Model Predictive Control Theory Computation and Design 2Nd Edition Rawlings James B Full Chapter

Model Predictive Control: Theory,
Computation, and Design, 2nd Edition

Rawlings James B.
Visit to download the full and correct content document:
https://ebookmass.com/product/model-predictive-control-theory-computation-and-desi
gn-2nd-edition-rawlings-james-b/
Model Predictive Control:
Theory, Computation, and Design
2nd Edition
ISBN 9780975937730
9 780975 937730
Model Predictive Control:
Theory, Computation, and Design
2nd Edition
James B. Rawlings
Department of Chemical Engineering
University of California
Santa Barbara, California, USA
David Q. Mayne
Department of Electrical and Electronic Engineering
Imperial College London
London, England
Moritz M. Diehl
Department of Microsystems Engineering and
D
Department of Mathematics
University of Freiburg
Freiburg, Germany
b Hi
No ll Publishing
Santa Barbara, California

This book was set in Lucida using LATEX, and printed and bound by
Worzalla. It was printed on Forest Stewardship CouncilÉ certified acid-
free recycled paper.
Cover design by Cheryl M. and James B. Rawlings.
Copyright c 2019 by Nob Hill Publishing, LLC
All rights reserved.
Nob Hill Publishing, LLC

Cheryl M. Rawlings, publisher
Santa Barbara, CA 93117
orders@nobhillpublishing.com
http://www.nobhillpublishing.com
No part of this book may be reproduced, in any form or by any means,

without permission in writing from the publisher.
Library of Congress Control Number: 2017909542
Printed in the United States of America.
First Edition
First Printing August 2009
Electronic Download November 2013
Electronic Download (2nd) April 2014
Electronic Download (3rd) July 2014
Electronic Download (4th) October 2014
Electronic Download (5th) February 2015
Second Edition
First Printing October 2017
Electronic Download October 2018
Electronic Download (2nd) February 2019
To Cheryl, Josephine, and Stephanie,
for their love, encouragement, and patience.

Preface to the Second Edition
In the eight years since the publication of the first edition, the field
of model predictive control (MPC) has seen tremendous progress. First
and foremost, the algorithms and high-level software available for solv-
ing challenging nonlinear optimal control problems have advanced sig-
nificantly. For this reason, we have added a new chapter, Chapter 8,
“Numerical Optimal Control,” and coauthor, Professor Moritz M. Diehl.
This chapter gives an introduction into methods for the numerical so-
lution of the MPC optimization problem. Numerical optimal control
builds on two fields: simulation of differential equations, and numeri-
cal optimization. Simulation is often covered in undergraduate courses
and is therefore only briefly reviewed. Optimization is treated in much
more detail, covering topics such as derivative computations, Hessian
approximations, and handling inequalities. Most importantly, the chap-
ter presents some of the many ways that the specific structure of opti-
mal control problems arising in MPC can be exploited algorithmically.
We have also added a software release with the second edition of
the text. The software enables the solution of all of the examples and
exercises in the text requiring numerical calculation. The software is
based on the freely available CasADi language, and a high-level set of
Octave/MATLAB functions, MPCTools, to serve as an interface to CasADi.
These tools have been tested in several MPC short courses to audiences
composed of researchers and practitioners. The software can be down-
loaded from www.chemengr.ucsb.edu/~jbraw/mpc.
In Chapter 2, we have added sections covering the following topics:
• economic MPC
• MPC with discrete actuators
We also present a more recent form of suboptimal MPC that is prov-
ably robust as well as computationally tractable for online solution of
nonconvex MPC problems.
In Chapter 3, we have added a discussion of stochastic MPC, which
has received considerable recent research attention.
In Chapter 4, we have added a new treatment of state estimation
with persistent, bounded process and measurement disturbances. We
have also removed the discussion of particle filtering. There are two
vi
vii
reasons for this removal; first, we wanted to maintain a manageable

total length of the text; second, all of the available sampling strate-
gies in particle filtering come up against the “curse of dimensionality,”
which renders the state estimates inaccurate for dimension higher than
about five. The material on particle filtering remains available on the
text website.
In Chapter 6, we have added a new section for distributed MPC of
nonlinear systems.
In Chapter 7, we have added the software to compute the critical
regions in explicit MPC.
Throughout the text, we support the stronger KL-definition of asymp-
totic stability, in place of the classical definition used in the first edition.
The most significant notational change is to denote a sequence with
(a, b, c, . . .) instead of with {a, b, c, . . .} as in the first edition.
JBR DQM MMD

Madison, Wis., USA London, England Freiburg, Germany
Preface
Our goal in this text is to provide a comprehensive and foundational

treatment of the theory and design of model predictive control (MPC).
By now several excellent monographs emphasizing various aspects of
MPC have appeared (a list appears at the beginning of Chapter 1), and
the reader may naturally wonder what is offered here that is new and
different. By providing a comprehensive treatment of the MPC foun-
dation, we hope that this text enables researchers to learn and teach
the fundamentals of MPC without continuously searching the diverse
control research literature for omitted arguments and requisite back-
ground material. When teaching the subject, it is essential to have a
collection of exercises that enables the students to assess their level of
comprehension and mastery of the topics. To support the teaching and
learning of MPC, we have included more than 200 end-of-chapter exer-
cises. A complete solution manual (more than 300 pages) is available
for course instructors.
Chapter 1 is introductory. It is intended for graduate students in en-
gineering who have not yet had a systems course. But it serves a second
purpose for those who have already taken the first graduate systems
course. It derives all the results of the linear quadratic regulator and
optimal Kalman filter using only those arguments that extend to the
nonlinear and constrained cases to be covered in the later chapters.
Instructors may find that this tailored treatment of the introductory
systems material serves both as a review and a preview of arguments
to come in the later chapters.
Chapters 2–4 are foundational and should probably be covered in
any graduate level MPC course. Chapter 2 covers regulation to the ori-
gin for nonlinear and constrained systems. This material presents in a
unified fashion many of the major research advances in MPC that took
place during the last 20 years. It also includes more recent topics such
as regulation to an unreachable setpoint that are only now appearing in
the research literature. Chapter 3 addresses MPC design for robustness,
with a focus on MPC using tubes or bundles of trajectories in place of
the single nominal trajectory. This chapter again unifies a large body of
research literature concerned with robust MPC. Chapter 4 covers state
estimation with an emphasis on moving horizon estimation, but also
viii
ix
covers extended and unscented Kalman filtering, and particle filtering.

Chapters 5–7 present more specialized topics. Chapter 5 addresses
the special requirements of MPC based on output measurement instead
of state measurement. Chapter 6 discusses how to design distributed
MPC controllers for large-scale systems that are decomposed into many
smaller, interacting subsystems. Chapter 7 covers the explicit optimal
control of constrained linear systems. The choice of coverage of these
three chapters may vary depending on the instructor’s or student’s own
research interests.
Three appendices are included, again, so that the reader is not sent
off to search a large research literature for the fundamental arguments
used in the text. Appendix A covers the required mathematical back-
ground. Appendix B summarizes the results used for stability analysis
including the various types of stability and Lyapunov function theory.
Since MPC is an optimization-based controller, Appendix C covers the
relevant results from optimization theory. In order to reduce the size
and expense of the text, the three appendices are available on the web:
www.chemengr.ucsb.edu/~jbraw/mpc. Note, however, that all mate-
rial in the appendices is included in the book’s printed table of contents,
and subject and author indices. The website also includes sample ex-
ams, and homework assignments for a one-semester graduate course
in MPC. All of the examples and exercises in the text were solved with
Octave. Octave is freely available from www.octave.org.
JBR DQM
Madison, Wisconsin, USA London, England
Acknowledgments
Both authors would like to thank the Department of Chemical and Bio-
logical Engineering of the University of Wisconsin for hosting DQM’s
visits to Madison during the preparation of this monograph. Funding
from the Paul A. Elfers Professorship provided generous financial sup-
port.
JBR would like to acknowledge the graduate students with whom
he has had the privilege to work on model predictive control topics:
Rishi Amrit, Dennis Bonné, John Campbell, John Eaton, Peter Findeisen,
Rolf Findeisen, Eric Haseltine, John Jørgensen, Nabil Laachi, Scott Mead-
ows, Scott Middlebrooks, Steve Miller, Ken Muske, Brian Odelson, Mu-
rali Rajamani, Chris Rao, Brett Stewart, Kaushik Subramanian, Aswin
Venkat, and Jenny Wang. He would also like to thank many colleagues
with whom he has collaborated on this subject: Frank Allgöwer, Tom
Badgwell, Bhavik Bakshi, Don Bartusiak, Larry Biegler, Moritz Diehl,
Jim Downs, Tom Edgar, Brian Froisy, Ravi Gudi, Sten Bay Jørgensen,
Jay Lee, Fernando Lima, Wolfgang Marquardt, Gabriele Pannocchia, Joe
Qin, Harmon Ray, Pierre Scokaert, Sigurd Skogestad, Tyler Soderstrom,
Steve Wright, and Robert Young.
DQM would like to thank his colleagues at Imperial College, espe-
cially Richard Vinter and Martin Clark, for providing a stimulating and
congenial research environment. He is very grateful to Lucien Polak
and Graham Goodwin with whom he has collaborated extensively and
fruitfully over many years; he would also like to thank many other col-
leagues, especially Karl Åström, Roger Brockett, Larry Ho, Petar Koko-
tovic, and Art Krener, from whom he has learned much. He is grateful
to past students who have worked with him on model predictive con-
trol: Ioannis Chrysochoos, Wilbur Langson, Hannah Michalska, Sasa
Raković, and Warren Schroeder; Hannah Michalska and Sasa Raković,
in particular, contributed very substantially. He owes much to these
past students, now colleagues, as well as to Frank Allgöwer, Rolf Find-
eisen Eric Kerrigan, Konstantinos Kouramus, Chris Rao, Pierre Scokaert,
and Maria Seron for their collaborative research in MPC.
Both authors would especially like to thank Tom Badgwell, Bob Bird,
Eric Kerrigan, Ken Muske, Gabriele Pannocchia, and Maria Seron for
their careful and helpful reading of parts of the manuscript. John Eaton
x
xi
again deserves special mention for his invaluable technical support dur-
ing the entire preparation of the manuscript.
Added for the second edition. JBR would like to acknowledge the
most recent generation of graduate students with whom he has had the
privilege to work on model predictive control research topics: Doug Al-
lan, Travis Arnold, Cuyler Bates, Luo Ji, Nishith Patel, Michael Risbeck,
and Megan Zagrobelny.
In preparing the second edition, and, in particular, the software re-
lease, the current group of graduate students far exceeded expectations
to help finish the project. Quite simply, the project could not have been
completed in a timely fashion without their generosity, enthusiasm,
professionalism, and selfless contribution. Michael Risbeck deserves
special mention for creating the MPCTools interface to CasADi, and
updating and revising the tools used to create the website to distribute
the text- and software-supporting materials. He also wrote code to cal-
culate explicit MPC control laws in Chapter 7. Nishith Patel made a
major contribution to the subject index, and Doug Allan contributed
generously to the presentation of moving horizon estimation in Chap-
ter 4.
A research leave for JBR in Fall 2016, again funded by the Paul A.
Elfers Professorship, was instrumental in freeing up time to complete
the revision of the text and further develop computational exercises.
MMD wants to especially thank Jesus Lago Garcia, Jochem De Schut-
ter, Andrea Zanelli, Dimitris Kouzoupis, Joris Gillis, Joel Andersson,
and Robin Verschueren for help with the preparation of exercises and
examples in Chapter 8; and also wants to acknowledge the following
current and former team members that contributed to research and
teaching on optimal and model predictive control at the Universities of
Leuven and Freiburg: Adrian Bürger, Hans Joachim Ferreau, Jörg Fis-
cher, Janick Frasch, Gianluca Frison, Niels Haverbeke, Greg Horn, Boris
Houska, Jonas Koenemann, Attila Kozma, Vyacheslav Kungurtsev, Gio-
vanni Licitra, Rien Quirynen, Carlo Savorgnan, Quoc Tran-Dinh, Milan
Vukov, and Mario Zanon. MMD also wants to thank Frank Allgöwer, Al-
berto Bemporad, Rolf Findeisen, Larry Biegler, Hans Georg Bock, Stephen
Boyd, Sébastien Gros, Lars Grüne, Colin Jones, John Bagterp Jørgensen,
Christian Kirches, Daniel Leineweber, Katja Mombaur, Yurii Nesterov,
Toshiyuki Ohtsuka, Goele Pipeleers, Andreas Potschka, Sebastian Sager,
Johannes P. Schlöder, Volker Schulz, Marc Steinbach, Jan Swevers, Phil-
ippe Toint, Andrea Walther, Stephen Wright, Joos Vandewalle, and Ste-
fan Vandewalle for inspiring discussions on numerical optimal control
xii
methods and their presentation during the last 20 years.

