(Ebook) Adaptive Approximation Based Control:

Unifying Neural, Fuzzy and Traditional Adaptive

Approximation Approaches by Jay A. Farrell,
Marios M. Polycarpou ISBN 9780471727880,
0471727881 download
https://ebooknice.com/product/adaptive-approximation-based-
control-unifying-neural-fuzzy-and-traditional-adaptive-
approximation-approaches-4540446

Explore and download more ebooks at ebooknice.com

We have selected some products that you may be interested in
Click the link to download now or visit ebooknice.com
for more options!.

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles,

James ISBN 9781459699816, 9781743365571, 9781925268492,
1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

(Ebook) Power System Load Frequency Control: Classical and

Adaptive Fuzzy Approaches by Hassan A. Yousef ISBN
9781498745574, 9782016037379, 2016037377, 1498745571

https://ebooknice.com/product/power-system-load-frequency-control-
classical-and-adaptive-fuzzy-approaches-49432982

(Ebook) Matematik 5000+ Kurs 2c Lärobok by Lena Alfredsson, Hans

Heikne, Sanna Bodemyr ISBN 9789127456600, 9127456609

https://ebooknice.com/product/matematik-5000-kurs-2c-larobok-23848312

(Ebook) SAT II Success MATH 1C and 2C 2002 (Peterson's SAT II

Success) by Peterson's ISBN 9780768906677, 0768906679

https://ebooknice.com/product/sat-ii-success-math-1c-and-2c-2002-peterson-
s-sat-ii-success-1722018
(Ebook) Power System Load Frequency Control: Classical and
Adaptive Fuzzy Approaches by Hassan A. Yousef ISBN
9781315166292, 9781351679572, 9781498745574, 1315166291,
1351679570, 1498745571
https://ebooknice.com/product/power-system-load-frequency-control-
classical-and-adaptive-fuzzy-approaches-5892108

(Ebook) Stable Adaptive Control and Estimation for Nonlinear

Systems: Neural and Fuzzy Approximator Techniques by Jeffrey T.
Spooner, Manfredi Maggiore, Raúl Ordóñez, Kevin M. Passino ISBN
0471415464
https://ebooknice.com/product/stable-adaptive-control-and-estimation-for-
nonlinear-systems-neural-and-fuzzy-approximator-techniques-2099854

(Ebook) Neural-Based Orthogonal Data Fitting: The EXIN Neural

Networks (Adaptive and Learning Systems for Signal Processing,
Communications and Control Series) by Giansalvo Cirrincione,
Maurizio Cirrincione ISBN 0471322709
https://ebooknice.com/product/neural-based-orthogonal-data-fitting-the-
exin-neural-networks-adaptive-and-learning-systems-for-signal-processing-
communications-and-control-series-2443968

(Ebook) Master SAT II Math 1c and 2c 4th ed (Arco Master the SAT
Subject Test: Math Levels 1 & 2) by Arco ISBN 9780768923049,
0768923042

https://ebooknice.com/product/master-sat-ii-math-1c-and-2c-4th-ed-arco-
master-the-sat-subject-test-math-levels-1-2-2326094

(Ebook) Cambridge IGCSE and O Level History Workbook 2C - Depth

Study: the United States, 1919-41 2nd Edition by Benjamin
Harrison ISBN 9781398375147, 9781398375048, 1398375144,
1398375047
https://ebooknice.com/product/cambridge-igcse-and-o-level-history-
workbook-2c-depth-study-the-united-states-1919-41-2nd-edition-53538044
ADAPTIVE APPROXIMATION
BASED CONTROL
ADAPTIVE APPROXI MATlON
BASED CONTROL
Unifying Neural, Fuzzy and Traditional
Adaptive Approximation Approaches

Jay A. Farrell
University of California Riverside

Marios M. Polycarpou
University of Cyprus and University of Cincinnati

WILEY-
INTERSCIENCE
A JOHN WILEY 81SONS, INC., PUBLICATION
Copyright 0 2006 by John Wiley & Sons, Inc. All rights reserved.

Published by John Wiley & Sons, Inc., Hoboken, New Jersey.

Published simultaneously in Canada.

No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form
or by any means, electronic, mechanical, photocopying, recording, scanning, or otherwise, except as
permitted under Section 107 or 108 of the 1976 United States Copyright Act, without either the prior
written permission of the Publisher, or authorization through payment of the appropriate per-copy fee to
the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, (978) 750-8400, fax
(978) 750-4470, or on the web at www.copyright.com. Requests to the Publisher for permission should
be addressed to the Permissions Department, John Wiley & Sons, Inc., 11 1 River Street, Hoboken, NJ
07030, (201) 748-601 1, fax (201) 748-6008, or online at http:llwww.wiley.coxn/go/permission.

Limit of LiabilityiDisclaimer of Warranty: While the publisher and author have used their best efforts in
preparing this book, they make no representations or warranties with respect to the accuracy or
completeness of the contents of this book and specifically disclaim any implied warranties of
merchantability or fitness for a particular purpose. No warranty may be created or extended by sales
representatives or written sales materials. The advice and strategies contnined lierciti m,iy not be
suitable for your situation. You should consult with a professional where appropriate. Neither the
publisher nor author shall be liable for any loss of profit or any other commercial damages, including
but not limited to special, incidental, consequential, or other damages.

For general information on our other products and services or for technical support, please contact our
Customer Care Department within the United States at (800) 762-2974, outside the United States at
(317) 572-3993 or fax (317) 572-4002.

Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may
not be available in electronic formats. For more information about Wiley products, visit our web site at
www.wiley.com.

Library of Congress Cataloging-in-Publication Data:

Farrell, Jay.
Adaptive approximation based control : unifying neural, fuzzy and traditional adaptive
approximation approaches / Jay A. Farrell, Marios M. Polycarpou.
p. cm.
Includes bibliographical references and index.
ISBN-I 3 978-0-471-72788-0 (cloth)
ISBN-I0 0-471-72788-1 (cloth)
1. Adaptive control systems. 2. Feedback control systems. I. Polycarpou, Marios. 11. Title.
TJ217.F37 2006
629.8'3Wc22
2005021 385

Printed in the United States of America.

10 9 8 7 6 5 4 3 2 1
To our families and friends.
CONTENTS

...
Preface Xlll

1 Introduction 1
1.I Systems and Control Terminology 1
1.2 Nonlinear Systems 3
1.3 Feedback Control Approaches 4
1.3.1 Linear Design 4
1.3.2 Adaptive Linear Design 6
1.3.3 Nonlinear Design 9
1.3.4 Adaptive Approximation Based Design 11
1.3.5 Example Summary 13
1.4 Components of Approximation Based Control 15
1.4.1 Control Architecture 15
1.4.2 Function Approximator 16
1.4.3 Stable Training Algorithm 17
1.5 Discussion and Philosophical Comments 18
1.6 Exercises and Design Problems 19

2 Approximation Theory 23
2.1 Motivating Example 24
2.2 Interpolation 29
vii
Viii CONTENTS

2.3 Function Approximation 30

2.3.1 Offline (Batch) Function Approximation 31
2.3.2 Adaptive Function Approximation 33
2.4 Approximator Properties 39
2.4.1 Parameter (Non)Linearity 39
2.4.2 Classical Approximation Results 43
2.4.3 Network Approximators 46
2.4.4 Nodal Processors 48
2.4.5 Universal Approximator 50
2.4.6 Best Approximator Property 52
2.4.7 Generalization 54
2.4.8 Extent of Influence Function Support 56
2.4.9 Approximator Transparency 65
2.4.10 Haar Conditions 66
2.4.1 1 Multivariable Approximationby Tensor Products 67
2.5 Summary 68
2.6 Exercises and Design Problems 69

3 Approximation Structures 71
3.1 Model Types 72
3.1.1 Physically Based Models 72
3.1.2 Structure (Model) Free Approximation 73
3.1.3 Function Approximation Structures 74
3.2 Polynomials 75
3.2.1 Description 75
3.2.2 Properties 77
3.3 Splines 78
3.3.1 Description 78
3.3.2 Properties 83
3.4 Radial Basis Functions 84
3.4.1 Description 84
3.4.2 Properties 86
3.5 Cerebellar Model Articulation Controller 87
3.5.1 Description 88
3.5.2 Properties 89
3.6 Multilayer Perceptron 93
3.6.1 Description 93
3.6.2 Properties 95
3.7 Fuzzy Approximation 96
3.7.1 Description 96
3.7.2 Takagi-Sugeno Fuzzy Systems 104
3.7.3 Properties 105
 CONTENTS ix

3.8 Wavelets 106

3.8.1 Multiresolution Analysis ( M U ) 108
3.8.2 M R 4 Properties 110
3.9 Further Reading 112
3.10 Exercises and Design Problems 112

