97)
4.7 (98) 5.1 (112) 5.2 (114) 5.3 (114) 5.4 (114) 5.5 (115)
5.6 (116) 6.1 (134) 9.1 (203) 9.2 (208) 9.3 (210) 9.4 (211)
9.5 (211) 9.6 (224) 9.7 (226) 11.1 (265) 11.2 (267) 11.3 (176)
12.1 (284) B.1 (354) B.2 (354)

Theorem Index

3.1 (41) 3.2 (44) 3.3 (47) 3.4 (56) 3.5 (57) 3.6 (58)
3.7 (60) 3.8 (68) 3.9 (69) 4.1 (78) 4.2 (78) 4.3 (78)
4.4 (87) 4.5 (89) 4.6 (95) 4.7 (98) 6.1 (127) 6.2 (131)
6.3 (131) 6.4 (133) 6.5 (134) 6.6 (135) 6.7 (137) 6.8 (137)
7.1 (143) 7.2 (143) 7.3 (146) 7.4 (147) 7.5 (149) 7.6 (150)
7.7 (152) 7.8 (157) 7.9 (161) 7.10 (165) 8.1 (176) 8.2 (183)
8.3 (191) 9.1 (218) 9.2 (222) 10.1 (245) 10.2 (247) 10.3 (253)
10.4 (254) 12.1 (286) 12.2 (295) 12.3 (297) 12.4. (298) 12.5 (302)
12.6 (302)

1 6.1(130) means Corollary 6.1 appears on page 130. The Lemma and Theorem Indexes are
written similarly.