All three authors would especially like to thank Joel Andersson and
Joris Gillis for having developed CasADi and for continuing its support,
and for having helped to improve some of the exercises in the text.
Contents
1 Getting Started with Model Predictive Control 1

1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Models and Modeling . . . . . . . . . . . . . . . . . . . . . . 1
1.2.1 Linear Dynamic Models . . . . . . . . . . . . . . . . . 2
1.2.2 Input-Output Models . . . . . . . . . . . . . . . . . . 3
1.2.3 Distributed Models . . . . . . . . . . . . . . . . . . . 4
1.2.4 Discrete Time Models . . . . . . . . . . . . . . . . . . 5
1.2.5 Constraints . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.6 Deterministic and Stochastic . . . . . . . . . . . . . 9
1.3 Introductory MPC Regulator . . . . . . . . . . . . . . . . . . 11
1.3.1 Linear Quadratic Problem . . . . . . . . . . . . . . . 11
1.3.2 Optimizing Multistage Functions . . . . . . . . . . . 12
1.3.3 Dynamic Programming Solution . . . . . . . . . . . 18
1.3.4 The Infinite Horizon LQ Problem . . . . . . . . . . . 21
1.3.5 Controllability . . . . . . . . . . . . . . . . . . . . . . 23
1.3.6 Convergence of the Linear Quadratic Regulator . . 24
1.4 Introductory State Estimation . . . . . . . . . . . . . . . . . 26
1.4.1 Linear Systems and Normal Distributions . . . . . 27
1.4.2 Linear Optimal State Estimation . . . . . . . . . . . 29
1.4.3 Least Squares Estimation . . . . . . . . . . . . . . . 33
1.4.4 Moving Horizon Estimation . . . . . . . . . . . . . . 39
1.4.5 Observability . . . . . . . . . . . . . . . . . . . . . . . 41
1.4.6 Convergence of the State Estimator . . . . . . . . . 43
1.5 Tracking, Disturbances, and Zero Offset . . . . . . . . . . 46
1.5.1 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.2 Disturbances and Zero Offset . . . . . . . . . . . . . 49
1.6 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
2 Model Predictive Control—Regulation 89

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
2.2 Model Predictive Control . . . . . . . . . . . . . . . . . . . . 91
2.3 Dynamic Programming Solution . . . . . . . . . . . . . . . 107
2.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
2.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 112
xiii
xiv Contents
2.4.2 Stabilizing Conditions . . . . . . . . . . . . . . . . . 114

2.4.3 Exponential Stability . . . . . . . . . . . . . . . . . . 120
2.4.4 Controllability and Observability . . . . . . . . . . . 120
2.4.5 Time-Varying Systems . . . . . . . . . . . . . . . . . 123
2.5 Examples of MPC . . . . . . . . . . . . . . . . . . . . . . . . . 131
2.5.1 The Unconstrained Linear Quadratic Regulator . . 132
2.5.2 Unconstrained Linear Periodic Systems . . . . . . . 133
2.5.3 Stable Linear Systems with Control Constraints . 134
2.5.4 Linear Systems with Control and State Constraints 136
2.5.5 Constrained Nonlinear Systems . . . . . . . . . . . 139
2.5.6 Constrained Nonlinear Time-Varying Systems . . . 141
2.6 Is a Terminal Constraint Set Xf Necessary? . . . . . . . . 144
2.7 Suboptimal MPC . . . . . . . . . . . . . . . . . . . . . . . . . 147
2.7.1 Extended State . . . . . . . . . . . . . . . . . . . . . . 150
2.7.2 Asymptotic Stability of Difference Inclusions . . . 150
2.8 Economic Model Predictive Control . . . . . . . . . . . . . 153
2.8.1 Asymptotic Average Performance . . . . . . . . . . 155
2.8.2 Dissipativity and Asymptotic Stability . . . . . . . 156
2.9 Discrete Actuators . . . . . . . . . . . . . . . . . . . . . . . . 160
2.10 Concluding Comments . . . . . . . . . . . . . . . . . . . . . 163
2.11 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 166
2.12 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
3 Robust and Stochastic Model Predictive Control 193

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
3.1.1 Types of Uncertainty . . . . . . . . . . . . . . . . . . 193
3.1.2 Feedback Versus Open-Loop Control . . . . . . . . 195
3.1.3 Robust and Stochastic MPC . . . . . . . . . . . . . . 200
3.1.4 Tubes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202
3.1.5 Difference Inclusion Description of Uncertain Sys-
tems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
3.2 Nominal (Inherent ) Robustness . . . . . . . . . . . . . . . . 204
3.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 204
3.2.2 Difference Inclusion Description of Discontinu-
ous Systems . . . . . . . . . . . . . . . . . . . . . . . 206
3.2.3 When Is Nominal MPC Robust? . . . . . . . . . . . . 207
3.2.4 Robustness of Nominal MPC . . . . . . . . . . . . . 209
3.3 Min-Max Optimal Control: Dynamic Programming Solution 214
3.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 214
3.3.2 Properties of the Dynamic Programming Solution 216
Contents xv
3.4 Robust Min-Max MPC . . . . . . . . . . . . . . . . . . . . . . 220

3.5 Tube-Based Robust MPC . . . . . . . . . . . . . . . . . . . . 223
3.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 223
3.5.2 Outer-Bounding Tube for a Linear System with
Additive Disturbance . . . . . . . . . . . . . . . . . . 224
3.5.3 Tube-Based MPC of Linear Systems with Additive
Disturbances . . . . . . . . . . . . . . . . . . . . . . . 228
3.5.4 Improved Tube-Based MPC of Linear Systems with
Additive Disturbances . . . . . . . . . . . . . . . . . 234
3.6 Tube-Based MPC of Nonlinear Systems . . . . . . . . . . . 236
3.6.1 The Nominal Trajectory . . . . . . . . . . . . . . . . 238
3.6.2 Model Predictive Controller . . . . . . . . . . . . . . 238
3.6.3 Choosing the Nominal Constraint Sets Ū and X̄ . . 242
3.7 Stochastic MPC . . . . . . . . . . . . . . . . . . . . . . . . . . 246
3.7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 246
3.7.2 Stabilizing Conditions . . . . . . . . . . . . . . . . . 248
3.7.3 Stochastic Optimization . . . . . . . . . . . . . . . . 248
3.7.4 Tube-Based Stochastic MPC for Linear Constrained
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 249
3.8 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
3.9 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
4 State Estimation 269

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
4.2 Full Information Estimation . . . . . . . . . . . . . . . . . . 269
4.2.1 State Estimation as Optimal Control of Estimate
Error . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
4.2.2 Duality of Linear Estimation and Regulation . . . . 281
4.3 Moving Horizon Estimation . . . . . . . . . . . . . . . . . . 283
4.3.1 Zero Prior Weighting . . . . . . . . . . . . . . . . . . 283
4.3.2 Nonzero Prior Weighting . . . . . . . . . . . . . . . . 287
4.3.3 Constrained Estimation . . . . . . . . . . . . . . . . 294
4.3.4 Smoothing and Filtering Update . . . . . . . . . . . 295
4.4 Bounded Disturbances . . . . . . . . . . . . . . . . . . . . . 300
4.5 Other Nonlinear State Estimators . . . . . . . . . . . . . . . 308
4.5.1 Particle Filtering . . . . . . . . . . . . . . . . . . . . . 308
4.5.2 Extended Kalman Filtering . . . . . . . . . . . . . . . 309
4.5.3 Unscented Kalman Filtering . . . . . . . . . . . . . . 310
4.5.4 EKF, UKF, and MHE Comparison . . . . . . . . . . . 312
4.6 On combining MHE and MPC . . . . . . . . . . . . . . . . . 318
xvi Contents
4.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

4.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
5 Output Model Predictive Control 339

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
5.2 A Method for Output MPC . . . . . . . . . . . . . . . . . . . 341
5.3 Linear Constrained Systems: Time-Invariant Case . . . . 344
5.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . 344
5.3.2 State Estimator . . . . . . . . . . . . . . . . . . . . . . 344
5.3.3 Controlling x̂ . . . . . . . . . . . . . . . . . . . . . . . 346
5.3.4 Output MPC . . . . . . . . . . . . . . . . . . . . . . . . 348
5.3.5 Computing the Tightened Constraints . . . . . . . 352
5.4 Linear Constrained Systems: Time-Varying Case . . . . . 353
5.5 Offset-Free MPC . . . . . . . . . . . . . . . . . . . . . . . . . 353
5.5.1 Estimation . . . . . . . . . . . . . . . . . . . . . . . . . 355
5.5.2 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . 356
5.5.3 Convergence Analysis . . . . . . . . . . . . . . . . . 360
5.6 Nonlinear Constrained Systems . . . . . . . . . . . . . . . . 363
5.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
5.8 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 366
6 Distributed Model Predictive Control 369

6.1 Introduction and Preliminary Results . . . . . . . . . . . . 369
6.1.1 Least Squares Solution . . . . . . . . . . . . . . . . . 370
6.1.2 Stability of Suboptimal MPC . . . . . . . . . . . . . . 375
6.2 Unconstrained Two-Player Game . . . . . . . . . . . . . . . 380
6.2.1 Centralized Control . . . . . . . . . . . . . . . . . . . 382
6.2.2 Decentralized Control . . . . . . . . . . . . . . . . . 383
6.2.3 Noncooperative Game . . . . . . . . . . . . . . . . . 384
6.2.4 Cooperative Game . . . . . . . . . . . . . . . . . . . . 392
6.2.5 Tracking Nonzero Setpoints . . . . . . . . . . . . . . 398
6.2.6 State Estimation . . . . . . . . . . . . . . . . . . . . . 405
6.3 Constrained Two-Player Game . . . . . . . . . . . . . . . . 406
6.3.1 Uncoupled Input Constraints . . . . . . . . . . . . . 408
6.3.2 Coupled Input Constraints . . . . . . . . . . . . . . 411
6.3.3 Exponential Convergence with Estimate Error . . . 413
6.3.4 Disturbance Models and Zero Offset . . . . . . . . 415
6.4 Constrained M-Player Game . . . . . . . . . . . . . . . . . . 419
6.5 Nonlinear Distributed MPC . . . . . . . . . . . . . . . . . . . 421
6.5.1 Nonconvexity . . . . . . . . . . . . . . . . . . . . . . . 421
6.5.2 Distributed Algorithm for Nonconvex Functions . 423
Contents xvii
6.5.3 Distributed Nonlinear Cooperative Control . . . . 425

6.5.4 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . 428
6.6 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 430
6.7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 435
7 Explicit Control Laws for Constrained Linear Systems 451

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 451
7.2 Parametric Programming . . . . . . . . . . . . . . . . . . . . 452
7.3 Parametric Quadratic Programming . . . . . . . . . . . . . 457
7.3.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . 457
7.3.2 Preview . . . . . . . . . . . . . . . . . . . . . . . . . . 458
7.3.3 Optimality Condition for a Convex Program . . . . 459
7.3.4 Solution of the Parametric Quadratic Program . . 462
7.3.5 Continuity of V 0 (·) and u0 (·) . . . . . . . . . . . . 466
7.4 Constrained Linear Quadratic Control . . . . . . . . . . . . 467
7.5 Parametric Piecewise Quadratic Programming . . . . . . . 469
7.6 DP Solution of the Constrained LQ Control Problem . . . 475
7.7 Parametric Linear Programming . . . . . . . . . . . . . . . 476
7.7.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . 476
7.7.2 Minimizer u0 (x) is Unique for all x ∈ X . . . . . . 478
7.8 Constrained Linear Control . . . . . . . . . . . . . . . . . . 481
7.9 Computation . . . . . . . . . . . . . . . . . . . . . . . . . . . 482
7.10 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 483
7.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484
8 Numerical Optimal Control 491

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 491
8.1.1 Discrete Time Optimal Control Problem . . . . . . 492
8.1.2 Convex Versus Nonconvex Optimization . . . . . . 493
8.1.3 Simultaneous Versus Sequential Optimal Control 496
8.1.4 Continuous Time Optimal Control Problem . . . . 498
8.2 Numerical Simulation . . . . . . . . . . . . . . . . . . . . . . 501
8.2.1 Explicit Runge-Kutta Methods . . . . . . . . . . . . . 502
8.2.2 Stiff Equations and Implicit Integrators . . . . . . . 506
8.2.3 Implicit Runge-Kutta and Collocation Methods . . 507
8.2.4 Differential Algebraic Equations . . . . . . . . . . . 511
8.2.5 Integrator Adaptivity . . . . . . . . . . . . . . . . . . 513
8.3 Solving Nonlinear Equation Systems . . . . . . . . . . . . . 513
8.3.1 Linear Systems . . . . . . . . . . . . . . . . . . . . . . 513
8.3.2 Nonlinear Root-Finding Problems . . . . . . . . . . 514
8.3.3 Local Convergence of Newton-Type Methods . . . 517
xviii Contents
8.3.4 Affine Invariance . . . . . . . . . . . . . . . . . . . . . 519