4 Parameter Estimation Methods 115

4.1 Formulation for Adaptive Approximation 1 I6
4. I . 1 Illustrative Example 116
4.1.2 Motivating Simulation Examples 118
4.1.3 Problem Statement 124
4.1.4 Discussion of Issues in Parametric Estimation 125
4.2 Derivation of Parametric Models 127
4.2.1 Problem Formulation for Full-State Measurement 128
4.2.2 Filtering Techniques 129
4.2.3 SPR Filtering 131
4.2.4 Linearly Parameterized Approximators 131
4.2.5 Parametric Models in State Space Form 133
4.2.6 Parametric Models of Discrete-Time Systems 134
4.2.7 Parametric Models of Input-Output Systems 136
4.3 Design of Online Learning Schemes 138
4.3.1 Error Filtering Online Learning (EFOL) Scheme 138
4.3.2 Regressor Filtering Online Learning (RFOL) Scheme 140
4.4 Continuous-Time Parameter Estimation 141
4.4.1 Lyapunov-Based Algorithms 143
4.4.2 Optimization Methods 148
4.4.3 Summary 154
4.5 Online Learning: Analysis 154
4.5.1 Analysis of LIP EFOL Scheme with Lyapunov Synthesis Method 155
4.5.2 Analysis of LIP RFOL Scheme with the Gradient Algorithm 158
4.5.3 Analysis of LIP RFOL Scheme with RLS Algorithm 160
4.5.4 Persistency of Excitation and Parameter Convergence 161
4.6 Robust Learning Algorithms 163
4.6.1 Projection Modification 165
4.6.2 a-Modification 168
4.6.3 c-Modification 169
4.6.4 Dead-Zone Modification 170
4.6.5 Discussion and Comparison 172
4.7 Concluding Summary 173
4.8 Exercises and Design Problems 173
X CONTENTS

5 Nonlinear Control Architectures 179

5.1 Small-Signal Linearization 180
5.1.1 Linearizing Around an Equilibrium Point 181
5.1.2 Linearizing Around a Trajectory 183
5.I .3 Gain Scheduling 186
5.2 Feedback Linearization 188
5.2.1 Scalar Input-State Linearization 188
5.2.2 Higher-Order Input-State Linearization 190
5.2.3 Coordinate Transformations and Diffeomorphisms 193
5.2.4 Input-Output Feedback Linearization 196
5.3 Backstepping 203
5.3.1 Second Order System 203
5.3.2 Higher Order Systems 205
5.3.3 Command Filtering Formulation 207
5.4 Robust Nonlinear Control Design Methods 211
5.4.1 Bounding Control 21 1
5.4.2 Sliding Mode Control 212
5.4.3 Lyapunov Redesign Method 215
5.4.4 Nonlinear Damping 219
5.4.5 Adaptive Bounding Control 220
5.5 Adaptive Nonlinear Control 222
5.6 Concluding Summary 225
5.7 Exercises and Design Problems 226

6 Adaptive Approximation: Motivation and Issues 231

6.1 Perspective for Adaptive Approximation Based Control 232
6.2 Stabilization of a Scalar System 236
6.2.1 Feedback Linearization 231
6.2.2 Small-Signal Linearization 238
6.2.3 Unknown Nonlinearity with Known Bounds 239
6.2.4 Adaptive Bounding Methods 24 1
6.2.5 Approximating the Unknown Nonlinearity 243
6.2.6 Combining Approximation with Bounding Methods 250
6.2.7 Combining Approximation with Adaptive Bounding Methods 252
6.2.8 Summary 252
6.3 Adaptive Approximation Based Tracking 253
6.3.1 Feedback Linearization 253
6.3.2 Tracking via Small-Signal Linearization 253
6.3.3 Unknown Nonlinearities with Known Bounds 256
6.3.4 Adaptive Bounding Design 258
6.3.5 Adaptive Approximation of the Unknown Nonlinearities 262
 CONTENTS xi

6.3.6 Robust Adaptive Approximation 264

6.3.7 Combining Adaptive Approximation with Adaptive Bounding 266
6.3.8 Advanced Adaptive Approximation Issues 27 1
6.4 Nonlinear Parameterized Adaptive Approximation 278
6.5 Concluding Summary 280
6.6 Exercises and Design Problems 28 1

7 Adaptive Approximation Based Control: General Theory 285

7.1 Problem Formulation 286
7.1.1 Trajectory Tracking 286
7.1.2 System 286
7.1 .3 Approximator 287
7.1.4 Control Design 288
7.2 Approximation Based Feedback Linearization 288
7.2.1 Scalar System 289
7.2.2 Input-State 294
7.2.3 Input-Output 306
7.2.4 Control Design Outside the Approximation Region 23 308
7.3 Approximation Based Backstepping 309
7.3.1 Second Order Systems 309
7.3.2 Higher Order Systems 316
7.3.3 Command Filtering Approach 323
7.3.4 Robustness Considerations 328
7.4 Concluding Summary 330
7.5 Exercises and Design Problems 33 1

8 Adaptive Approximation Based Control for Fixed-Wing Aircraft 333

8.1 Aircraft Model Introduction 334
8.1.1 Aircraft Dynamics 334
8.1.2 Nondimensional Coefficients 335
8.2 Angular Rate Control for Piloted Vehicles 336
8.2.1 Model Representation 337
8.2.2 Baseline Controller 337
8.2.3 Approximation Based Controller 338
8.2.4 Simulation Results 345
8.3 Full Control for Autonomous Aircraft 349
8.3.1 Airspeed and Flight Path Angle Control 350
8.3.2 Wind-Axes Angle Control 355
8.3.3 Body Axis Angular Rate Control 359
8.3.4 Control Law and Stability Properties 362
8.3.5 Approximator Definition 365
xii CONTENTS

8.3.6 Simulation Analysis 367

8.3.7 Conclusions 371
8.4 Aircraft Notation 371

Appendix A: Systems and Stability Concepts 377

A. 1 Systems Concepts 377
A.2 Stability Concepts 379
A.2.1 Stability Definitions 379
A.2.2 Stability Analysis Tools 381
A.2.3 Strictly Positive Real Transfer Functions 391
A.3 General Results 392
A.4 Trajectory Generation Filters 394
A S A Useful Inequality 391
A.6 Exercises and Design Problems 398

Appendix B: Recommended Implementation and Debugging Approach 399

References 40 1

Index 417
PREFACE

During the last few years there have been significant developments in the control of highly
uncertain, nonlinear dynamical systems. For systems with parametric uncertainty, adaptive
nonlinear control has evolved as a powerful methodology leading to global stability and
tracking results for a class of nonlinear systems. Advances in geometric nonlinear control
theory, in conjunction with the development and refinement of new techniques, such as
the backstepping procedure and tuning functions, have brought about the design of control
systems with proven stability properties. In addition, there has been a lot of research
activity on robust nonlinear control design methods, such as sliding mode control, Lyapunov
redesign method, nonlinear damping, and adaptive bounding control. These techniques are
based on the assumption that the uncertainty in the nonlinear functions is within some
known, or partially known, bounding functions.
In parallel with developments in adaptive nonlinear control, there has been a tremendous
amount of activity in neural control and adaptive fuzzy approaches. In these studies, neural
networks or fuzzy approximators are used to approximate unknown nonlinearities. The
input/output response of the approximator is modified by adjusting the values of certain
parameters, usually referred to as weights. From a mathematical control perspective, neural
networks and fuzzy approximators represent just two classes of function approximators.
Polynomials, splines, radial basis functions, and wavelets are examples of other function
approximators that can be used-and have been used-in a similar setting. We refer to
such approximation models with adaptivity features as adaptive approximators, and control
methodologies that are based on them as adaptive approximation based control.
Adaptive approximation based control encompasses a variety of methods that appear
in the literature: intelligent control, neural control, adaptive fuzzy control, memory-based
control, knowledge-based control, adaptive nonlinear control, and adaptive linear control.

xiii
xiv PREFACE

Researchers in these fields have diverse backgrounds: mathematicians, engineers, and