8.3.5 Globalization for Newton-Type Methods . . . . . . 519
8.4 Computing Derivatives . . . . . . . . . . . . . . . . . . . . . 520
8.4.1 Numerical Differentiation . . . . . . . . . . . . . . . 521
8.4.2 Algorithmic Differentiation . . . . . . . . . . . . . . 522
8.4.3 Implicit Function Interpretation . . . . . . . . . . . 523
8.4.4 Algorithmic Differentiation in Forward Mode . . . 526
8.4.5 Algorithmic Differentiation in Reverse Mode . . . 528
8.4.6 Differentiation of Simulation Routines . . . . . . . 531
8.4.7 Algorithmic and Symbolic Differentiation Software 533
8.4.8 CasADi for Optimization . . . . . . . . . . . . . . . . 533
8.5 Direct Optimal Control Parameterizations . . . . . . . . . 536
8.5.1 Direct Single Shooting . . . . . . . . . . . . . . . . . 538
8.5.2 Direct Multiple Shooting . . . . . . . . . . . . . . . . 540
8.5.3 Direct Transcription and Collocation Methods . . 544
8.6 Nonlinear Optimization . . . . . . . . . . . . . . . . . . . . . 548
8.6.1 Optimality Conditions and Perturbation Analysis 549
8.6.2 Nonlinear Optimization with Equalities . . . . . . . 552
8.6.3 Hessian Approximations . . . . . . . . . . . . . . . . 553
8.7 Newton-Type Optimization with Inequalities . . . . . . . 556
8.7.1 Sequential Quadratic Programming . . . . . . . . . 557
8.7.2 Nonlinear Interior Point Methods . . . . . . . . . . 558
8.7.3 Comparison of SQP and Nonlinear IP Methods . . 560
8.8 Structure in Discrete Time Optimal Control . . . . . . . . 561
8.8.1 Simultaneous Approach . . . . . . . . . . . . . . . . 562
8.8.2 Linear Quadratic Problems (LQP) . . . . . . . . . . . 564
8.8.3 LQP Solution by Riccati Recursion . . . . . . . . . . 564
8.8.4 LQP Solution by Condensing . . . . . . . . . . . . . 566
8.8.5 Sequential Approaches and Sparsity Exploitation 568
8.8.6 Differential Dynamic Programming . . . . . . . . . 570
8.8.7 Additional Constraints in Optimal Control . . . . . 572
8.9 Online Optimization Algorithms . . . . . . . . . . . . . . . 573
8.9.1 General Algorithmic Considerations . . . . . . . . . 574
8.9.2 Continuation Methods and Real-Time Iterations . 577
8.10 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 580
8.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 583
Author Index 601
Citation Index 608

Contents xix
Subject Index 614
A Mathematical Background 624

A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
A.2 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 624
A.3 Range and Nullspace of Matrices . . . . . . . . . . . . . . . 624
A.4 Linear Equations — Existence and Uniqueness . . . . . . 625
A.5 Pseudo-Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . 625
A.6 Partitioned Matrix Inversion Theorem . . . . . . . . . . . . 628
A.7 Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . . 629
A.8 Norms in Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
A.9 Sets in Rn . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 631
A.10Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 632
A.11Continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 633
A.12Derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 636
A.13Convex Sets and Functions . . . . . . . . . . . . . . . . . . . 641
A.13.1 Convex Sets . . . . . . . . . . . . . . . . . . . . . . . . 641
A.13.2 Convex Functions . . . . . . . . . . . . . . . . . . . . 646
A.14Differential Equations . . . . . . . . . . . . . . . . . . . . . . 648
A.15Random Variables and the Probability Density . . . . . . 654
A.16Multivariate Density Functions . . . . . . . . . . . . . . . . 659
A.16.1 Statistical Independence and Correlation . . . . . . 668
A.17Conditional Probability and Bayes’s Theorem . . . . . . . 672
A.18Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 678
B Stability Theory 693

B.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 693
B.2 Stability and Asymptotic Stability . . . . . . . . . . . . . . 696
B.3 Lyapunov Stability Theory . . . . . . . . . . . . . . . . . . . 700
B.3.1 Time-Invariant Systems . . . . . . . . . . . . . . . . 700
B.3.2 Time-Varying, Constrained Systems . . . . . . . . . 707
B.3.3 Upper bounding K functions . . . . . . . . . . . . . 709
B.4 Robust Stability . . . . . . . . . . . . . . . . . . . . . . . . . 709
B.4.1 Nominal Robustness . . . . . . . . . . . . . . . . . . 709
B.4.2 Robustness . . . . . . . . . . . . . . . . . . . . . . . . 711
B.5 Control Lyapunov Functions . . . . . . . . . . . . . . . . . . 713
B.6 Input-to-State Stability . . . . . . . . . . . . . . . . . . . . . 717
B.7 Output-to-State Stability and Detectability . . . . . . . . . 719
B.8 Input/Output-to-State Stability . . . . . . . . . . . . . . . . 720
B.9 Incremental-Input/Output-to-State Stability . . . . . . . . 722
B.10 Observability . . . . . . . . . . . . . . . . . . . . . . . . . . . 722
xx Contents
B.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724
C Optimization 729
C.1 Dynamic Programming . . . . . . . . . . . . . . . . . . . . . 729
C.1.1 Optimal Control Problem . . . . . . . . . . . . . . . 731
C.1.2 Dynamic Programming . . . . . . . . . . . . . . . . . 733
C.2 Optimality Conditions . . . . . . . . . . . . . . . . . . . . . 737
C.2.1 Tangent and Normal Cones . . . . . . . . . . . . . . 737
C.2.2 Convex Optimization Problems . . . . . . . . . . . . 741
C.2.3 Convex Problems: Polyhedral Constraint Set . . . 743
C.2.4 Nonconvex Problems . . . . . . . . . . . . . . . . . . 745
C.2.5 Tangent and Normal Cones . . . . . . . . . . . . . . 746
C.2.6 Constraint Set Defined by Inequalities . . . . . . . 750
C.2.7 Constraint Set; Equalities and Inequalities . . . . . 753
C.3 Set-Valued Functions and Continuity of Value Function . 755
C.3.1 Outer and Inner Semicontinuity . . . . . . . . . . . 757
C.3.2 Continuity of the Value Function . . . . . . . . . . . 759
C.4 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 767
List of Figures
1.1 System with input u, output y, and transfer function ma-

trix G connecting them; the model is y = Gu. . . . . . . . . 3
1.2 Typical input constraint sets U for (a) continuous actua-
tors and (b) mixed continuous/discrete actuators. . . . . . 9
1.3 Output of a stochastic system versus time. . . . . . . . . . . 10
1.4 Two quadratic functions and their sum. . . . . . . . . . . . . 15
1.5 Schematic of the moving horizon estimation problem. . . . 39
1.6 MPC controller consisting of: receding horizon regulator,
state estimator, and target selector. . . . . . . . . . . . . . . 52
1.7 Schematic of the well-stirred reactor. . . . . . . . . . . . . . . 54
1.8 Three measured outputs versus time after a step change
in inlet flowrate at 10 minutes; nd = 2. . . . . . . . . . . . . 57
1.9 Two manipulated inputs versus time after a step change
1.10 Three measured outputs versus time after a step change
1.11 Two manipulated inputs versus time after a step change
1.12 Plug-flow reactor. . . . . . . . . . . . . . . . . . . . . . . . . . . 60
1.13 Pendulum with applied torque. . . . . . . . . . . . . . . . . . 62
1.14 Feedback control system with output disturbance d, and
setpoint ysp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
2.1 Example of MPC. . . . . . . . . . . . . . . . . . . . . . . . . . . 101

2.2 Feasible region U2 , elliptical cost contours and ellipse
center a(x), and constrained minimizers for different
values of x. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
2.3 First element of control constraint set U3 (x) (shaded)
and control law κ3 (x) (line) versus x = (cos(θ), sin(θ)),
θ ∈ [−π , π ] on the unit circle for a nonlinear system with
terminal constraint. . . . . . . . . . . . . . . . . . . . . . . . . 106
2.4 Optimal cost V30 (x) versus x on the unit circle. . . . . . . . 107
xxi
xxii List of Figures
2.5 Closed-loop economic MPC versus tracking MPC starting

at x = (−8, 8) with optimal steady state (8, 4). Both con-
trollers asymptotically stabilize the steady state. Dashed
contours show cost functions for each controller. . . . . . . 159
2.6 Closed-loop evolution under economic MPC. The rotated
0
cost function V e is a Lyapunov function for the system. . . 160
2.7 Diagram of tank/cooler system. Each cooling unit can be
either on or off, and if on, it must be between its (possibly
nonzero) minimum and maximum capacities. . . . . . . . . 163
2.8 Feasible sets XN for two values of Q̇min . Note that for
Q̇min = 9 (right-hand side), XN for N ≤ 4 are discon-
nected sets. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
2.9 Phase portrait for closed-loop evolution of cooler system
with Q̇min = 9. Line colors show value of discrete actuator u2 . 165
2.10 Region of attraction (shaded region) for constrained MPC
controller of Exercise 2.6. . . . . . . . . . . . . . . . . . . . . . 174
2.11 The region Xf , in which the unconstrained LQR control
law is feasible for Exercise 2.7. . . . . . . . . . . . . . . . . . . 175
2.12 The region of attraction for terminal constraint x(N) ∈
Xf and terminal penalty Vf (x) = (1/2)x 0 Πx and the es-
timate of X̄N for Exercise 2.8. . . . . . . . . . . . . . . . . . . 177
2.13 Inconsistent setpoint (xsp , usp ), unreachable stage cost
`(x, u), and optimal steady states (xs , us ), and stage costs
`s (x, u) for constrained and unconstrained systems. . . . 181
2.14 Stage cost versus time for the case of unreachable setpoint. 182
2.15 Rotated cost-function contour è (x, u) = 0 (circles) for
λ = 0, −8, −12. Shaded region shows feasible region where
è (x, u) < 0. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185
3.1 Open-loop and feedback trajectories. . . . . . . . . . . . . . . 198

3.2 The sets XN , Rb , and Rc . . . . . . . . . . . . . . . . . . . . . . 214
3.3 Outer-bounding tube X(z, ū). . . . . . . . . . . . . . . . . . . . 228
3.4 Minimum feasible α for varying N. Note that we require
α ∈ [0, 1). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
3.5 Bounds on tightened constraint set Z̄ for varying N. Bounds
are |x1 | ≤ χ1 , |x2 | ≤ χ2 , and |u| ≤ µ. . . . . . . . . . . . . . . 233
3.6 Comparison of 100 realizations of standard and tube-
based MPC for the chemical reactor example. . . . . . . . . 244
3.7 Comparison of standard and tube-based MPC with an ag-
gressive model predictive controller. . . . . . . . . . . . . . . 245
List of Figures xxiii
3.8 Concentration versus time for the ancillary model predic-

tive controller with sample time ∆ = 12 (left) and ∆ = 8
(right). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
3.9 Observed probability εtest of constraint violation. Distri-
bution is based on 500 trials for each value of ε. Dashed
line shows the outcome predicted by formula (3.23), i.e.,
εtest = ε. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
3.10 Closed-loop robust MPC state evolution with uniformly
distributed |w| ≤ 0.1 from four different x0 . . . . . . . . . . 263
4.1 MHE arrival cost Ẑk (p), underbounding prior weighting

Γk (p), and MHE optimal value V̂k0 . . . . . . . . . . . . . . . . . 289
4.2 Concatenating two MHE sequences to create a single state
estimate sequence. . . . . . . . . . . . . . . . . . . . . . . . . . 292
4.3 Smoothing update. . . . . . . . . . . . . . . . . . . . . . . . . . 297
4.4 Comparison of filtering and smoothing updates for the
batch reactor system. Second column shows absolute es-
timate error. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
4.5 Evolution of the state (solid line) and EKF state estimate
(dashed line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 315
4.6 Evolution of the state (solid line) and UKF state estimate
(dashed line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 316
4.7 Evolution of the state (solid line) and MHE state estimate
(dashed line). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317
4.8 Perturbed trajectories terminating in Xf . . . . . . . . . . . . 321
4.9 Closed-loop performance of combined nonlinear MHE/MPC
with no disturbances. First column shows system states,
and second column shows estimation error. Dashed line
shows concentration setpoint. Vertical lines indicate times
of setpoint changes. . . . . . . . . . . . . . . . . . . . . . . . . 323
4.10 Closed-loop performance of combined nonlinear MHE/MPC
for varying disturbance size. The system is controlled be-
tween two steady states. . . . . . . . . . . . . . . . . . . . . . . 324
4.11 Relationships between i-IOSS, FSO, and observable for K-
continuous nonlinear systems. . . . . . . . . . . . . . . . . . . 330
5.1 State estimator tube. The solid line x̂(t) is the center of
the tube, and the dashed line is a sample trajectory of x(t). 342
5.2 The system with disturbance. The state estimate lies in
the inner tube, and the state lies in the outer tube. . . . . . 343
xxiv List of Figures
p p p+1 p+1
6.1 Convex step from (u1 , u2 ) to (u1 , u2 ). . . . . . . . . . 386
6.2 Ten iterations of noncooperative steady-state calculation. . 403
6.3 Ten iterations of cooperative steady-state calculation. . . . 403
6.4 Ten iterations of noncooperative steady-state calculation;
reversed pairing. . . . . . . . . . . . . . . . . . . . . . . . . . . 404
6.5 Ten iterations of cooperative steady-state calculation; re-
versed pairing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404
6.6 Cooperative control stuck on the boundary of U under
coupled constraints . . . . . . . . . . . . . . . . . . . . . . . . 412
6.7 Cost contours for a two-player, nonconvex game. . . . . . . 422
6.8 Nonconvex function optimized with the distributed gra-
dient algorithm. . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
6.9 Closed-loop state and control evolution with (x1 (0), x2 (0)) =
(3, −3). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 431
6.10 Contours of V (x(0), u1 , u2 ) for N = 1. . . . . . . . . . . . . . 432
6.11 Optimizing a quadratic function in one set of variables at
a time. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 440
6.12 Constrained optimality conditions and the normal cone. . 445
7.1 The sets Z, X, and U(x). . . . . . . . . . . . . . . . . . . . . . 454

7.2 Parametric linear program. . . . . . . . . . . . . . . . . . . . . 454
7.3 Unconstrained parametric quadratic program. . . . . . . . . 455
7.4 Parametric quadratic program. . . . . . . . . . . . . . . . . . . 455
7.5 Polar cone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 460
7.6 Regions Rx , x ∈ X for a second-order example. . . . . . . . 468
7.7 Solution to a parametric LP. . . . . . . . . . . . . . . . . . . . 479
7.8 Solution times for explicit and implicit MPC for N = 20. . . 486
8.1 Feasible set and reduced objective ψ(u(0)) of the non-

linear MPC Example 8.1. . . . . . . . . . . . . . . . . . . . . . . 496
8.2 Performance of different integration methods. . . . . . . . . 505
8.3 Polynomial approximation x e 1 (t) and true trajectory x1 (t)
of the first state and its derivative. . . . . . . . . . . . . . . . 510
8.4 Performance of implicit integration methods on a stiff ODE. 512
8.5 Newton-type iterations for solution of R(z) = 0 from Ex-
ample 8.5. Left: exact Newton method. Right: constant
Jacobian approximation. . . . . . . . . . . . . . . . . . . . . . 516
8.6 Convergence of different sequences as a function of k. . . 518
8.7 A hanging chain at rest. See Exercise 8.6(b). . . . . . . . . . 587
List of Figures xxv
8.8 Direct single shooting solution for (8.63) without path con-
straints. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 589
8.9 Open-loop simulation for (8.63) using collocation. . . . . . . 592
8.10 Gauss-Newton iterations for the direct multiple-shooting
method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594
A.1 The four fundamental subspaces of matrix A . . . . . . . . 626

A.2 Matrix A maps into R(A). . . . . . . . . . . . . . . . . . . . . . 627
A.3 Pseudo-inverse of A maps into R(A0 ). . . . . . . . . . . . . . 627
A.4 Subgradient. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 640
A.5 Separating hyperplane. . . . . . . . . . . . . . . . . . . . . . . 642
A.6 Polar cone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645
A.7 A convex function. . . . . . . . . . . . . . . . . . . . . . . . . . 646
A.8 Normal distribution. . . . . . . . . . . . . . . . . . . . . . . . . 658
A.9 Multivariate normal in two dimensions. . . . . . . . . . . . . 660
A.10 The geometry of quadratic form x 0 Ax = b. . . . . . . . . . . 661
A.11 A nearly singular normal density in two dimensions. . . . . 665
A.12 The region X(c) for y = max(x1 , x2 ) ≤ c. . . . . . . . . . . . 667
A.13 A joint density function for the two uncorrelated random
variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 670
A.14 The probability distribution and inverse distribution for
random variable ξ. . . . . . . . . . . . . . . . . . . . . . . . . . 687
B.1 Stability of the origin. . . . . . . . . . . . . . . . . . . . . . . . 697

B.2 An attractive but unstable origin. . . . . . . . . . . . . . . . . 698
C.1 Routing problem. . . . . . . . . . . . . . . . . . . . . . . . . . . 730

C.2 Approximation of the set U . . . . . . . . . . . . . . . . . . . . 738
C.3 Tangent cones. . . . . . . . . . . . . . . . . . . . . . . . . . . . 738
C.4 Normal at u. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 739
C.5 Condition of optimality. . . . . . . . . . . . . . . . . . . . . . . 745
C.6 Tangent and normal cones. . . . . . . . . . . . . . . . . . . . . 747
C.7 Condition of optimality. . . . . . . . . . . . . . . . . . . . . . . 749
C.8 FU (u) 6= TU (u). . . . . . . . . . . . . . . . . . . . . . . . . . . . 751
C.9 Graph of set-valued function U(·). . . . . . . . . . . . . . . . 756
C.10 Graphs of discontinuous set-valued functions. . . . . . . . . 757
C.11 Outer and inner semicontinuity of U(·). . . . . . . . . . . . . 758
C.12 Subgradient of f (·). . . . . . . . . . . . . . . . . . . . . . . . . 762
List of Examples and Statements
1.1 Example: Sum of quadratic functions . . . . . . . . . . . . . 15

1.2 Lemma: Hautus lemma for controllability . . . . . . . . . . . 24
1.3 Lemma: LQR convergence . . . . . . . . . . . . . . . . . . . . . 24
1.4 Lemma: Hautus lemma for observability . . . . . . . . . . . 42
1.5 Lemma: Convergence of estimator cost . . . . . . . . . . . . 43
1.6 Lemma: Estimator convergence . . . . . . . . . . . . . . . . . 44
1.7 Assumption: Target feasibility and uniqueness . . . . . . . 48
1.8 Lemma: Detectability of the augmented system . . . . . . . 50
1.9 Corollary: Dimension of the disturbance . . . . . . . . . . . 50
1.10 Lemma: Offset-free control . . . . . . . . . . . . . . . . . . . . 52
1.11 Example: More measured outputs than inputs and zero
offset . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
1.12 Lemma: Hautus lemma for stabilizability . . . . . . . . . . . 68
1.13 Lemma: Hautus lemma for detectability . . . . . . . . . . . . 72
1.14 Lemma: Stabilizable systems and feasible targets . . . . . . 83
2.1 Proposition: Continuity of system solution . . . . . . . . . . 94

2.2 Assumption: Continuity of system and cost . . . . . . . . . 97
2.3 Assumption: Properties of constraint sets . . . . . . . . . . 98
2.4 Proposition: Existence of solution to optimal control problem 98
2.5 Example: Linear quadratic MPC . . . . . . . . . . . . . . . . . 99
2.6 Example: Closer inspection of linear quadratic MPC . . . . 101
2.7 Theorem: Continuity of value function and control law . . 104
2.8 Example: Discontinuous MPC control law . . . . . . . . . . . 105
2.9 Definition: Positive and control invariant sets . . . . . . . . 109
2.10 Proposition: Existence of solutions to DP recursion . . . . . 110
2.11 Definition: Asymptotically stable and GAS . . . . . . . . . . 112
2.12 Definition: Lyapunov function . . . . . . . . . . . . . . . . . . 113
2.13 Theorem: Lyapunov stability theorem . . . . . . . . . . . . . 113
2.14 Assumption: Basic stability assumption . . . . . . . . . . . . 114
2.15 Proposition: The value function VN0 (·) is locally bounded . 115
2.16 Proposition: Extension of upper bound to XN . . . . . . . . 115
2.17 Assumption: Weak controllability . . . . . . . . . . . . . . . . 116
2.18 Proposition: Monotonicity of the value function . . . . . . . 118
2.19 Theorem: Asymptotic stability of the origin . . . . . . . . . 119
xxvii
xxviii List of Examples and Statements
2.20 Definition: Exponential stability . . . . . . . . . . . . . . . . . 120

2.21 Theorem: Lyapunov function and exponential stability . . 120
2.22 Definition: Input/output-to-state stable (IOSS) . . . . . . . . 121
2.23 Assumption: Modified basic stability assumption . . . . . . 121
2.24 Theorem: Asymptotic stability with stage cost `(y, u) . . . 122
2.25 Assumption: Continuity of system and cost; time-varying
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
2.26 Assumption: Properties of constraint sets; time-varying case 124
2.27 Definition: Sequential positive invariance and sequential
control invariance . . . . . . . . . . . . . . . . . . . . . . . . . . 125
2.28 Proposition: Continuous system solution; time-varying case 125
2.29 Proposition: Existence of solution to optimal control prob-
lem; time-varying case . . . . . . . . . . . . . . . . . . . . . . . 125
2.30 Definition: Asymptotically stable and GAS for time-varying
systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
2.31 Definition: Lyapunov function: time-varying, constrained
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
2.32 Theorem: Lyapunov theorem for asymptotic stability (time-
varying, constrained) . . . . . . . . . . . . . . . . . . . . . . . . 126
2.33 Assumption: Basic stability assumption; time-varying case 127
2.34 Proposition: Optimal cost decrease; time-varying case . . . 127
2.35 Proposition: MPC cost is less than terminal cost . . . . . . . 127
2.36 Proposition: Optimal value function properties; time-varying
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
2.37 Assumption: Uniform weak controllability . . . . . . . . . . 128
2.38 Proposition: Conditions for uniform weak controllability . 128
2.39 Theorem: Asymptotic stability of the origin: time-varying
MPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
2.40 Lemma: Entering the terminal region . . . . . . . . . . . . . . 145
2.41 Theorem: MPC stability; no terminal constraint . . . . . . . 146
2.42 Proposition: Admissible warm start in Xf . . . . . . . . . . . 149
2.43 Algorithm: Suboptimal MPC . . . . . . . . . . . . . . . . . . . 149
2.44 Proposition: Linking warm start and state . . . . . . . . . . . 150
2.45 Definition: Asymptotic stability (difference inclusion) . . . 150
2.46 Definition: Lyapunov function (difference inclusion) . . . . 151
2.47 Proposition: Asymptotic stability (difference inclusion) . . 151
2.48 Theorem: Asymptotic stability of suboptimal MPC . . . . . 151
2.49 Assumption: Continuity of system and cost . . . . . . . . . 154
2.50 Assumption: Properties of constraint sets . . . . . . . . . . 154
2.51 Assumption: Cost lower bound . . . . . . . . . . . . . . . . . 154
List of Examples and Statements xxix
2.52 Proposition: Asymptotic average performance . . . . . . . . 155

2.53 Definition: Dissipativity . . . . . . . . . . . . . . . . . . . . . . 156
2.54 Assumption: Continuity at the steady state . . . . . . . . . . 157
2.55 Assumption: Strict dissipativity . . . . . . . . . . . . . . . . . 157
2.56 Theorem: Asymptotic stability of economic MPC . . . . . . 157
2.57 Example: Economic MPC versus tracking MPC . . . . . . . . 158
2.58 Example: MPC with mixed continuous/discrete actuators . 162
2.59 Theorem: Lyapunov theorem for asymptotic stability . . . 177
2.60 Proposition: Convergence of state under IOSS . . . . . . . . 178
2.61 Lemma: An equality for quadratic functions . . . . . . . . . 178
2.62 Lemma: Evolution in a compact set . . . . . . . . . . . . . . . 179
3.1 Definition: Robust global asymptotic stability . . . . . . . . 207

3.2 Theorem: Lyapunov function and RGAS . . . . . . . . . . . . 208
3.3 Theorem: Robust global asymptotic stability and regular-
ization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
3.4 Proposition: Bound for continuous functions . . . . . . . . . 211
3.5 Proposition: Robustness of nominal MPC . . . . . . . . . . . 214
3.6 Definition: Robust control invariance . . . . . . . . . . . . . 217
3.7 Definition: Robust positive invariance . . . . . . . . . . . . . 217
3.8 Assumption: Basic stability assumption; robust case . . . . 218
3.9 Theorem: Recursive feasibility of control policies . . . . . . 218
3.10 Definition: Set algebra and Hausdorff distance . . . . . . . . 224
3.11 Definition: Robust asymptotic stability of a set . . . . . . . 230
3.12 Proposition: Robust asymptotic stability of tube-based
MPC for linear systems . . . . . . . . . . . . . . . . . . . . . . 230
3.13 Example: Calculation of tightened constraints . . . . . . . . 231
3.14 Proposition: Recursive feasibility of tube-based MPC . . . . 235
3.15 Proposition: Robust exponential stability of improved tube-
based MPC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
3.16 Proposition: Implicit satisfaction of terminal constraint . . 239
3.17 Proposition: Properties of the value function . . . . . . . . . 240
3.18 Proposition: Neighborhoods of the uncertain system . . . . 241
3.19 Proposition: Robust positive invariance of tube-based MPC
for nonlinear systems . . . . . . . . . . . . . . . . . . . . . . . 241
3.20 Example: Robust control of an exothermic reaction . . . . . 243
3.21 Assumption: Feasibility of robust control . . . . . . . . . . . 251
3.22 Assumption: Robust terminal set condition . . . . . . . . . 252
3.23 Example: Constraint tightening via sampling . . . . . . . . . 254
xxx List of Examples and Statements
4.1 Definition: i-IOSS . . . . . . . . . . . . . . . . . . . . . . . . . . 272

4.2 Proposition: Convergence of state under i-IOSS . . . . . . . 272
4.3 Definition: β-convergent sequence . . . . . . . . . . . . . . . 272
4.4 Proposition: Bounded, convergent sequences are β-convergent 272
4.5 Proposition: Bound for sum cost of convergent sequence. . 273
4.6 Assumption: β-convergent disturbances . . . . . . . . . . . . 273
4.7 Assumption: Positive definite stage cost . . . . . . . . . . . . 273
4.8 Definition: Robustly globally asymptotically stable esti-
mation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
4.9 Proposition: Boundedness and convergence of estimate error 274
4.10 Theorem: FIE with β-convergent disturbances is RGAS . . . 277
4.11 Lemma: Duality of controllability and observability . . . . . 282
4.12 Theorem: Riccati iteration and regulator stability . . . . . . 282
4.13 Definition: Observability . . . . . . . . . . . . . . . . . . . . . 284
4.14 Definition: Final-state observability . . . . . . . . . . . . . . . 285
4.15 Definition: Globally K-continuous . . . . . . . . . . . . . . . 285
4.16 Proposition: Observable and global K-continuous imply FSO 285
4.17 Definition: RGAS estimation (observable case) . . . . . . . . 285
4.18 Theorem: MHE is RGAS (observable case) with zero prior
weighting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 286
4.19 Definition: Full information arrival cost . . . . . . . . . . . . 287
4.20 Lemma: MHE and FIE equivalence . . . . . . . . . . . . . . . . 287
4.21 Definition: MHE arrival cost . . . . . . . . . . . . . . . . . . . 288
4.22 Assumption: Prior weighting . . . . . . . . . . . . . . . . . . . 288
4.23 Proposition: Arrival cost of full information greater than
MHE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
4.24 Definition: MHE-detectable system . . . . . . . . . . . . . . . 290
4.25 Theorem: MHE is RGAS . . . . . . . . . . . . . . . . . . . . . . 290
4.26 Assumption: Estimator constraint sets . . . . . . . . . . . . 294
4.27 Theorem: Constrained full information is RGAS . . . . . . . 295
4.28 Theorem: Constrained MHE is RGAS . . . . . . . . . . . . . . 295
4.29 Assumption: Prior weighting for linear system . . . . . . . . 296
4.30 Assumption: Polyhedral constraint sets . . . . . . . . . . . . 296
4.31 Corollary: Constrained MHE is RGAS . . . . . . . . . . . . . . 296
4.32 Example: Filtering and smoothing updates . . . . . . . . . . 299
4.33 Definition: i-IOSS (max form) . . . . . . . . . . . . . . . . . . . 301
4.34 Assumption: Positive definite cost function . . . . . . . . . 302
4.35 Assumption: Lipschitz continuity of stage-cost bound com-
positions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 302
4.36 Assumption: Uniform i-IOSS contractivity . . . . . . . . . . . 302
List of Examples and Statements xxxi
4.37 Proposition: Locally Lipschitz upper bound . . . . . . . . . . 302