computer scientists. Therefore, the perspective of the various papers in this area is also
varied. However, the objective of the various practitioners is typically similar: to design a
controller that can be guaranteed to be stable and achieve a high level of control performance
for systems that contain poorly modeled nonlinear effects, or the dynamics of the system
change during operation (for example, due to system faults). This objective is achieved
by adaptively developing an approximating function to compensate the nonlinear effects
during the operation of the system.
Many of the original papers on neural or adaptive fizzy control were motivated by such
concepts as ease of use, universal approximation, and fault tolerance. Often, ease of use
meant that researchers without a control or systems background could experiment with and
often succeed at controlling certain dynamics systems, at least in simulation. The rise of
interest in the neural and adaptive fuzzy control approaches occurred at a time when desktop
computers and dynamic simulation tools were becoming sufficiently cheap at reasonable
levels of performance to support such research on a wide basis.
However, prior to application on systems of high economic value, the control system
designer must carefully consider any new approach within a sound analytical framework that
allows rigorous analysis of conditions for stability and robustness. This approach opens
a variety of questions that have been of interest to various researchers: What properties
should the function approximator have? Are certain families of approximators superior
to others? How should the parameters of the approximator be estimated? What can be
guaranteed about the properties of the signals within the control system? Can the stability
of the approximator parameters be guaranteed? Can the convergence of the approximator
parameters be guaranteed? Can such control systems be designed to be robust to noise,
disturbances, and unmodeled effects. Can this approach handle significant changes in the
dynamics due to, for example, a system failure. What types of nonlinear dynamic systems
are amenable to the approach? What are the limitations? The objective of this textbook is
to provide readers with a framework for rigorously considering such questions.
Adaptive approximation based control can be viewed as one of the available tools that
a control designer should have in herihis control toolbox. Therefore, it is desirable for the
reader not only to be able to apply, for example, neural network techniques to a certain
class of systems, but more importantly to gain enough intuition and understanding about
adaptive approximation so that shelhe knows when it is a useful tool to be used and how to
make necessary modifications or how to combine it with other control tools, so that it can
be applied to a system that has not be encountered before.
The book has been written at the level of a first-year graduate student in any engineering
field that includes an introduction to basic dynamic systems concepts such as state variables
and Laplace transforms. We hope that this book has appeal to a wide audience. For use as
a graduate text, we have included exercises, examples, and simulations. Sufficient detail is
included in examples and exercises to allow students to replicate and extend results. Simu-
lation implementation of the methods developed herein is a virtually necessary component
of understanding implications of the approach. The book extensively uses ideas from sta-
bility theory. The advantage of this approach is that the adaptive law is derived based on the
Lyapunov synthesis method and therefore the stability properties of the closed-loop system
are more readily determined. Therefore, an appendix has been included as an aid to readers
who are not familiar with the ideas ofLyapunov stability analysis. For theoretically oriented
readers, the book includes complete stability analysis of the methods that are presented.
 PREFACE XV

Organization. To understand and effectively implement adaptive approximation based

control systems that have guaranteed stability properties, the designer must become familiar
with concepts of dynamic systems, stability theory, function approximation, parameter
estimation, nonlinear control methods, and the mechanisms to apply these various tools in
a unified methodology.
Chapter 1 introduces the idea of adaptive approximation for addressing unknown nonlin-
ear effects. This chapter includes a simple example comparing various control approaches
and concludes with a discussion of components of an adaptive approximation based control
system with pointers to the locations in the text where each topic is discussed.
Function approximation and data interpolation have long histories and are important
fields in their own right. Many of the concepts and results from these fields are impor-
tant relative to adaptive approximation based control. Chapter 2 discuss various properties
of function approximators as they relate to adaptive function approximation for control
purposes. Chapter 3 presents various function approximation structures that have been
considered for implementation of adaptive approximation based controllers. All of the ap-
proximators of this chapter are presented using a single unifying notation. The presentation
includes a comparative discussion of the approximators relative to the properties presented
in Chapter 2.
Chapter 4 focuses on issues related to parameter estimation. First we study the formu-
lation of parametric models for the approximation problem. Then we present the design of
online learning schemes; and finally, we derive parameter estimation algorithms with cer-
tain stability and robustness properties. The parameter estimation problem is formulated
in a continuous-time framework. The chapter includes a discussion of robust parame-
ter estimation algorithms, which will prove to be critical to the design of stable adaptive
approximation based control systems.
Chapter 5 reviews various nonlinear control system design methodologies. The objective
of this chapter is to introduce the methods, analysis tools, and key issues of nonlinear
control design. The chapter begins with a discussion of small-signal linearization and gain
scheduling. Then we focus on feedback linearization and backstepping, which are two of
the key design methods for nonlinear control design. The chapter presents a set of robust
nonlinear control design techniques. These methods include bounding control, sliding mode
control, Lyapunov redesign method, nonlinear damping, and adaptive bounding. Finally,
we briefly study the adaptive nonlinear control methodology. For each approach we present
the basic method, discuss necessary theoretical ideas related to each approach, and discuss
the effect (and accommodation) of modeling error.
Chapters 6 and 7 bring together the ideas of Chapters 1-5 to design and analyze con-
trol systems using adaptive approximation to compensate for poorly modeled nonlinear
effects. Chapter 6 considers scalar dynamic systems. The intent of this chapter is to al-
low a detailed discussion of important issues without the complications of working with
higher numbers of state variables. The ideas, intuition, and methods developed in Chapter
6 are important to successful applications to higher order systems. Chapter 7 will aug-
ment feedback linearization and backstepping with adaptive approximation capabilities to
achieve high-performance tracking for systems with significant unmodeled nonlinearities.
The presentation of each approach includes a rigorous Lyapunov analysis.
Chapter 8 presents detailed design and analysis of adaptive approximation based con-
trollers applied to fixed-wing aircraft. We study two control situations. First, an angular
rate controller is designed and analyzed. This controller is applicable in piloted aircraft
applications where the stick motion of the pilot is processed into body-frame angular rate
commands. Then we develop a full vehicle controller suitable for uninhabited air vehicles
XVi PREFACE

(UAVs). The control design is based on the approximation based backstepping methodol-
ogy.
Acknowledgments. The authors would like to thank the various sponsors that have sup-
ported the research that has resulted in this book: the National Science Foundation (Paul
Werbos), Air Force Wright-Patterson Laboratory (Mark Mears), Naval Air Development
Center (Marc Steinberg), and the Research Promotion Foundation of Cyprus. We would
like to thank our current and past employers who have directly and indirectly enabled this
research: University of California, Riverside; University of Cyprus; University of Cincin-
nati; and Draper Laboratory. In addition, we wish to acknowledge the many colleagues,
collaborators, and students who have contributed to the ideas presented herein, especially:
P. Antsaklis, W. L. Baker, J.-Y. Choi, M. Demetriou, S. Ge, J. Harrison, P. A. Ioannou, H. K.
Khalil, P. Kokotovic, F. L. Lewis, D. Liu, M. Mears, A. N. Michel, A. Minai, J. Nakanishi,
K. Narendra, C. Panayiotou, T. Parisini, K. M. Passino, T. Samad, S. Schaal, M. Sharma,
J.-J. Slotine, E. Sontag, G. Tao, A. Vemuri, H. Wang, S. Weaver, Y. Yang, X. Zhang, Y.
Zhao, and P. Zufiria. Finally, we would like to thank our families for their constant support
and encouragement throughout the long period that it took for this book to be completed.

Jay A. Farrell
Marios M. Polycarpou

Riverside, California and Nicosia, Cyprus

(1 0 hours time difference)
July 2005
CHAPTER I

INTRODUCTION

This book presents adaptive function estimation and feedback control methodologies that
develop and use approximations to portions ofthe nonlinear functions describing the system
dynamics while the system is in online operation. Such methodologies have been proposed
and analyzed under a variety of titles: neural control, adaptive fuzzy control, learning
control, and approximation-based control. A primary objective of this text is to present the
methods systematically in a unifying framework that will facilitate discussion of underlying
properties and comparison of alternative techniques.
This introductory chapter discusses some fundamental issues such as: (i) motivations
for using adaptive approximation-based control; (ii) when adaptive approximation-based
control methods are appropriate; (iii) how the problem can be formulated; and (iv) what
design decisions are required. These issues are illustrated through the use of a simple
simulation example.

1.1 SYSTEMS AND CONTROL TERMINOLOGY

Researchers interested in this area come from a diverse set of backgrounds other than
control; therefore, we start with a brief review of terminology standard to the field of
control systems, as depicted in Figure 1.1. The plant is the system to be controlled. The
plant will by modeled herein by a typically nonlinear set of ordinary differential equations.
The plant model is assumed to include the actuator and sensor models. The control system
is designed to achieve certain control objectives. As indicated in Figure 1.1, the inputs
to the control system include the reference input yc(t) (which is possibly passed through
Adaptive Approximation Based Control: Unifiing Neural, Fuzzy and Truditional Adaptive 1
Approximation Approaches. By Jay A. Farretl and Marios M. Polycarpou
Copyright @ 2006 John Wiley & Sons, Inc.
2 INTRODUCTION

-YJt)
Prefilter Control
System
u(t)
Plant - Y (t)

Figure 1.1 : Standard control system block diagram.