4.38 Proposition: The nameless lemma . . . . . . . . . . . . . . . 303
4.39 Theorem: MHE is RAS . . . . . . . . . . . . . . . . . . . . . . . 304
4.40 Example: MHE of linear time-invariant system . . . . . . . . 306
4.41 Example: EKF, UKF, and MHE performance comparison . . 313
4.42 Definition: i-UIOSS . . . . . . . . . . . . . . . . . . . . . . . . . 319
4.43 Assumption: Bounded estimate error . . . . . . . . . . . . . 319
4.44 Definition: Robust positive invariance . . . . . . . . . . . . . 320
4.45 Definition: Robust asymptotic stability . . . . . . . . . . . . 320
4.46 Definition: ISS Lyapunov function . . . . . . . . . . . . . . . . 320
4.47 Proposition: ISS Lyapunov stability theorem . . . . . . . . . 320
4.48 Theorem: Combined MHE/MPC is RAS . . . . . . . . . . . . . 322
4.49 Example: Combined MHE/MPC . . . . . . . . . . . . . . . . . . 323
5.1 Definition: Positive invariance; robust positive invariance . 345

5.2 Proposition: Proximity of state and state estimate . . . . . 345
5.3 Proposition: Proximity of state estimate and nominal state 347
5.4 Assumption: Constraint bounds . . . . . . . . . . . . . . . . . 348
5.5 Algorithm: Robust control algorithm (linear constrained
systems) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
5.6 Proposition: Exponential stability of output MPC . . . . . . 351
5.7 Algorithm: Robust control algorithm (offset-free MPC) . . . 359
6.1 Algorithm: Suboptimal MPC (simplified) . . . . . . . . . . . . 375

6.2 Definition: Lyapunov stability . . . . . . . . . . . . . . . . . . 376
6.3 Definition: Uniform Lyapunov stability . . . . . . . . . . . . 377
6.4 Definition: Exponential stability . . . . . . . . . . . . . . . . . 377
6.5 Lemma: Exponential stability of suboptimal MPC . . . . . . 378
6.6 Lemma: Global asymptotic stability and exponential con-
vergence with mixed powers of norm . . . . . . . . . . . . . 379
6.7 Lemma: Converse theorem for exponential stability . . . . 380
6.8 Assumption: Unconstrained two-player game . . . . . . . . 386
6.9 Example: Nash equilibrium is unstable . . . . . . . . . . . . . 389
6.10 Example: Nash equilibrium is stable but closed loop is
unstable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 390
6.11 Example: Nash equilibrium is stable and the closed loop
is stable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
6.12 Example: Stability and offset in the distributed target cal-
culation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 401
6.13 Assumption: Constrained two-player game . . . . . . . . . . 407
xxxii List of Examples and Statements
6.14 Lemma: Global asymptotic stability and exponential con-

vergence of perturbed system . . . . . . . . . . . . . . . . . . 414
6.15 Assumption: Disturbance models . . . . . . . . . . . . . . . . 415
6.16 Lemma: Detectability of distributed disturbance model . . 415
6.17 Assumption: Constrained M-player game . . . . . . . . . . . 420
6.18 Lemma: Distributed gradient algorithm properties . . . . . 424
6.19 Assumption: Basic stability assumption (distributed) . . . . 426
6.20 Proposition: Terminal constraint satisfaction . . . . . . . . 427
6.21 Theorem: Asymptotic stability . . . . . . . . . . . . . . . . . . 429
6.22 Example: Nonlinear distributed control . . . . . . . . . . . . 429
6.23 Lemma: Local detectability . . . . . . . . . . . . . . . . . . . . 443
7.1 Definition: Polytopic (polyhedral) partition . . . . . . . . . . 456

7.2 Definition: Piecewise affine function . . . . . . . . . . . . . . 456
7.3 Assumption: Strict convexity . . . . . . . . . . . . . . . . . . . 457
7.4 Definition: Polar cone . . . . . . . . . . . . . . . . . . . . . . . 459
7.5 Proposition: Farkas’s lemma . . . . . . . . . . . . . . . . . . . 459
7.6 Proposition: Optimality conditions for convex set . . . . . . 459
7.7 Proposition: Optimality conditions in terms of polar cone 461
7.8 Proposition: Optimality conditions for linear inequalities . 461
7.9 Proposition: Solution of P(w), w ∈ Rx0 . . . . . . . . . . . . 463
7.10 Proposition: Piecewise quadratic (affine) cost (solution) . . 464
7.11 Example: Parametric QP . . . . . . . . . . . . . . . . . . . . . . 464
7.12 Example: Explicit optimal control . . . . . . . . . . . . . . . . 465
7.13 Proposition: Continuity of cost and solution . . . . . . . . . 467
7.14 Assumption: Continuous, piecewise quadratic function . . 470
7.15 Definition: Active polytope (polyhedron) . . . . . . . . . . . 471
7.16 Proposition: Solving P using Pi . . . . . . . . . . . . . . . . . 471
7.17 Proposition: Optimality of u0x (w) in Rx . . . . . . . . . . . . 474
7.18 Proposition: Piecewise quadratic (affine) solution . . . . . . 474
7.19 Proposition: Optimality conditions for parametric LP . . . 478
7.20 Proposition: Solution of P . . . . . . . . . . . . . . . . . . . . . 481
7.21 Proposition: Piecewise affine cost and solution . . . . . . . 481
8.1 Example: Nonlinear MPC . . . . . . . . . . . . . . . . . . . . . 495

8.2 Example: Sequential approach . . . . . . . . . . . . . . . . . . 498
8.3 Example: Integration methods of different order . . . . . . 504
8.4 Example: Implicit integrators for a stiff ODE system . . . . 511
8.5 Example: Finding a fifth root with Newton-type iterations . 516
8.6 Example: Convergence rates . . . . . . . . . . . . . . . . . . . 517
List of Examples and Statements xxxiii
8.7 Theorem: Local contraction for Newton-type methods . . . 518

8.8 Corollary: Convergence of exact Newton’s method . . . . . 519
8.9 Example: Function evaluation via elementary operations . 522
8.10 Example: Implicit function representation . . . . . . . . . . 524
8.11 Example: Forward algorithmic differentiation . . . . . . . . 526
8.12 Example: Algorithmic differentiation in reverse mode . . . 528
8.13 Example: Sequential optimal control using CasADi from
Octave . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534
8.14 Theorem: KKT conditions . . . . . . . . . . . . . . . . . . . . . 549
8.15 Theorem: Strong second-order sufficient conditions for
optimality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 551
8.16 Theorem: Tangential predictor by quadratic program . . . 551
A.1 Theorem: Schur decomposition . . . . . . . . . . . . . . . . . 629

A.2 Theorem: Real Schur decomposition . . . . . . . . . . . . . . 630
A.3 Theorem: Bolzano-Weierstrass . . . . . . . . . . . . . . . . . . 632
A.4 Proposition: Convergence of monotone sequences . . . . . 633
A.5 Proposition: Uniform continuity . . . . . . . . . . . . . . . . . 634
A.6 Proposition: Compactness of continuous functions of com-
pact sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 635
A.7 Proposition: Weierstrass . . . . . . . . . . . . . . . . . . . . . 636
A.8 Proposition: Derivative and partial derivative . . . . . . . . 637
A.9 Proposition: Continuous partial derivatives . . . . . . . . . . 638
A.10 Proposition: Chain rule . . . . . . . . . . . . . . . . . . . . . . 638
A.11 Proposition: Mean value theorem for vector functions . . . 638
A.12 Definition: Convex set . . . . . . . . . . . . . . . . . . . . . . . 641
A.13 Theorem: Caratheodory . . . . . . . . . . . . . . . . . . . . . . 641
A.14 Theorem: Separation of convex sets . . . . . . . . . . . . . . 642
A.15 Theorem: Separation of convex set from zero . . . . . . . . 643
A.16 Corollary: Existence of separating hyperplane . . . . . . . . 643
A.17 Definition: Support hyperplane . . . . . . . . . . . . . . . . . 644
A.18 Theorem: Convex set and halfspaces . . . . . . . . . . . . . . 644
A.19 Definition: Convex cone . . . . . . . . . . . . . . . . . . . . . . 644
A.20 Definition: Polar cone . . . . . . . . . . . . . . . . . . . . . . . 644
A.21 Definition: Cone generator . . . . . . . . . . . . . . . . . . . . 645
A.22 Proposition: Cone and polar cone generator . . . . . . . . . 645
A.23 Theorem: Convexity implies continuity . . . . . . . . . . . . 647
A.24 Theorem: Differentiability and convexity . . . . . . . . . . . 647
A.25 Theorem: Second derivative and convexity . . . . . . . . . . 647
A.26 Definition: Level set . . . . . . . . . . . . . . . . . . . . . . . . 648
xxxiv List of Examples and Statements
A.27 Definition: Sublevel set . . . . . . . . . . . . . . . . . . . . . . 648

A.28 Definition: Support function . . . . . . . . . . . . . . . . . . . 648
A.29 Proposition: Set membership and support function . . . . . 648
A.30 Proposition: Lipschitz continuity of support function . . . 648
A.31 Theorem: Existence of solution to differential equations . 651
A.32 Theorem: Maximal interval of existence . . . . . . . . . . . . 651
A.33 Theorem: Continuity of solution to differential equation . 651
A.34 Theorem: Bellman-Gronwall . . . . . . . . . . . . . . . . . . . 651
A.35 Theorem: Existence of solutions to forced systems . . . . . 653
A.36 Example: Fourier transform of the normal density. . . . . . 659
A.37 Definition: Density of a singular normal . . . . . . . . . . . . 662
A.38 Example: Marginal normal density . . . . . . . . . . . . . . . 663
A.39 Example: Nonlinear transformation . . . . . . . . . . . . . . . 666
A.40 Example: Maximum of two random variables . . . . . . . . . 667
A.41 Example: Independent implies uncorrelated . . . . . . . . . 668
A.42 Example: Does uncorrelated imply independent? . . . . . . 669
A.43 Example: Independent and uncorrelated are equivalent
for normals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671
A.44 Example: Conditional normal density . . . . . . . . . . . . . 674
A.45 Example: More normal conditional densities . . . . . . . . . 675
B.1 Definition: Equilibrium point . . . . . . . . . . . . . . . . . . . 694

B.2 Definition: Positive invariant set . . . . . . . . . . . . . . . . . 694
B.3 Definition: K, K∞ , KL, and PD functions . . . . . . . . . . 695
B.4 Definition: Local stability . . . . . . . . . . . . . . . . . . . . . 696
B.5 Definition: Global attraction . . . . . . . . . . . . . . . . . . . 696
B.6 Definition: Global asymptotic stability . . . . . . . . . . . . . 696
B.7 Definition: Various forms of stability . . . . . . . . . . . . . . 697
B.8 Definition: Global asymptotic stability (KL version) . . . . . 698
B.9 Proposition: Connection of classical and KL global asymp-
totic stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 698
B.10 Definition: Various forms of stability (constrained) . . . . . 699
B.11 Definition: Asymptotic stability (constrained, KL version) . 700
B.12 Definition: Lyapunov function (unconstrained and con-
strained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 700
B.13 Theorem: Lyapunov function and GAS (classical definition) 701
B.14 Lemma: From PD to K∞ function (Jiang and Wang (2002)) 702
B.15 Theorem: Lyapunov function and global asymptotic sta-
bility (KL definition) . . . . . . . . . . . . . . . . . . . . . . . . 703
B.16 Proposition: Improving convergence (Sontag (1998b)) . . . 705
List of Examples and Statements xxxv
B.17 Theorem: Converse theorem for global asymptotic stability 705

B.18 Theorem: Lyapunov function for asymptotic stability (con-
strained) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 706
B.19 Theorem: Lyapunov function for exponential stability . . . 706
B.20 Lemma: Lyapunov function for linear systems . . . . . . . . 706
B.21 Definition: Sequential positive invariance . . . . . . . . . . . 707
B.22 Definition: Asymptotic stability (time-varying, constrained) 707
B.23 Definition: Lyapunov function: time-varying, constrained
case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 708
B.24 Theorem: Lyapunov theorem for asymptotic stability (time-
varying, constrained) . . . . . . . . . . . . . . . . . . . . . . . . 708
B.25 Proposition: Global K function overbound . . . . . . . . . . 709
B.26 Definition: Nominal robust global asymptotic stability . . . 710
B.27 Theorem: Nominal robust global asymptotic stability and
Lyapunov function . . . . . . . . . . . . . . . . . . . . . . . . . 710
B.28 Definition: Positive invariance with disturbances . . . . . . 711
B.29 Definition: Local stability (disturbances) . . . . . . . . . . . . 712
B.30 Definition: Global attraction (disturbances) . . . . . . . . . . 712
B.31 Definition: GAS (disturbances) . . . . . . . . . . . . . . . . . . 712
B.32 Definition: Lyapunov function (disturbances) . . . . . . . . . 712
B.33 Theorem: Lyapunov function for global asymptotic sta-
bility (disturbances) . . . . . . . . . . . . . . . . . . . . . . . . 713
B.34 Definition: Global control Lyapunov function (CLF) . . . . . 714
B.35 Definition: Global stabilizability . . . . . . . . . . . . . . . . . 714
B.36 Definition: Positive invariance (disturbance and control) . . 715
B.37 Definition: CLF (disturbance and control) . . . . . . . . . . . 715
B.38 Definition: Positive invariance (constrained) . . . . . . . . . 715
B.39 Definition: CLF (constrained) . . . . . . . . . . . . . . . . . . . 715
B.40 Definition: Control invariance (disturbances, constrained) 716
B.41 Definition: CLF (disturbances, constrained) . . . . . . . . . . 716
B.42 Definition: Input-to-state stable (ISS) . . . . . . . . . . . . . . 717
B.43 Definition: ISS-Lyapunov function . . . . . . . . . . . . . . . . 718
B.44 Lemma: ISS-Lyapunov function implies ISS . . . . . . . . . . 718
B.45 Definition: ISS (constrained) . . . . . . . . . . . . . . . . . . . 718
B.46 Definition: ISS-Lyapunov function (constrained) . . . . . . . 718
B.47 Lemma: ISS-Lyapunov function implies ISS (constrained) . 719
B.48 Definition: Output-to-state stable (OSS) . . . . . . . . . . . . 719
B.49 Definition: OSS-Lyapunov function . . . . . . . . . . . . . . . 719
B.50 Theorem: OSS and OSS-Lyapunov function . . . . . . . . . . 720
B.51 Definition: Input/output-to-state stable (IOSS) . . . . . . . . 720
xxxvi List of Examples and Statements
B.52 Definition: IOSS-Lyapunov function . . . . . . . . . . . . . . . 720

B.53 Theorem: Modified IOSS-Lyapunov function . . . . . . . . . 721
B.54 Conjecture: IOSS and IOSS-Lyapunov function . . . . . . . . 721
B.55 Definition: Incremental input/output-to-state stable . . . . 722
B.56 Definition: Observability . . . . . . . . . . . . . . . . . . . . . 722
B.57 Assumption: Lipschitz continuity of model . . . . . . . . . . 723
B.58 Lemma: Lipschitz continuity and state difference bound . 723
B.59 Theorem: Observability and convergence of state . . . . . . 723
C.1 Lemma: Principle of optimality . . . . . . . . . . . . . . . . . 734

C.2 Theorem: Optimal value function and control law from DP 734
C.3 Example: DP applied to linear quadratic regulator . . . . . . 736
C.4 Definition: Tangent vector . . . . . . . . . . . . . . . . . . . . 739
C.5 Proposition: Tangent vectors are closed cone . . . . . . . . 739
C.6 Definition: Regular normal . . . . . . . . . . . . . . . . . . . . 739
C.7 Proposition: Relation of normal and tangent cones . . . . . 740
C.8 Proposition: Global optimality for convex problems . . . . 741
C.9 Proposition: Optimality conditions—normal cone . . . . . . 742
C.10 Proposition: Optimality conditions—tangent cone . . . . . 743
C.11 Proposition: Representation of tangent and normal cones . 743
C.12 Proposition: Optimality conditions—linear inequalities . . 744
C.13 Corollary: Optimality conditions—linear inequalities . . . . 744
C.14 Proposition: Necessary condition for nonconvex problem . 746
C.15 Definition: General normal . . . . . . . . . . . . . . . . . . . . 748
C.16 Definition: General tangent . . . . . . . . . . . . . . . . . . . . 748
C.17 Proposition: Set of regular tangents is closed convex cone 748
C.18 Definition: Regular set . . . . . . . . . . . . . . . . . . . . . . . 749
C.19 Proposition: Conditions for regular set . . . . . . . . . . . . 749
C.20 Proposition: Quasiregular set . . . . . . . . . . . . . . . . . . 751
C.21 Proposition: Optimality conditions nonconvex problem . . 752
C.22 Proposition: Fritz-John necessary conditions . . . . . . . . . 753
C.23 Definition: Outer semicontinuous function . . . . . . . . . . 757
C.24 Definition: Inner semicontinuous function . . . . . . . . . . 758
C.25 Definition: Continuous function . . . . . . . . . . . . . . . . . 758
C.26 Theorem: Equivalent conditions for outer and inner semi-
continuity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 759
C.27 Proposition: Outer semicontinuity and closed graph . . . . 759
C.28 Theorem: Minimum theorem . . . . . . . . . . . . . . . . . . . 760
C.29 Theorem: Lipschitz continuity of the value function, con-
stant U . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 761
List of Examples and Statements xxxvii
C.30 Definition: Subgradient of convex function . . . . . . . . . . 762

C.31 Theorem: Clarke et al. (1998) . . . . . . . . . . . . . . . . . . 762
C.32 Corollary: A bound on d(u, U(x 0 )) for u ∈ U (x) . . . . . . 763
C.33 Theorem: Continuity of U(·) . . . . . . . . . . . . . . . . . . . 765
C.34 Theorem: Continuity of the value function . . . . . . . . . . 765
C.35 Theorem: Lipschitz continuity of the value function—U (x) 766
Notation
Mathematical notation
∃ there exists
∈ is an element of
∀ for all
=⇒ ⇐= implies; is implied by
=6 ⇒ ⇐=
6 does not imply; is not implied by
a := b a is defined to be equal to b.
a =: b b is defined to be equal to a.
≈ approximately equal
V (·) function V
V :A→B V is a function mapping set A into set B
x , V (x) function V maps variable x to value V (x)
x+ value of x at next sample time (discrete time system)
ẋ time derivative of x (continuous time system)
fx partial derivative of f (x) with respect to x
∇ nabla or del operator
δ unit impulse or delta function
|x| absolute value of scalar; norm of vector (two-norm unless
stated otherwise); induced norm of matrix
x sequence of vector-valued variable x, (x(0), x(1), . . .)
kxk sup norm over a sequence, supi≥0 |x(i)|
kxka:b max a≤i≤b |x(i)|
tr(A) trace of matrix A
det(A) determinant of matrix A
eig(A) set of eigenvalues of matrix A
ρ(A) spectral radius of matrix A, max i |λi | for λi ∈ eig(A)
A−1 inverse of matrix A
A †
pseudo-inverse of matrix A
A0 transpose of matrix A
inf infimum or greatest lower bound
min minimum
sup supremum or least upper bound
max maximum
xxxix
xl Notation
arg argument or solution of an optimization

s.t. subject to
I integers
I≥0 nonnegative integers
In:m integers in the interval [n, m]
R real numbers
R≥0 nonnegative real numbers
Rn real-valued n-vectors
m×n
R real-valued m × n matrices
C complex numbers
B ball in Rn of unit radius
x ∼ px random variable x has probability density px
E(x) expectation of random variable x
var(x) variance of random variable x
cov(x, y) covariance of random variables x and y
N(m, P ) normal distribution (mean m, covariance P ), x ∼ N(m, P )
n(x, m, P ) normal probability density, px (x) = n(x, m, P )
∅ the empty set
aff(A) affine hull of set A
int(A) interior of set A
co(A) convex hull of the set A
A closure of set A
epi(f ) epigraph of function f
leva V sublevel set of function V , {x | V (x) ≤ a}
f ◦g composition of functions f and g, f ◦ g (s) := f (g(s))
a⊕b maximum of scalars a and b, Chapter 4
A⊕B set addition of sets A and B, Chapters 3 and 5
A B set subtraction of set B from set A
A\B elements of set A not in set B
A∪B union of sets A and B
A∩B intersection of sets A and B
A⊆B set A is a subset of set B
A⊇B set A is a superset of set B
A⊂B set A is a proper (or strict) subset of set B
A⊃B set A is a proper (or strict) superset of set B
d(a, B) Distance between element a and set B
dH (A, B) Hausdorff distance between sets A and B
x & y (x % y) x converges to y from above (below)
sat(x) saturation, sat(x) = x if |x| ≤ 1, −1 if x < −1, 1 if x > 1
Notation xli
Symbols
A, B, C system matrices, discrete time, x + = Ax + Bu, y = Cx
Ac , Bc system matrices, continuous time, ẋ = Ac x + Bc u
Aij state transition matrix for player i to player j
Ai state transition matrix for player i
ALi estimate error transition matrix Ai − Li Ci
Bd input disturbance matrix
Bij input matrix of player i for player j’s inputs
Bi input matrix of player i
Cij output matrix of player i for player j’s interaction states
Ci output matrix of player i
Cd output disturbance matrix
C controllability matrix
C∗ polar cone of cone C
d integrating disturbance
E, F constraint matrices, F x + Eu ≤ e
f,h system functions, discrete time, x + = f (x, u), y = h(x)
fc (x, u) system function, continuous time, ẋ = fc (x, u)
F (x, u) difference inclusion, x + ∈ F (x, u), F is set valued
G input noise-shaping matrix
Gij steady-state gain of player i to player j
H controlled variable matrix
I(x, u) index set of constraints active at (x, u)
0
I (x) index set of constraints active at (x, u0 (x))
k sample time
K optimal controller gain
`(x, u) stage cost
`N (x, u) final stage cost
L optimal estimator gain
m input dimension
M cross-term penalty matrix x 0 Mu
M number of players, Chapter 6
M class of admissible input policies, µ ∈ M
n state dimension
N horizon length
O observability matrix, Chapters 1 and 4
O compact robust control invariant set containing the origin,
Chapter 3
p output dimension
xlii Notation
p optimization iterate, Chapter 6

pξ probability density of random variable ξ
P
ps (x) sampled probability density, ps (x) = i wi δ(x − xi )
P covariance matrix in the estimator
Pf terminal penalty matrix
P polytopic partition, Chapter 3
P polytopic partition, Chapter 7
PN (x) MPC optimization problem; horizon N and initial state x
q importance function in importance sampling
Q state penalty matrix
r controlled variable, r = Hy
R input penalty matrix
s number of samples in a sampled probability density
S input rate of change penalty matrix
S(x, u) index set of active polytopes at (x, u)
0
S (x) index set of active polytopes at (x, u0 (x))
t time
T current time in estimation problem
u input (manipulated variable) vector
+
u
e warm start for input sequence
u+ improved input sequence
UN (x) control constraint set
U input constraint set
v output disturbance, Chapters 1 and 4
v nominal control input, Chapters 3 and 5
VN (x, u) MPC objective function
VN0 (x) MPC optimal value function
VT (χ, ω) Full information state estimation objective function at time T
with initial state χ and disturbance sequence ω
V̂T (χ, ω) MHE objective function at time T with initial state χ and distur-
bance sequence ω
Vf (x) terminal penalty
VN (z) nominal control input constraint set
V output disturbance constraint set
w disturbance to the state evolution
wi weights in a sampled probability density, Chapter 4
wi convex weight for player i, Chapter 6
wi normalized weights in a sampled probability density
W class of admissible disturbance sequences, w ∈ W
Notation xliii
W state disturbance constraint set

x state vector
xi sample values in a sampled probability density
xij state interaction vector from player i to player j
x(0) mean of initial state density
X(k; x, µ) state tube at time k with initial state x and control policy µ
Xj set of feasible states for optimal control problem at stage j
X state constraint set
Xf terminal region
y output (measurement) vector
Y output constraint set
z nominal state, Chapters 3 and 5
ZT (x) full information arrival cost
ẐT (x) MHE arrival cost
e T (x)
Z MHE smoothing arrival cost
Z system constraint region, (x, u) ∈ Z
Zf terminal constraint region, (x, u) ∈ Zf
ZN (x, u) constraint set for state and input sequence
Greek letters
ΓT (χ) MHE prior weighting on state at time T

∆ sample time
κ control law
κj control law at stage j
κf control law applied in terminal region Xf
µi (x) control law at stage i
µ(x) control policy or sequence of control laws
ν output disturbance decision variable in estimation problem
Π cost-to-go matrix in regulator, Chapter 1
Π covariance matrix in the estimator, Chapter 5
ρi objective function weight for player i
Σi Solution to Lyapunov equation for player i
φ(k; x, u) state at time k given initial state x and input sequence u
φ(k; x, i, u) state at time k given state at time i is x and input sequence u
φ(k; x, u, w) state at time k given initial state is x, input sequence is u, and
disturbance sequence is w
χ state decision variable in estimation problem
ω state disturbance decision variable in estimation problem
xliv Notation
Subscripts, superscripts, and accents

x̂ estimate
x̂ −
estimate before measurement
x
e estimate error
xs steady state
xi subsystem i in a decomposed large-scale system
xsp setpoint
V0 optimal
V uc unconstrained
V sp unreachable setpoint
Acronyms
AD algorithmic (or automatic) differentiation