a prefilter to yield a smoother function y d ( t ) and its first T time derivatives y ! ’ ( t ) for
i = 1, . . . ,T ) and a set of measurable plant outputs y ( t ) . The control system processes
its inputs to produce the control system output u ( t ) that is applied to the plant actuators to
affect the desired change in the plant output. The control system output u(t)is sometimes
referred to as control signal orplant input. Figure 1.1 depicts as a block diagram a standard
closed-loop control system configuration.
The control system determines the stability of the closed-loop system and the response
to disturbances d ( t ) and initial condition errors. A disturbance is any unmodeled physical
effect on the plant state, usually caused by the environment. A disturbance is distinct from
measurement noise. The former directly and physically affects the system to be controlled.
The latter affects the measurement of the physical quantity without directly affecting the
physical quantity. The physical quantity may be indirectly affected by the noise through
the feedback control process.
Control design typically distinguishes regulation from tracking objectives. Regulation
is concerned with designing a control system to achieve convergence of the system state,
with a desirable transient response, from any initial condition within a desired domain of
attraction, to a single operating point. In this case, the signal yc(t) is constant. Tracking is
concerned with the design of a control system to cause the system output y ( t ) to converge
to and accurately follow the signal y d ( t ) . Although the input signal y c ( t ) to a tracking
controller could be a constant, it typically is time-varying in a manner that is not known
at the time that the control system is designed. Therefore, the designer of a tracking
controller must anticipate that the plant state may vary significantly on a persistent basis. It
is reasonable to expect that the designer of the open-loop physical system and the designer
of the feedback control system will agree on an allowable range of variation of the state
of the system. Herein, we will denote this operating envelope by V.The designer of the
physical system ensures safe operation when the state of the system is in V.The designer
ofthe controller must ensure that the state the system remains in V.Implicitly it is assumed
that the state required to track Yd lies entirely in V.
To illustrate the control terminology let us consider the example of a simple cruise control
system for automobiles. In this case, the control objective is to make the vehicle follow
a desired speed profile y c ( t ) , which is set by the driver. The measured output y ( t ) is the
sensed vehicle speed and the control system output u ( t ) is the throttle angle and/or fuel
injection rate. The disturbance d ( t ) may arise due to the wind or road incline. In addition to
disturbances, which are external factors influencing the state, there may also be modeling
errors. In the cruise control example, the plant model describes the effect of changing
the throttle angle on the actual vehicle speed. Hence, modeling errors may arise from
simplifications or inaccuracies in characterizing the effect of changing the throttle angle
on the vehicle speed. Modeling errors (especially nonlinearities), whether they arise due
 NONLINEAR SYSTEMS 3

to inaccuracies or intentional model simplifications, constitute one of the key motivations

for employing adaptive approximation-based control, and thus are crucial to the techniques
developed in this book.
In general, the objectives of a control system design are:

1. to stabilize the closed-loop system;

2. to achieve satisfactory reference input tracking in transient and at steady state;

3. to reduce the effect of disturbances;

4. to achieve the above in spite of modeling error;

5. to achieve the above in spite of noise introduced by sensors required to implement

the feedback mechanism.

Introductory textbooks in control systems provide linear-based design and analysis tech-
niques for achieving the above objectives and discuss some basic robustness and imple-
mentation issues [61, 66, 86, 1401. The theoretical foundations of linear systems analysis
and design are presented in more advanced textbooks (see, for example, [lo, 19,39, 130]),
where issues such as controllability, observability, and model reduction are examined.

1.2 NONLINEAR SYSTEMS

Most dynamic systems encountered in practice are inherently nonlinear. The control system
design process builds on the concept of a model. Linear control design methods can some-
times be applied to nonlinear systems over limited operating regions (i.e., 2)is sufficiently
small), through the process of small-signal linearization. However, the desired level of
performance or tracking problems with a sufficiently large operating region 2)may require
in which the nonlinearities be directly addressed in the control system design. Depending
on the type of nonlinearity and the manner that the nonlinearity affects the system, various
nonlinear control design methods are available [121, 134, 159, 234, 249, 2791. Some of
these methods are reviewed in Chapter 5.
Nonlinearity and model accuracy directly affect the achievable control system perfor-
mance. Nonlinearity can impose hard constraints on achievable performance. The challenge
of addressing nonlinearities during the control design process is further complicated when
the description of the nonlinearities involves significant uncertainty. When portions of the
plant model are unknown or inaccurately defined, or they change during operation, the con-
trol performance may need to be severely limited to ensure safe operation. Therefore there
is often an interest to improve the model accuracy. Especially in tracking applications this
will typically necessitate the use of nonlinear models. The focus of this text is on adaptively
improving models of nonlinear effects during online operation.
In such applications the level of achievable performance may be enhanced by using
adaptive function approximation techniques to increase the accuracy of the model of the
nonlinearities. Such adaptive approximation-based control methods include the popular
areas of adaptive fuzzy and neural control. This chapter introduces various issues related to
adaptive approximation-based control. This introductory discussion will direct the reader
to the appropriate sections of the text where more detailed discussion of each issue can be
found.
4 INTRODUCTION

1.3 FEEDBACK CONTROL APPROACHES

To introduce the concept of adaptive approximation-based control, consider the following

example, where the objective is to control the dynamic system

in a manner such that y ( t ) accurately tracks an externally generated reference input signal
yd(t). Therefore, the control objective is achieved if the tracking error Q(t) = y ( t ) - y d ( t )
is forced to zero. The performance specification is for the closed-loop system to have a
rate of convergence corresponding to a linear system with a dominant time constant T of
about 5.0 s. With this time constant, tracking errdrs due to disturbances or initial conditions
should decay to zero in approximately 15 s (= 37). The system is expected to normally
operate within y E 120,601, but may safely operate on the region 23 = {y E [0, loo]}. Of
course, all signals in the controller and plant must remain bounded during operation.
However, the plant model is not completely accurate. The best model available to the
control system designer is given by

where f,(y) = -y and go(y) = 1.0 0 . 3 ~The + . actual system dynamics are not known or
available to the designer. For implementation of the following simulation results, the actual
dynamics will be

f(y) = -1 -0.01y2

Therefore, there exists significant error between the design model and the actual dynamics
over the desired domain of operation.
This section will consider four alternative control system design approaches. The ex-
ample will allow a concrete, comparative discussion, but none of the designs have been
optimized. The objective is to highlight the similarities, distinctions, complexity, and com-
plicating factors of each approach. The details of each design have been removed from this
discussion so as not to distract from the main focus. The details are included in the problem
section of this chapter to allow further exploration. These methodologies and various others
will be analyzed in substantially greater detail throughout the remainder of the book.

1.3.1 Linear Design

Given the design model and performance specification, the objective in this subsection is
to design a linear controller for the system

y ( t ) = h ( y ( t ) , u ( t )=
) -y(t) + (1.0 + O . S y ( t ) ) u ( t ) (1.3)
so that the linearized closed-loop system is stable (stability concepts are reviewed in Ap-
pendix A) and has the desired tracking error convergence rate. This controller is designed
based on the idea of small-signal linearization and is approximate, even relative to the model.
Section 1.3.3 will consider feedback linearization, which is a nonlinear design approach
that exactly linearizes the model using the feedback control signal.
 FEEDBACK CONTROL APPROACHES 5

For the scalar system g = h(y, u), an operatingpoint is a pair of real numbers (y*, u')
such that h(y*, u*)= 0. If y = y* and u = u*, then jr = 0. In a more general setting,
the designer may need to linearize around a time-varying nominal trajectory ( y * ( t ) , u * ( t ) ) .
Note that operating points may be stable or unstable (see the discussion in Appendix A). An
operating point analysis only indicates the values of y at which it is possible, by appropriate
choice of u, for the system to be in steady state. For our example, the set of operating points
is defined by (y', u*)such that
u*= - Y*
+
1 0.3y*'
Therefore, the design model indicates that the system can operate at any y E D.
The operating point analysis does not indicate how u ( t ) should be selected to get con-
vergence to any particular operating point. Convergence to a desired operating point is an
objective for the control system design. In a linear control design, the best available model
is linearized around an operating point and a linear controller is designed for that linearized
model. If we choose the operating point (y*, u') = (40, fi)
as the design point, then the
linearized dynamics are (see Exercise 1.1)
1
-by = ---by
13
+ 13&,
where by = y - 40 and bu = u - 3. The linear controller
40 0.2(s+ L,
U ( S )= --
13 13s
l3 F(s)
used with the design model results in a stable system that achieves the specification at
y* = 40. In the above, s is the Laplace variable, U ( s ) denotes the Laplace transform of
u ( t ) ,C(t)= y ( t ) - y d ( t ) , and y d ( t ) is the reference input. Ofcourse, D is large enough that
a linear controller designed to achieve the specification at one operating point will probably
not achieve the specification at all operating points in D or for yd(t) varying with time over
the region D.
Figure 1.2 shows the performance using the linear controller of eqn. (1.4) for a series
of amplitude step inputs changing between yd = 20 and yd = 60. Note that the response
exhibits two different convergence rates indicated by T~ and 7 2 . One is significantly slower
than the desired 5 s. Therefore, the linear controller does not operate as designed. There
are two reasons for this. First, there is significant error between the design model and the
actual dynamics of the system. Second, an inherent assumption of linear design is that
the linear controller will only be used in a reasonably small neighborhood of the operating
point for which the controller was designed. The degree of reasonableness depends on the
nonlinear system of interest. For these two reasons, the actual linearized dynamics at the
two points y* = 20 and yc = 60 are distinct from the linearized dynamics of the design
model at the design point y* = 40. The design methodology to determine eqn. (1.4) relied
on cancelling the pole of the linearized dynamics. With modeling error, even for a linear
system, the pole is not cancelled; instead, there are two poles. One near the desired pole
and one near the origin. The second pole is dominant and yields the slowly converging
error dynamics.
Improved performance using linear methods could be achieved by various methods.
First, additional modeling efforts could decrease the error between the actual dynamics
and the design model, but may be expensive and will not solve the problem of operating
far from the linearization point. Second, high gain control will decrease the sensitivity to
6 INTRODUCTION

65

I I t I
0 10 20 30 40 50 60 70 80 90 100
Time, t

Figure 1.2: Performance of the linear control system of eqn. (1.4) with the dynamic system
of eqn. (1.1). The solid curve is y ( t ) . The dashed curve is yd(t).

modeling error, but will result in a higher bandwidth closed-loop system as well as a large
control effort. Third, gain scheduling methods (although not truly linear) address the issue
of limiting the use of a linear controller to a region about its design point by switching
between a set of linear controllers as a function of the system state. Each linear controller
is designed to meet the performance specification (for the design model) on a small region
of operation Di. The regions Q are defined such that they cover the region of operation 2)
(i.e., D C U z , D i ) . Gain scheduling a set of linear controllers does not address the issue
of error between the actual system and the design model.

1.3.2 Adaptive Linear Design

Through linearization, the dynamics near a fixed operating point ( y * ,u') are approximated
by
$ ( t ) = a' + +
b * y ( t ) c*u(t), (1.5)
where a', b', and c* are parameters that depend on (y*,u*). In one possible adaptive
control approach, the control law is

U
1
= - (-a - by
C
+ y d + 0.2(Yd - y)) , (1.6)

where yd E C ' ( D ) (i.e., the first derivative of yd exists and is continuous within the region
D),and a , b, care parameter estimates of a * , b', and c*, respectively. Note that if ( a ,b, c) =
(a', b', c*), then exact cancellation occurs and the resulting error dynamics are
s = -0.29,
Random documents with unrelated
content Scribd suggests to you:
 “I beg your pardon, my love—” he would have stopped her; but
she saw what in particular had offended him, and ran on.
“I am your wife, I know. But I am a person, too; and I own that I
would rather be with people who—who respect me for what I am in
myself, as well as for what you have made me. Forgive me for
saying so; it is rather natural, I think. And it happens that I should
like to see my parents again, and my sisters. It is six months since I
was at Blackheath. So that would be an opportunity, and a reason—
while you were at the Rectory.”
“You wish me to go there alone?” She could guess at the scalding
spot beneath his armour-plate.
“I should love to go with you,” she said, “if—if it could be
managed.”
“I may mention to you,” he said coldly, “that you will not find an
old acquaintance there. Since his mother’s death my young relative,
Tristram Duplessis, has bestirred himself. He has sold the cottage.”
She had not been prepared for an attack in flank, and blenched
before it. Then she told her fib. “My reason against going with you
had nothing to do with Mr. Duplessis,” she said; and, watching her,
he did not believe her.
He turned to his papers. “It shall be as you wish, my love,” he
said. “I will write to Constantia. It may well be that I shall not care
to resume a broken habit. Are you going up to dress? If so, and if
you should happen to see Wilbraham, would you tell him that I am
ready?”
She hovered about his studious back, as if on the brink of
speech; but thought better of it and went slowly out of the room.
Intensely conscious of her going, he cowered at his desk, looking
sideways—until he heard the door close. Then he began to read,
with lips pressed close together.
In the hall Mrs. Germain almost ran into the arms of Wilbraham,
who, scarlet in the face and wet as with rain, was racing to his room.
“By jove, I beg your pardon, Mrs. Germain!”
“You only made me jump,” she laughed. “Have you been playing
all this time?”
“I know, I know! It was Gunner’s fault, upon my honour.”
 “It always is Lord Gunner’s fault. Mr. Germain asked me to tell
you that he was ready.”
“Good Lord!” cried the unhappy youth. “And I’m sw—as hot as
anything.”
“Go and change,” she said kindly. “I’ll go back to him.”
He was fervent. “You are an angel! But I’ve told you that before.”
Their eyes met; they laughed together. He pelted upstairs.
“Mr. Wilbraham will be with you in a second,” she said, entering
the library again. Had she seen him spring round as she came in? No
doubt of it. “I left my book down by the lake—and I know you don’t
like that. Do you?”
“No, dearest, no. I confess the foible.” His eyes invited her
nearer. She advanced to his table and stood by him, her hand
touched his shoulder. He was inordinately happy, though he made no
immediate sign. But presently his arm went about her waist, and
then she bent down and leaned her cheek for his kiss. They
remained together, saying nothing, until she heard Wilbraham
coming down, three stairs at a time. Then she slipped away and just
caught him outside the door.
“I had to tell a fib,” she told him. “I said that I had left my book
by the lake.”
“Well!” He looked at her. “I’ll bet that’s not a fib.”
“No,” she laughed. “But it was meant to be. Now I’m going to get
it myself.”
“You are an angel!” he said. “Don’t. I’ll go presently. I should love
to.”
“No. I shall go myself. I deserve it.”
“You deserve—!” He stopped himself. “Look here,” he said, “send
Gunner. No, he’s changing. Send young Chaveney.”
She opened her eyes—fatal use! “Is Mr. Chaveney here? I
thought he said——”
Wilbraham chuckled. “Did you suppose he’d go when it came to
the point? Not he! Why, before we’d played half a set he came to
borrow some clothes off me.”
He glided smoothly into the library. Mrs. Germain fetched her
book.
 II
REFLECTIONS ON HONEYMOONS AND SUCHLIKE