AS asymptotically stable
BFGS Broyden-Fletcher-Goldfarb-Shanno
CLF control Lyapunov function
DAE differential algebraic equation
DARE discrete algebraic Riccati equation
DDP differential dynamic programming
DP dynamic programming
END external numerical differentiation
FIE full information estimation
FLOP floating point operation
FSO final-state observable
GAS globally asymptotically stable
GES globally exponentially stable
GL Gauss-Legendre
GPC generalized predictive control
EKF extended Kalman filter
i-IOSS incrementally input/output-to-state stable
IND internal numerical differentiation
i-OSS incrementally output-to-state stable
IOSS input/output-to-state stable
IP interior point
ISS input-to-state stable
i-UIOSS incrementally uniformly input/output-to-state stable
KF Kalman filter
KKT Karush-Kuhn-Tucker
LAR linear absolute regulator
LICQ linear independence constraint qualification
LP linear program
LQ linear quadratic
LQG linear quadratic Gaussian
LQP linear quadratic problem
LQR linear quadratic regulator
MHE moving horizon estimation
xlv
Another random document with
no related content on Scribd:
Bartlett—[Gruffly.] I’ve naught to forget—leastways naught that’s in
your mind. But they’s things about the stubborn will o’ woman I’d like to
forget. [They look at each other across the table. There is a pause. Finally
he cannot stand her accusing glance. He looks away, gets to his feet, walks
about, then sits down again, his face set determinedly—with a grim smile.]
Well, here we be, Sarah—alone together for the first time since—
Mrs. Bartlett—[Quickly.] Since that night, Isaiah.
Bartlett—[As if he hadn’t heard.] Since I come back to you, almost.
Did ye ever stop to think o’ how strange it be we’d ever come to this? I
never dreamed a day’d come when ye’d force me to sleep away from ye,
alone in a shed like a mangy dog!
Mrs. Bartlett—[Gently.] I didn’t drive you away, Isaiah. You came o’
your own will.
Bartlett—Because o’ your naggin’ tongue, woman—and the wrong ye
thought o’ me.
Mrs. Bartlett—[Shaking her head, slowly.]
It wasn’t me you ran from, Isaiah. You ran away from your own self—
the conscience God put in you that you think you can fool with lies.
Bartlett—[Starting to his feet—angrily.] Lies?
Mrs. Bartlett—It’s the truth, Isaiah, only you be too weak to face it.
Bartlett—[With defiant bravado.] Ye’ll find I be strong enough to face
anything, true or lie! [Then protestingly.] What call have ye to think evil o’
me, Sarah? It’s mad o’ ye to hold me to account for things I said in my
sleep—for the damned nightmares that set me talkin’ wild when I’d just
come home and my head was still cracked with the thirst and the sun I’d
borne on that island. Is that right, woman, to be blamin’ me for mad
dreams?
Mrs. Bartlett—You confessed the rest of what you said was true—of
the gold you’d found and buried there.
Bartlett—[With a sudden fierce exultation.] Aye—that be true as
Bible, Sarah. When I’ve sailed back in the schooner, ye’ll see for yourself.
There be a big chest o’ it, yellow and heavy, and fixed up with diamonds,
emeralds and sech, that be worth more, even, nor the gold. We’ll be rich,
Sarah—rich like I’ve always dreamed we’d be! There’ll be silks and
carriages for ye—all the woman’s truck in the world ye’ve a mind to want
—and all that Nat and Sue’ll want, too.
Mrs. Bartlett—[With a shudder.] Are you tryin’ to bribe me, Isaiah—
with a treasure that’s been cursed by God?
Bartlett—[As if he hadn’t heard.] D’ye remember long ago, back East,
just after we was married, and I was skipper o’ my first whalin’ ship, how
that foreigner come to me with the map o’ the pirates’ gold and asked me to
charter the ship? D’ye remember o’ how I’d talk to ye o’ findin’ ambergris,
a pile o’ it on one vige that’d make us rich? Ye used to take interest then,
and all th’ voyage with me ye’d be hopin’ I’d find it, too.
Mrs. Bartlett—That was my sin o’ greed that I’m bein’ punished for
now.
Bartlett—[Again as if he hadn’t heard.] And now when it’s come to us
at last—bigger nor I ever dreamed on—ye drive me away from ye and say
it’s cursed.
Mrs. Bartlett—[Inexorably.] Cursed with the blood o’ the man and
boy ye murdered!
Bartlett—[In a mad rage.] Ye lie, woman! I spoke no word!
Mrs. Bartlett—That’s what you kept repeatin’ in your sleep, night
after night that first week you was home, till I knew the truth, and could
bear no more. “I spoke no word!” you kept sayin’, as if ’twas your own soul
had you at the bar of judgment. And “That cook, he didn’t believe ’twas
gold,” you’d say, and curse him.
Bartlett—[Wildly.] He was lyin’, the thief! Lyin’ so’s he and the boy
could steal th’ gold. I made him own up he was lyin’. What if it’s all true,
what ye heard? Hadn’t we the right to do away with two thieves? And we
was all mad with thirst and sun. Can ye hold madmen to account for the
things they do?
Mrs. Bartlett—You wasn’t so crazed but you remember.
Bartlett—I remember I spoke no word, Sarah—as God’s my judge!
Mrs. Bartlett—But you could have prevented it with a word, couldn’t
you, Isaiah? That heathen savage lives in the fear of you. He’d not have
done it if——
Bartlett—[Gloomily.] That’s woman’s talk. There be three o’ us can
swear in any court I spoke no word.
Mrs. Bartlett—What are courts? Can you swear it to yourself? You
can’t, and it’s that’s drivin’ you mad, Isaiah. Oh, I’d never have believed it
of you for all you said in sleep, if it wasn’t for the way you looked and
acted out of sleep. I watched you that first week, Isaiah, till the fear of it had
me down sick. I had to watch you, you was so strange and fearful to me. At
first I kept sayin’, ’twas only you wasn’t rid o’ the thirst and the sun yet.
But then, all to once, God gave me sight, and I saw ’twas guilt written on
your face, on the queer stricken way you acted, and guilt in your eyes. [She
stares into them.] I see it now, as I always see it when you look at me. [She
covers her face with her hands with a sob.]
Bartlett—[His face haggard and drawn—hopelessly, as if he were too
beaten to oppose her further—in a hoarse whisper.] What would ye have
me do, Sarah?
Mrs. Bartlett—[Taking her hands from her face—her eyes lighting up
with religious fervor.] Confess your sin, Isaiah! Confess to God and men,
and make your peace and take your punishment. Forget that gold that’s
cursed and the voyage you be settin’ out on, and make your peace.
[Passionately.] I ask you to do this for my sake and the children’s, and your
own most of all! I’ll get down on my knees, Isaiah, and pray you to do it, as
I’ve prayed to God to send you his grace! Confess and wash your soul of
the stain o’ blood that’s on it. I ask you that, Isaiah—and God asks you—to
make your peace with Him.
Bartlett—[His face tortured by the inward struggle—as if the word
strangled him.] Confess and let someone steal the gold! [This thought
destroys her influence over him in a second. His obsession regains
possession of him instantly, filling him with rebellious strength. He laughs
harshly.] Ye’d make an old woman o’ me, would ye, Sarah?—an old,
Sunday go-to-meetin’ woman snivvelin’ and prayin’ to God for pardon!
Pardon for what? Because two sneakin’ thieves are dead and done for? I
spoke no word, I tell ye—but if I had, I’d not repent it. What I’ve done I’ve
done, and I’ve never asked pardon o’ God or men for ought I’ve done, and
never will. Confess, and give up the gold I’ve dreamed of all my life that
I’ve found at last! By thunder, ye must think I’m crazed!
Mrs. Bartlett—[Seeming to shrivel up on her chair as she sees she has
lost—weakly.] You be lost, Isaiah—no one can stop you.
Bartlett—[Triumphantly.] Aye, none’ll stop me. I’ll go my course
alone. I’m glad ye see that, Sarah.
Mrs. Bartlett—[Feebly trying to get to her feet.] I’ll go to home.
Bartlett—Ye’ll stay, Sarah. Ye’ve had your say, and I’ve listened to
ye; now I’ll have mine and ye listen to me. [Mrs. Bartlett sinks back in
her chair exhaustedly. Bartlett continues slowly.] The schooner sails at
dawn on the full tide. I ask ye again and for the last time, will ye christen
her with your name afore she sails?
Mrs. Bartlett—[Firmly.] No.
Bartlett—[Menacingly.] Take heed, Sarah, o’ what ye’re sayin’! I’m
your husband ye’ve sworn to obey. By right I kin order ye, not ask.
Mrs. Bartlett—I’ve never refused in anything that’s right—but this be
wicked wrong.
Bartlett—It’s only your stubborn woman’s spite makes ye refuse.
Ye’ve christened every ship I’ve ever been skipper on, and it’s brought me
luck o’ a kind, though not the luck I wanted. And we’ll christen this one
with your own name to bring me the luck I’ve always been seekin’.
Mrs. Bartlett—[Resolutely.] I won’t, Isaiah.
Bartlett—Ye will, Sarah, for I’ll make ye. Ye force me to it.
Mrs. Bartlett—[Again trying to get up.] Is this the way you talk to me
who’ve been a good wife to you for more than thirty years?
Bartlett—[Commandingly.] Wait! [Threateningly.] If ye don’t christen
her afore she sails, I’ll take Nat on the vige along with me. [Mrs. Bartlett
sinks back in her chair, stunned.] He wants to go, ye know it. He’s asked me
a hundred times. He s’spects—’bout the gold—but he don’t know for sartin.
But I’ll tell him the truth o’ it, and he’ll come with me, unless—
Mrs. Bartlett—[Looking at him with terror-stricken eyes—
imploringly.] You won’t do that, Isaiah? You won’t take Nat away from me
and drag him into sin? I know he’ll go if you give him the word, in spite of
what I say. [Pitifully.] You be only frightenin’ me! You can’t be so wicked
cruel as that.
Bartlett—I’ll do it, I take my oath—unless—
Mrs. Bartlett—[With hysterical anger.] Then I’ll tell him myself—of
the murders you did, and—
Bartlett—[Grimly.] And I’ll say ’twas done in fair fight to keep them
from stealin’ the gold! I’ll tell him your’s is a woman’s notion, and he’ll
believe me, not you. He’s his father’s son, and he’s set to go. Ye know it,
Sarah. [She falls back in the chair hopelessly staring at him with horrified
eyes. He turns away and adds after a pause.] So ye’ll christen the Sarah
Allen in the mornin’ afore she sails, won’t ye, Sarah?
Mrs. Bartlett—[In a terrified tone.] Yes—if it’s needful to save Nat—
and God’ll forgive me when He sees my reason. But you—Oh, Isaiah! [She
shudders and then breaks down, sobbing.]
Bartlett—[After a pause, turns to her humbly as if asking her
forgiveness.] Ye mustn’t think hard o’ me that I want your name. It’s
because it’s a good woman’s name, and I know it’ll bring luck to our vige.
I’d find it hard to sail without it—the way things be.
Mrs. Bartlett—[Getting to her feet—in a state of feverish fear of him.]
I’m goin’ to home.
Bartlett—[Going to her.] I’ll help ye to the top o’ the hill, Sarah.
Mrs. Bartlett—[Shrinking from him in terror.] No. Don’t you touch
me! Don’t you touch me! [She hobbles quickly out of the door in the rear,
looking back frightenedly over her shoulder to see if he is following as
[The Curtain Falls]