The years fly, we know, and come not again, and there’s balm in
that for the wounds they leave. For we forget a good deal, and Hope
is a faithful lover, and never quits us for long together; and then
there’s honest Use-and-Wont, surely our friend. Because you were a
fool yesterday, you’re wise to-day; and if you’re a fool to-morrow—
why, the alternation is established. There’s a progression; it is like
the rotation of crops.
There’s a mort of healing in a brace of longish years. The county,
which had found little Mrs. Germain stiff when she came home from
her honeymoon, now looked to her for stiffness when it felt relaxed.
Her idiosyncrasy was accepted, you see; once admitted to be a
person, she became a personage. And, discovered by the county,
she discovered herself. She found out that she had a character; she
had never known that before, nor had any others who had had to do
with her: Mrs. James, to wit, Miss de Speyne, her husband. The
process of these discoveries ought to entertain us for a chapter, and
its resolution shall be attempted. But the county learned it first,
when it came to rely upon her stiffness. The Chaveneys, the Gerald
Swetebredes, the Trevor-Waynes, the Perceforest people, before the
two years were over, forgot that they had ever eyed each other, with
brows inquiring “Colonial?” or spelling “Hopeless, my dear!” Such
looks had passed, but now, on the contrary, they leaned—some
heavily. Lady Chaveney was one. “She is charming with Guy,” she
said more than once, “quite charming. An influence—in the nicest
way.” She added, once, as if the news was sacred, “I believe he’s
told her everything.” Guy was the Chaveney heir, the florid, assured
youth whom we met just now on our visit; he had been pronounced
“wild” by Mr. Germain; and he had told her everything. She took
herself quite seriously with Guy, in the elder-sister fashion, Mr.
Germain, at first approving, as, at first, he had approved every sign
of her making way. He came, before the end of two years, to feel
differently, lost touch with the sense of his benevolence, felt to be
losing grip of many things. But in the early days he had approved,
there’s no doubt—in those days of stress and taut nerves when,
returning from a honeymoon by much too long, she had found Mrs.
James pervading the great, orderly house, and had, without knowing
it, braced herself for a tussle, and unawares found herself in it, and
amazingly the winner. Her husband had backed her up there, in his
quiet way. Short, quick, breathless work it had been—a fight in
spasms. She had been crossing the hall when the great lady came
out of the Little Library.
“Ah, Mary—A Mrs. Burgess has called, I see—wife of some one in
Farlingbridge. She called while you were out. A politeness very
natural under the circumstances—but not the custom here, I think.
Lady Diana, I happen to know, never—I suppose you will send cards
by the carriage. That would answer the purpose very well. We have
never known the townspeople, you know—in that sort of way. There
is a tenants’ party in the summer. They come to that.”
Mary had listened. She was pale, but her eyes smouldered.
“I can’t do that, Mrs. Germain. I mean, I must return the call.”
“Ah? It will be against my recommendation.”
“I am very sorry. I asked Mrs. Burgess to call when I met her the
other day at Waysford.”
“Really? Waysford? One would meet her there, I suppose. A Sale
of Work?”
“Yes. But I asked her to call upon me. It was kind of her to come
so soon.”
Mrs. James pressed her lips together. So soon! Why, the woman
would fly! “Does my brother know of this, may I ask?”
“I don’t know,” said Mary, out of breath. She was scared, but
meant to go on.
“It will be better that he should be told.”
“If you think it will interest him—yes,” Mary said, and went
upstairs—to stare out of window, clench and unclench her hands.
Mrs. James reported the case to her brother-in-law, and Mary drove,
the next day, to Farlingbridge—her husband with her—and returned
the call. Nothing more was said; nor, when the visit of a Colonel
Dermott, V.C., and his lady, townspeople, too, had to be witnessed,
was a word of warning uttered. But Mrs. James left within a fortnight
of her rout, staying only for the first dinner-party at Southover. That
was how she learned that Mary Middleham had character. It shocked
her; and it was annoying, too, that she could expect no sympathy
from James.
The house-parties for the winter shooting, and those dinner-
parties for the county had to be gone through with somehow. She
set herself squarely to the task, and was glad enough to believe
towards the end of her two years that she was learning the
business. There was little to do, indeed, but be agreeable, but she
found that more than enough. Agreeable she could be when she felt
happy; her nature was as sweet as an apple. But if she felt hurt she
must show it, and she discovered that that was a cardinal sin. Then
there was the language to master, the queer, impertinent, leisurely
laconics of these people—expensive, perfectly complacent, incredibly
idle young men, old men without reticence, airy, free-spoken
women, and girls who unaffectedly ignored her. To cope with such as
these she must be even as they were, or seem so. The quickness of
their give-and-take in conversation, the ripple and flow, the ease of
the thing, asked an alertness of her which excited while it tried her
to death. Perpetually flagging at the game, she spurred herself
perpetually; for she discovered that there is no more deadly sin in
the code than an awkward pause, that being all of a piece with the
end and aim of living—which is smooth running. A woman should die
sooner than drop a conversation, or murder it.
She was at her best with the men, as perhaps she might expect.
She could run, she could walk all day, chatter, laugh outright, seem
to be herself; they paid her the compliment of approving looks. But
among the women she knew that she must be herself, a very
different thing. She felt infinitely small, ill-dressed, ill-mannered,
clumsy, and a dunce. It was from them, however, that she gained
her reputation of being stiff; she had them to thank for that. It had
come to her in a flash of spirit one day in the summer of her first
year, that if ignoring was in the wind, she could ignore with the best.
She chose to ignore Mrs. Chilmarke, Mrs. Ralph Chilmarke, a beauty,
a dainty blonde and a wit. She did it steadily for three days, at what
a cost she could never have guessed when she began it, and her
reward was great. Mrs. Chilmarke respected her for it, and the
Duchess—a duchess was in the house—was frankly delighted, and
said so. She had watched out the match, and had backed the brune.
Under such exertions as these character will out, while it may
slumber through years of pedagogy. But she worked hard at her
lessons directly she had found out what she wanted, and was
tolerably equipped for her tour in France and Italy when the time
came. She made no way with Latin—Mr. Germain had to give that
up; and English literature made her yawn. She insisted on botany,
for reasons unknown to the good gentleman, and became great
friends with the head gardener, a Scotchman, who made the initial
mistake of supposing her a little fool, and was ever afterwards her
obedient servant. Shall we do wrong in putting this study down to
Senhouse’s credit? I think not. Quietly and methodically, after a
method all her own, Mary Germain began to find herself, as they
say. But before she did that her husband had to find her; and he,
poor gentleman, who had had to begin upon their wedding day, was
at the end of his discoveries before he was at the end of his
honeymoon. So far he struggled, but after that he suffered—dumbly
and in secret, within his plate armour. The fact is, there had been
too much honeymoon. His evident discomfort had made her self-
conscious, killed her ease, threatened her gratitude—upon which he
had proposed to subsist—and turned him from an improbable mate
into a rather unsuccessful father of his wife.
October is a bad time for honeymoons; the evenings are so long.
Nevertheless, at Torquay, her mind had been fairly easy about him.
He had liked the hotel. At Saltcombe he had been pretty miserable,
much on her conscience. He had taught her chess, it seems, and if
she had known what she was about, chess might have done pretty
well. But unfortunately she took to chess, and began to beat him at
it by audacious combinations and desperate sallies quite
unwarranted by science. That vexed him sadly. He abandoned the
game, telling her frankly that he could not help being irritated to see
skill out-vailed by temerity. “One plays, you see, my love, for the
pleasure of playing, not to win. That is the first condition of a
pastime.” She told him she was very sorry, and he kissed her. But
after that Villiers used to lay newspapers and reviews on the sitting-
room table while they were dining. She consoled herself with the
remembrance of that kiss on the lips; it was nearly the last of them.
He selected her forehead, from Saltcombe onwards, or her cheek.
From Saltcombe they went down into Cornwall—Truro, Penzance,
Sennen, St. Ives. There it was that she learned to be happy in her
own company. She spent hours alone, scrambling among the rocks,
watching the sea.
Her life was filling, her vistas opening. This was great gain, to
feel the triumph of discovery. She had never been so far afield
before, and the wild splendours of rocks and seas made her at times
like a thing inspired. She was amazed at herself—at the stinging
blood in her which made her heart beat. She used to get up early at
Sennen, steal, hatless, out of the sleeping inn, and fleet over turf to
the edge of the cliffs. There she stood motionless, with unwinking
eyes and parted lips, while the wind enfolded her. All was pure
ecstasy; she was like a nymph—bare-bosomed, ungirdled,
unfilletted, in the close arms of the Country God. From such hasty
blisses she returned drowsy-eyed, glossed with rose-colour, with a
sleek bloom upon her, and ministered to her husband’s needs,
dressed with care, with the neatness which he loved. She sat quietly
by him, hearing but not heeding his measured tones, dreaming of
she knew not what, save that the dreams were lyric, and sang of
freedom in her ears.
They took more tangible shape as they waxed bolder in outline
and scope. There was a tumble-down white cottage on the cliff
beyond the coastguard station; two rooms and a wash-house below
green eaves. It faced the open sea, but lay otherwise snugly below a
jutting boulder, and was so much of a piece with rock and turf that
the sea-pinks had seeded in the roof and encrusted it with emerald
tufts. Her fancy adorned this tenement; she saw herself there in a
cotton gown, alone with wind and sea. What a life! The freedom of
it, the space, the promise! Not a speck could she descry upon the
fair blue field of such a life. Childlike she built upon the airy fabric,
added to it, assured herself of it. Some day, some day she would be
there—free! The thought made her perfectly happy; she felt her
blood glow.
Mr. Germain complained of the damp Cornish air and took her to
St. Ives and Newquay on the way to Southover. Once on the
homeward path, he had no eyes for her in Cornwall; all his hopes
were now set upon the feast he should have of her, queening it
there in his hall—queen by his coronation. She, for her part, was all
for lingering good-byes to her glimpses of the wild. She went
obediently, but carried with her the assurance that she should see
her cottage again; and by some juggling of the mind, in the picture
of it which floated up before her at call, she came to see always near
it the tilt-cart and its occupant, her friend of the open Common. A
community down there! The tilt-cart stood in a hollow of the rocks
within sound and sight of the sea; the Ghost cropped the thyme
above it; Bingo ran barking out of the tent, and, seeing her, lowered
his head and came wriggling for a caress. Above them all, dominant,
stood her friend, bareheaded to the buffeting gale, so clearly at
times that she could see the wind bellying his white trousers or
flacking the points of his rolling collar. His face unfortunately was not
always to be seen; a mist over it baffled her, but egged her on. For a
flash, for a passing second, his bright, quizzing eyes might be upon
her; she could hear the greeting of the dawn laugh from them, and
feel her bosom swell as she answered it, and knew the long day
before them—and every long day to come. What a comradeship that
might be—what a comradeship! She came to thank God daily that
she had such a friend, and to declare stoutly to herself that she had
no need to see him. Friendship was independent of such needs; the
necessities of touching, eyeing, speaking—what were these but
fetters? Lovers might hug such chains and call them leading-strings.
Poor lovers could not walk without them. But friends had their pride
in each other and themselves. Each stood foursquare in the faith of
his friend; the independence of each was the pride of the other. So
far was she from loving Mr. Senhouse that she learned without a
pang of his visit to the Cantacutes in the following summer, of his
painting days with Hertha de Speyne, and was surprised at herself.
It drew the two girls closer together; it gave zest to letter-writing,
and brought Miss Hertha more than once to Southover. Senhouse
was the presiding genius of their fireside talks; between Hertha and
Senhouse Mary began to find herself—a person, with a reasonable
soul in human flesh.
Her wedding-day, and the days that followed it, had dismayed
the flesh; she could not be one to whom marriage was a sacred
mystery, to be unveiled to piercing music. She had cried herself to
sleep—once; but she cried no more. If she had been in love with her
husband, even if she had ever been in love with anybody, she might
have been won over by pity or by passion; but poor Mr. Germain was
incapable of the second, and somewhat to her surprise she found
herself unpersuaded, though she was touched, by the first. She did
pity him, she pitied him deeply, but she could not help him. Esteem
she gave him, gratitude, obedience, meekness, respect. But herself
—after that once—never, never! For that discharging of her
conscience of its poor little trivial, human load had been forced upon
her by pure generosity on her part (she knew it), and had cost her
an agony of shame. And it had chilled him to the bone—she had
seen his passion fade before her eyes, such passion as he had. Her
generosity had stultified her, played the traitor. She never taxed him
with want of magnanimity, didn’t know the word—but she found
herself resolute, and was as much surprised as he was. What dismay
she had, as the honeymoon wore on, was brought her by her own
position, not by her husband’s; that a girl such as she, with
undeniable proofs to hand of her attractiveness of face and person,
with experience of men and their ways, should find herself daughter
to her husband! An indulged, courted, only daughter, if you please—
but certainly a daughter. Here was an anti-climax, to say the least of
it; and her dismay endured through the honeymoon—until Cornish
cliffs gave her happier things to dream of. It disappeared as the
great red flank of Southover House filled up the scene. Tussles with
Mrs. James, the sweets and perils of victory, ordeals of shooting-
parties, dinner-parties, household cares, and, above all, routine—
such drugs as these sent her heart to sleep. By the time she had
been eighteen months a wife she had forgotten that she had never
been other than a maiden.
Now, what of Cratylus, poor Cratylus the mature, who, clasping
his simple Mero (or Marina) to his heart, found that he had to reckon
with her character first? Good, honest man, he had never supposed
her to have one; and the bitter thing was that the finding of her
character woke up his own. He saw himself again in full plate-
armour, cowering behind it, hiding from himself as well as from the
world a terrible deformity—an open sore in his self-esteem which
could never be healed again, which, at every chance of her daily life,
must bleed and ache. Oh, the pity of it, on how light a spring all this
had depended—a hair, a gossamer! Exeter—fatal day of Exeter! He
had believed himself young again. As she clung to him, half-sobbing,
after dinner, he had pressed her to his bosom, called her his bride,
his wife. She had not dared to look at him, had bowed her head,
hidden her face in his shoulder, let him feel the trembling, the wild
beating of her heart. Then her broken confessions; pitiful, pitiful!
What did they amount to, when all was told? But they, and what
followed upon them—his own conduct, his own curse; and her
conduct, and her curse—were his nightmare. He had found out that
he could not live if he must remember them. He fought, literally, for
life; and after a six months’ toil had succeeded in living. He spent
himself in benevolence and care, gave her everything she could
want, before she asked, taught her, prayed for her, watched over
her. She was never out of his thoughts—and, poor girl, without
knowing it, she stabbed him deeply every day.
He had his benevolence to fall back upon. He could be King of
Southover, of the Cophetua dynasty; he could dazzle her, take her
breath away, and have the delight, which he had promised himself,
of seeing her misty eyes and cheeks flushed with wonder. Yes, yes;
but the æsthetic nerve, you see, dulls with use, and the worst of a
king’s homage to a beggar maid is that the more obsequious the
homage the less beggar is the maid. If you set a coronet in her hair
she will blush deliciously for a week; but in two years’ time it will be
there as a matter of course, put there nightly by her woman—and
bang goes your joy of that. So with all the other enrichments of
society, travel, book-learning. The more she had of them, the more
she was able to take for herself. He who put her in the way of
knowledge could not grumble if she acted upon what he had taught
her. Such gifts as his destroy themselves. It had filled his eyes with
tears to see his wilding in the great terraced house, to watch the
little airs of dignity of matronhood, wifehood (alas, poor gentleman!)
flutter about her, and, like birds, take assurance, and alight. Her
cares were charming, too. It was pretty to see her knit her brows
over some tough nugget of Dante’s, exquisite when she came
faltering to him, coaxing for help. But then, naturally, the more help
she had the less she came. It grew to be her pride to get through
alone—her pride and his disaster. No. Tristram Duplessis had been
wiser in his generation than he. If you love to fill a thing you must
take care to keep it pretty empty. Thus it was that King Cophetua
kneeled in vain. He had kneeled too low.
But there’s a balm in the passing years for Cratylus as well as for
Marina. The musical clockwork of Southover, which he had promised
himself, became his. He went about his duties as landlord, county
magnate, patron of reasonable things, tolerably sure of a welcome
home from a pair of kind brown eyes. Kisses might be his if he chose
to call for them, clinging arms, a warm and grateful heart. Such
things had to be his solace; and sometimes they were. And he still
fought for his treasure, against all the odds, with his teeth set hard.
If he had lost grip it was because her muscles were more practised.
He must try another, and another, if he would whirl her in the air. He
must impress her anew, prove to her that he was a man, honour-
worthy and loveworthy. His ambitions were rekindled: that was the
result of his musings. In the spring of the year, when the tulips
blazed in the Italian gardens, and Mary Middleham had been Mary
Germain for a good eighteen months, we heard him speak with
young Mr. Wilbraham of Sir Gregory and the Farlingbridge division of
the county. There was a chance of lighting up the wonder again in a
pair of brown eyes. He hoarded the thought for the month, and by
June had made up his mind. Then he broke it to his Mary. “I will
gladly put my experience at the service of the country,” he told her,
“and convince you, if I can, that I am not too old for a public career.”
She had told him that he wasn’t old at all, and had kissed his
forehead. They happened to be alone for a few days just then; so
that he could draw her down to his knee and talk to her about
himself, and the part she would have to play for him in London. The
house in Hill-street must be reopened.
 III
MATTERS OF ELECTION

The country showed the periodic symptoms the moment

Parliament was dissolved; the market-place of Farlingbridge hummed
with rumour. Farmers in gigs pulled up to discuss the affairs of the
nation with farmers on horseback, with hedgers, ditchers, tinkers,
anybody. Class flowed over class, and The Reverend Stephen
Burgess, Vicar of the town, exchanged evening papers with
Reverend Samuel Rock, Congregational minister. Blue and red were
in the air; Mr. Germain had long sittings with young Mr. Wilbraham.
Presently a deputation attended, by leave, at Southover and was
received in the library. The seat was to be contested, it seemed; the
Honourable Leopold Levitt intended to fight. Now would Mr. Germain
fight him? In a weighed speech of twenty minutes Mr. Germain
declared his loins to be girded. “Pompous old boy,” said Mr. Tom
Blyth, the Liberal harness-maker, to Mr. Peake, the Liberal agent;
“but he’s good enough for the Honourable Levitt.” Mr. Peake thought
he was just good enough. It was to be a narrow thing, a close-run
thing. The addresses of the candidates showed as much. “Those
great institutions to which this country of ours—” cried the
Honourable Mr. Levitt in ink. “Those institutions to which this great
country of ours—” was the peroration of Mr. Germain. There was to
be very little in it. Mr. Peake, the agent, said that the ladies would do
the trick. The Honourable Mrs. Levitt was stout, and twenty years
older than her Leopold.
The writs were out in August, the election was to be in October.
Mary, who had begun to lose colour during the summer heats, grew
animated again at the prospect of the bustle. She had been getting
introspective, too, had been sometimes fretful, sometimes glum. She
thought more than was good for her about things which could not
be helped. But for a flying Sunday visit, she had not seen her own
people since her wedding day; for Mr. Germain had given up
Misperton once more, and seemed to dislike the notion of her
leaving him at such a time as this. Here, then, was a chance for her
to be useful. She told her husband that she felt sure of
Farlingbridge, and when he shook his head despondent she told him
why. “They like me there, you know,” she said, blushing and
laughing. “I know they do; besides Mrs. Blyth told me so. Oh, and
Colonel Dermott stopped me yesterday and said that you might be
easy. He’ll speak for you wherever you want him.” Colonel Dermott
was an introduction of hers to the penetralia of Southover; a fiery
Irishman with a turn for sarcasm. What he had really said to her was
that he’d go to the stake for her, but that it wouldn’t be necessary.
He admired her unaffectedly.
As the campaign progressed on its roaring way Mr. Germain
became conscious that greater efforts than his own were necessary.
The Honourable Mr. Levitt was untiring. He drove his own drag, and
seemed to have a speech on the tip of his tongue for every village
green. To Mr. Germain speeches were matters of enormous
preparation, literary and economic. He balanced his periods as
carefully as his convictions; he polished them, gave them form; but
he could not fire them, because he had no fire. “We must give it ’em
hot, Mr. Germain, we must indeed,” said Peake, the agent. Mr.
Germain knew very well that he gave it them cold. The charming
spectacle of his young wife, in red cloth, driving her ponies in red
harness, a red bow on her whip and red roses in her bosom, far
from kindling him, whitened the ashes of his hearth. She was pretty,
she was gay, she went again and again to the attack, and coaxed for
votes as a child for sweets. One great sensation was when Guy
Chaveney ratted, and wore red; another when Levitt publicly alluded
to her as his “fair enemy,” and was drowned in the cheers of his own
party. Colonel Dermott swept her into debate with his hand. “Here’s
the lady we follow and serve, gentlemen,” and he turned to her
where she sat glowing on the platform. “By the powers, gentlemen,
I’d run her up to Westminster by myself,” he went on; “but we’ll
share in the enterprise, if you please.” A little more of that and we
were in, said Mr. Peake.
 Help from on high was promised, of an exciting kind. The Right
Honourable Constantine Jess, President once of the Board of Trade,
now Secretary of State apparent, offered to come down and help his
old friend. He offered, I assume, in such a way that he could not be
refused; for his approach was announced to Mary one morning over
the breakfast table, and received by her with the calmness proper to
county ladies. But there had been more. “He brings Tristram, his
private secretary. You remember Tristram Duplessis, Mary?”
She managed it. “Yes,” she said. “I remember him very well.”
Mrs. Hartley—Mrs. Leonard Hartley, I mean—said that she had
heard him speak. He reminded her of Mirabeau. Sir James Plash had
asked, “Which Mirabeau?” and driven Mrs. Leonard into a corner.
“Oh, Carlyle’s, of course!” she answered—and the talk flowed
over Tristram Duplessis.
But behind her fortification of silver urns and coffee-cups, she did
remember him. Her eyes wide, sombre, and brooding, made no sign.
It is the prerogative of county eyes to be still, and of married eyes to
be indifferent. She did not smile at her thoughts, nor betray that
they were not of a smiling kind. But she felt her heart quicken its
beat, knew that she was to be put to the proof, and that her
husband had chosen it to be so. To the racing rhythm in her head
ran the refrain, “I knew he must come. He never forgets.”
Notes must be written and answers received. His was very short:
—
“Dear Mrs. Germain,—I am very glad to come and help you.
Certainly, we must bring him in. Yrs. sincerely, Tr. Duplessis.” It
required sharp scrutiny to read between the lines of such a letter,
and sharp scrutiny was applied—more than once. She pinched her
lip over it as she sat alone, and carried it with her as she walked the
park—but when she found herself doing that she tore it up. “I am
very glad to come and help you”; that “you” was an after-thought.
“Certainly, we must bring him in”; that “we” proved it. She knew,
better than most, how Tristram could imply himself in a note. He had
forgotten nothing, never would forget anything. No! No more had
she forgotten.
 Of all her former lovers this was the one man who could cause
her any disquiet, or have evoked any sensation. She could never
have recalled herself as she had been, two years ago, by any other
aid than his. John Rudd? Ambrose Perivale? It is doubtful if she
would have known them again. Sharper memories, a sharper
fragrance clung about Tristram. Of all of them, it was with him that
her relations had been the least explicit; but it had been he, also,
who had thereby implied the most. He was master of implication—
that delicate art which leaves it to the imagination of the object to
read what precisely is implied. Had Tristram implied love? She never
knew: that made Tristram’s dealing so exciting. Of course he had
admired her; his savage looks, as if she stung and vexed him, had
assured her of that. Her presence—her near presence—seemed
always to make him angry; her absence angrier still, since he always
came after her, and never forgot to let her see how angry he had
been. Yes, he admired her; but admired other things more, much
more: his books, his scholarship, the power he had, and, vastly
more, himself. He was endlessly interested in himself, only “liked”
her as showing him himself in new aspects; but she accepted that as
a part of him, like the cut of his clothes; and there was no doubt as
to her own feeling; she had admired Tristram on this side adultary,
just on this side. Tristram intended to be Somebody: he used to tell
her so, in a way which made her understand that he knew her to be
a little Nobody. All the same, he couldn’t keep his eyes off her, or his
steps from turning to where she was, or was like to be. In a sense,
then, she had drawn Tristram Duplessis; and that’s an exciting thing
for a little nobody to do.
If he had been her lover as well, it had been in a way of his own.
He had told her often and often that he disapproved of her—of her
too speaking eyes, for instance (which could and did speak in those
days), of her little affairs with Dick or Tom, as to which he had given
himself the trouble to be exactly informed; of her lack of ambition;
and because she was a dunce. And she had laughed or blushed, or
been offended—she had never been hurt; and had allowed herself to
be put under the rod of his tongue, or the gibe of his eyes again and
again. She thought now—with hot cheeks—that she ought to have
felt herself insulted, and, with hotter cheeks, that it was doubtful
even now whether she would feel herself so. To have a book thrown
into her lap, with the inference that she couldn’t read it; to be kissed
without leave asked, or to kiss again without notice taken—these
should have insulted Mary Middleham: but would they insult Mary
Germain? Tingling cheeks were no answer.
Tristram had indeed been very exciting; he had been
unaccountable, arbitrary, splendid; to have attracted his scowling
looks had been one of her triumphs. It had been a triumph, even,
that Misperton Brand knew all about it, and that part of it had been
scandalized. Yet—and for all that—thinking over it now, with his
coming again so close at hand, she knew perfectly well that she had
not been in love with him, and was not in love with him now. He had
treated her in too lordly a fashion altogether. Dimly she could guess
that love was another affair. It might be possible for a girl to worship
a man as a god—but that was never love. She knew better than that
now. But certainly she must confess, even now, to a tenderness for
her reminiscences of Tristram, who had singled her out of a herd,
watched, followed, engrossed her, and in his own morose and
grudging way had seemed to be in love with her. He had known how
to kiss, anyhow. As she inhaled the sharp fragrance of those days
she was again excited. There had been glamour. She recalled, with a
thrill, the Sunday afternoon when Mrs. James had caught him
reading Shelley to her under the apple tree, and blushed anew as
she had blushed then. And the continuous alarm of the affair! The
moments snatched in pauses of the chase! Yes, there had been
glamour, and it had been sweet—perilous and sweet. It was a thing
to remember, but not to fear. She didn’t think she need fear
anything, especially as she had told Mr. Germain all about it—or as
good as told him.
But it’s always ticklish work, meeting an old heart’s acquaintance
on new terms. Neither party to the business can face it quite
unmoved. For him, there’s the painful, curious inquiry:—“This, this is
she with whom I had fondly hoped—! Now, look, there is knowledge
stored within those limpid eyes—and I might have put it there! She
and I share experiences, which He—that interloper—can never
share. With this I must dress my wounded side.” All that his
handshake, or his bow, may convey to her. Upon her side—the
sedately conscious of two men’s regard—veiled within her eyes
there’s this for the ousted lover: “You may spare me the rod. I am
another’s, who might have been yours. You loved me once, you told
me; be charitable now!” And all that she will express in the flutter of
her greeting.
Tristram Duplessis, loose-limbed, flushed, frowning as of old, may
have implied it, or she, who played him hostess of Southover, may
have appealed in that fashion. “How d’you do?” was what he said in
words, when he took her hand, which she held out, in a nerveless
clasp. He had arrived late in the afternoon, when the hall was fully
occupied; stockinged young men, in from shooting, short-kirtled
ladies, in from getting in their way; a dowager or two reading
evening papers, and a whiskered professor in slippers. One must
imply skilfully in such a company.
And then, to be sure, there was Mr. Constantine Jess, ponderous,
benevolent, all for domesticity, to be reckoned with. All women liked
Mr. Jess because, although he was prodigiously learned, he owned
to a weakness for small talk and soft voices. It was he, then, who
had the triumph of the entry. “Ah, Mrs. Germain, this is a welcome
indeed. And doth not a meeting like this make amends?” His quick,
full-cushioned eyes swept the corners of the room—“My dear Lady
Barbara—! Lady Wentrode, your servant—How d’ye do? How d’ye
do?” These things accomplished, he turned to his hostess, cup in
hand, and sank into the cushions by her side. “We have not met, I
think, since that auspicious day—two years ago? Is it that? Dear me,
how Time makes sport with us! One should hear the Titans laugh. I
had promised myself an earlier contemplation of your felicity, but—
business! business!” He sighed, drained his teacup, and asked for
more. “It must have been within a week of your marriage that my
young friend and I took a fancy for each other. A marriage of minds!
Tristram, my dear fellow, when was it?” He had taught his secretary
the duty of playing chorus. That was very necessary to Mr. Jess.
Tristram, leisurely, as of old, sipped his tea before answering, got
up and waited for another cup while he collected his reply. “It’s a
long time—I know that. Thanks, no sugar.”
“Oh, I’m so sorry—I forgot.” She looked up at him hazardously.
“You always took it, I thought?”
“I know. But I’ve dropped all that. Better without it.” He spoke
lightly.
The campaign was broached by Mr. Jess. “Well, and how does
my friend in the field? Gallantly, I am sure; happily, I hope.”
Mrs. Germain thought that he would win. “He works very hard.
He’s speaking now, somewhere—out of the carriage, at a harvest
home. I ought to have been with him, but——”
“You stayed to be hospitable to us. We are grateful. At least, I
speak for myself. Tristram there takes kindness for granted.”
“Not Mrs. Germain’s,” said Tristram, and startled her.
However, she laughed. “I don’t think it was very kind of me; I
was glad to be let off. I’m sure everything will go right now. Did you
know that you must begin to-night, Mr. Jess? Do you mind? There’s
a meeting at the Corn Exchange at eight. We are dining early.”
Mr. Jess laid his comfortable hand upon his heart. “I follow my
leader. Where she calls me I am ever to be found.”
And then she raised her eyes to Tristram. “Will you speak for us,
Mr. Duplessis?” He started, as out of a stare.
“Who? I? Oh, I’ll do as I’m bid, of course.”
“Enlist him, my dear lady, enrol him,” cried Mr. Jess, twinkling,
“but if you love me, let him follow me. He has a note like a trumpet.”
“Really?” She opened her eyes upon Tristram.
“I can make a row,” he admitted. “But perhaps Germain won’t
like that.”
“I am sure he will like whatever you do,” said she. Duplessis
made no answer, but did not shirk the reflection that, if he did, it
would indicate a striking change in the gentleman’s views.
At this moment a fair-haired young lady in a riding-habit—Miss
Nina Swetebrede of Copestake—came in, craving tea. She
distributed her nods and smiles on either hand as she advanced to
the table. “Dear Mary, I’m so tired,” she pleaded. “Do feed me, and
make a fuss of me, and I shall love you.” The newly arrived
gentlemen were made known to her, and Mr. Jess courtly and
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com