ACT THREE
Scene—Dawn of the following morning—exterior of the Bartlett home,
showing the main entrance, facing left, toward the harbor. On either side
of the door, two large windows, their heavy green shutters tightly closed.
In front of the door, a small porch, the roof supported by four white
column. A flight of three steps goes up to this porch from the ground. Two
paths lead to the steps through the straggly patches of grass, one around
the corner of the house to the rear, the other straight to the left to the edge
of the cliff where there is a small projecting iron platform, fenced in by a
rail. The top of a steel ladder can be seen. This ladder leads up the side or
the cliff from the shore below to the platform. The edge of the cliff extends
from the left corner front, half-diagonally back to the right, rear-center.
In the grey half-light of the dawn, Horne, Cates, and Jimmy Kanaka are
discovered. Horne is standing on the steel platform looking down at the
shore below. Cates is sprawled on the ground nearby. Jimmy squats on his
haunches, his eyes staring out to sea as if he were trying to pierce the
distance to the warm islands of his birth. Cates wears dungarees, Jimmy
dungaree pants and a black jersey; Horne, the same as in Act Two.
Cates—[With sluggish indifference.] Ain’t she finished with it yet?
Horne—[Irritably.] No, damn her! I kin see ’em all together on the
wharf at the bow o’ the schooner. That old crow o’ a woman o’ his! Why
the hell don’t she christen her and be done with it and let us make sail?
Cates—[After a pause.] Funny, ain’t it—his orderin’ us to come up here
and wait till it’s all done.
Horne—[Angrily.] That’s her doin’, too. She thinks we ain’t good
enough to be where she is. [After a pause.] But there’s nothin’ funny to me
that he does no more. He’s still out o’ his head, d’ye know that, Cates?
Cates—[Stupidly.] I ain’t noticed nothin’ diff’rent ’bout him.
Horne—[Scornfully.] He axed me if I ever seen them two in my sleep—
that cook and the boy o’ the Triton. Said he did often.
Cates—[Immediately protesting uneasily as if he had been accused.]
They was with us in the boat b’fore we fetched the island, that’s all ’bout
’em I remember. I was crazy, after.
Horne—[Looking at him with contempt.] So was we all crazy, for the
matter o’ that. I’ll not call ye a liar, Cates, but—a hell o’ a man ye be! You
wasn’t so out o’ your head that ye forgot the gold, was ye?
Cates—[His eyes glistening.] That’s diff’rent. Any man’d remember
that, even if he was crazy.
Horne—[With a greedy grin.] Aye. That’s the one thing I see in my
sleep. [Gloatingly.] We’ll dig it up soon now. In three months we’d ought to
be there—an’ then we’ll be rich, by Christ! [There is the faint sound of cries
from the beach below. Horne starts and turns to look down again.] They
must ’a’ finished it. [Cates and Jimmy come to the edge to look down.]
Jimmy—[Suddenly—with an eager childish curiosity.] That falla wife
Captain she make strong falla spell on ship, we sail fast, plenty good wind?
Horne—[Contemptuously.] Aye, that’s as near as ye’ll come to it. She’s
makin’ a spell. Ye stay here, Jimmy, and tell us when the Old Man is
comin’. [Jimmy remains looking down. Horne motions Cates to follow him,
front—then in a low voice, disgustedly.] Did ye hear that damn fool nigger?
Cates—[Grumblingly.] Why the hell is the Old Man givin’ him a full
share? One piece o’ it’d be enough for a nigger like him.
Horne—[Craftily.] There’s a way to get rid o’ him—if it comes to that.
He knifed them two, ye remember.
Cates—Aye.
Horne—The two o’ us can take oath to that in any court.
Cates—Aye.
Horne—[After a calculating look into his companion’s greedy eyes—
meaningly.] We’re two sane men, Cates—and the other two to share is a
lunatic and a nigger. The skipper’s showed me where there’s a copy o’ his
map o’ the island locked up in the cabin—in case anything happens to him
I’m to bring back the gold to his woman, he says. [He laughs harshly.]
Bring it back! Catch me! The fool! I’ll be open with ye, Cates. If I could
navigate and find the island myself I wouldn’t wait for a cracked man to
take me there. No, be damned if I would! Me and you’d chance it alone
someway or other.
Cates—[Greedily.] The two o’ us—share and share alike! [Then shaking
his head warningly.] But he’s a hard man to git the best on.
Horne—[Grimly.] And I be a hard man, too. And he’s not right in his
head. We’ll keep our eyes peeled for a chance. Something may turn up—
and maybe—
Jimmy—[Turning to them.] Captain, he come. [Cates and Horne
separate hastily. Bartlett climbs into sight up the ladder to the platform.
He is breathing heavily but his expression is one of triumphant exultation.]
Bartlett—[Motions with his arms.] Down with ye and git aboard. The
schooner’s got a name now—a name that’ll bring us luck. We’ll sail on this
tide.
Horne—Aye—aye, sir.
Bartlett—I got to wait here till they climb up the path. I’ll be aboard
afore long. See that ye have her ready to cast off by then.
Horne—Aye—aye, sir. [He and Cates disappear down the ladder.
Jimmy lingers, looking sidewise at his Captain.]
Bartlett—[Noticing him—gruffly but almost kindly.] What are ye
waitin’ for?
Jimmy—[Volubly.] That old falla wife belong you, Captain, she make
strong falla spell for wind blow plenty? She catch strong devil charm for
schooner, Captain?
Bartlett—[Scowling.] What’s that, ye brown devil? [Then suddenly
laughing harshly.] Yes—a strong spell to bring us luck. [Roughly.] Git
aboard, ye dog! Don’t let her find ye here with me. [Jimmy disappears
hurriedly down the ladder. Bartlett remains at the edge looking down
after him. There is a sound of voices from the right and presently Mrs.
Bartlett, Sue, Drew and Nat enter, coming around the house from the
rear. Nat and Drew walk at either side of Mrs. Bartlett, who is in a state
of complete collapse, so that they are practically carrying her. Sue follows,
her handkerchief to her eyes. Nat keeps his eyes on the ground, his
expression fixed and gloomy. Drew casts a glance of angry indignation at
the Captain, who, after one indifferent look at them, has turned back to
watch the operations on the schooner below.]
Bartlett—[As they reach the steps of the house—intent on the work
below—makes a megaphone of his hands and shouts in stentorian tones.]
Look lively there, Horne!
Sue—[Protestingly.] Pa!
Bartlett—[Wheels about. When he meets his daughter’s eyes he
controls his angry impatience and speaks gently.] What d’ye want, Sue?
Sue—[Pointing to her mother who is being assisted through the door—
her voice trembling.] You mustn’t shout. She’s very sick.
Bartlett—[Dully, as if he didn’t understand.] Sick?
Sue—[Turning to the door.] Wait. I’ll be right back. [She enters the
house. As soon as she is gone all of Bartlett’s excitement returns. He
paces up and down with nervous impatience. Nat comes out of the house.]
Nat—[In a tone of anxiety.] Ma seems bad. We can’t do anything. I’m
going for the doctor. [As his father doesn’t seem to hear him—tapping him
on the shoulder, his voice breaking.] Why did you make her do it, Pa? It
was too much for her strength. Wouldn’t anyone else or any other name
have done just as well?
Bartlett—[Impatiently.] No. It had to be.
Nat—When she spoke the words—and fell back in a faint—I thought
she was dead.
Bartlett—[Vaguely.] Weakness. She’ll be all right again after a rest.
[He draws Nat’s attention to the schooner.] Smart lines on that schooner,
boy. She’ll sail hell bent in a breeze. I knowed what I was about when I
bought her.
Nat—[Staring down fascinatedly.] How long will the voyage take?
Bartlett—[Preoccupied.] How long?
Nat—[Insinuatingly.] To get to the island.
Bartlett—Three months at most—with fair luck. [Exultantly.] And I’ll
have luck now!
Nat—Then in six months you may be back—with it?
Bartlett—Aye, with—[Stopping abruptly, turns and stares into his
son’s eyes—angrily.] With what? What boy’s foolishness be ye talkin’?
Nat—[Pleading fiercely.] I want to go, Pa! There’s no good in my
staying here any more. I can’t think of anything but—Oh, why don’t you be
fair and let me sail with you!
Bartlett—[Sternly, to conceal his uneasiness.] Keep clear o’ this, boy,
I’ve warned ye!
Sue—[Appearing in doorway—indignantly.] Nat! Haven’t you gone for
the doctor yet?
Nat—[Shame-facedly.] I forgot.
Sue—Forgot!
Nat—[Starting off.] I’m going, Sue. [Then over his shoulder.] You won’t
sail before I come back, Pa? [Bartlett does not answer. Nat stands
miserably hesitating.]
Sue—Nat! For heaven’s sake! [Nat hurries off around the corner of the
house, rear. Sue comes to her father who is watching her with a queer,
humble, hunted expression.]
Bartlett—Well, Sue?
Sue—[Her voice trembling.] Oh, Pa, how can you do such terrible
things. How could you drag Ma out of bed at dawn to christen your old boat
—when you knew how sick she’s been!
Bartlett—[Avoiding her eyes.] It’s only weakness. She’ll get well o’ it
soon.
Sue—Pa! How can you say things like that—as if you didn’t care!
[Accusingly.] The way you’ve acted ever since you’ve been home almost,
anyone would think—you hated her!
Bartlett—[Wincing.] No!
Sue—Oh, Pa, what is it that has come between you? Can’t you tell me?
Can’t I help to set things right again?
Bartlett—[Mumblingly.] Nothin’—nothin’ ye kin help—nor me. Keep
clear o’ it, Sue. Danny—ye think o’ him, that’s enough for ye.
Sue—But things can’t go on like this. Don’t you see how it’s killing Ma?
Bartlett—She’ll forget her stubborn notions, now I be sailin’ away.
Sue—But you’re not—not going for a while now, are you?
Bartlett—Ain’t I been sayin’ I’d sail at dawn today? They’re makin’
her ready to cast off. I’m waitin’ for Horne to hail.
Sue—[Looking at him for a moment with shocked amazement.] But—
you can’t mean—right now!
Bartlett—[Keeping his face averted.] Aye—or we’ll miss this tide.
Sue—[Putting her hands on his shoulders and trying to look into his
face.] Pa! You can’t mean that! [His face is set with his obsessed
determination. She lets her hands fall with a shudder.] You can’t be as cruel
as that! Why, I thought, of course, you’d put off—[Wildly.] You have,
haven’t you, Pa? You did tell those men you couldn’t sail when you saw
how sick Ma was, didn’t you—when she fainted down on the wharf?
Bartlett—[Implacably.] I said I was sailin’ by this tide—and sail I
will, by thunder!
Sue—Pa! [Then pleadingly.] When the doctor comes and you hear what
he says—
Bartlett—[Roughly.] I ain’t stoppin’ on his word nor any man’s. I
know what’s best to do. [Intensely.] That schooner’s been fit to sail these
two weeks past. I been waitin’ on her stubborn will [he gestures toward the
house], eatin’ my heart out day and night. Then I swore I’d sail today. I tell
ye, Sue, I got a feelin’ in my bones if I don’t put out now I never will. Aye,
I feel it deep down inside me. [In a tone of superstitious awe.] And when
she christened the schooner—jest to the minute, mind ye!—a fair breeze
sprung up and come down out o’ the land to blow her out to sea—like a
sign o’ good luck.
Sue—[Aroused to angry indignation.] What kind of a man have you
become—to think of such things now! Oh, I can’t believe you’re the same
man who used to be my father!
Bartlett—Sue!
Sue—To talk cold-bloodedly of sailing away on a long voyage when
Ma’s inside—dying for all you seem to know or care! Oh, I hate you when
you’re like this! You’re not the father I love! You’ve changed into someone
else—hateful and cruel—and I hate him, I hate him! [She breaks down,
sobbing hysterically.]
Bartlett—[Who has listened to her with a face suddenly stricken by
fear and torturing remorse.] Sue! Ye don’t know what ye be sayin’, do ye?
Sue—I do! You’re not the same to me any more—or to any of us. I’m
afraid of you. And when you coldly propose to go away—now—I hate you,
yes I do! And I hate those three awful men who make you act this way. I
hate the schooner! I wish she and they were at the bottom of the sea!
Bartlett—[Frenziedly—putting his hand over her mouth to stop her
words.] Stop, girl! Don’t ye dare—
Sue—[Shrinking away from him—frightenedly.] Pa!
Bartlett—[Bewilderedly, pleading for forgiveness.] Don’t heed that,
Sue—I didn’t mean—ye git me so riled—I’d not hurt ye for all the gold in
the world. But don’t ye talk wrong o’ things ye can’t know on.
Sue—Oh, Pa, what kind of things must they be—when you’re ashamed
to tell them!
Bartlett—I ain’t ashamed. It ain’t that. On’y they be things a girl’s no
call to meddle in. They be men’s business and I be man enough to carry ’em
out alone. Ye’ll know all they be to know—and your Ma and Nat, too—
when I come back from this vige. And the sooner I sail, the quicker I’ll be
back to ye. Oh, ye’ll be glad enough then—when ye see with your own
eyes! Ye’ll bless me then ’stead o’ turning agin me! [Hesitating for a second
—then somberly.] On’y now—till it’s all over and done—ye’d best keep
clear o’ it.
Sue—[Passionately.] I don’t care—I don’t want to know anything about
it. What I do know is that you can’t sail now. Oh, Pa, don’t you see you
can’t? Haven’t you any heart at all? Can’t you see how bad Ma is?
Bartlett—It’s the sight o’ me sickens her. She’ll git better with me
away from her.
Sue—No. She needs you. She doesn’t want you to go. She called your
name just a while ago—the only word she’s spoken since she christened the
ship. Come in to her, Pa! Tell her you won’t go!
Bartlett—[Desperately.] I got to git away from her, I tell ye, Sue!
She’s been houndin’ me ever since I got back—houndin’ me with her
stubborn tongue till she’s druv me mad, a’most! Ye’ve been on’y givin’
thought to her, not me. They’s my side to it, too!
Sue—I’ll talk to her, Pa. She can’t realize she’s hurting you or she
wouldn’t—And then everything will be just the same as it used to be again.
Bartlett—[Shaking his head.] They be too much between. The only
chance for that be my plan—to sail away and come back with—what I be
seekin’. Then she’ll give over her stubborn naggin’—if she’s human
woman. It’s for her sake as much as my own I’m goin’—for her and you
and Nat. [With a sudden return of his old resolution.] I’ve made up my
mind, I tell ye, and in the end ye’ll know I be right. [A hail in Horne’s
voice comes thinly up from the shore below. Bartlett starts, his eyes
gleaming.] Ye hear? It’s Horne hailin’ me to come. They be ready to cast
off. I’ll git aboard. [He starts for the ladder.]
Sue—Pa! After all I’ve said—without one word of good-bye to Ma!
[Hysterically.] Oh, what can I do, what can I say to stop you! She hasn’t
spoken but that one call for you. She hardly seems to breathe. If it weren’t
for her eyes I’d believe she was dead—but her eyes look for you. She’ll die
if you go, Pa!
Bartlett—No!
Sue—You might just as well kill her now in cold blood as murder her
that way!
Bartlett—[Shaken—raising his hands as if to put them over his ears to
shut out her words—hoarsely.] No! Ye lie! She’ll live till I git back and
all’ll be as it was again!
Drew—[Appearing in the doorway, his face working with grief and
anger—harshly.] Captain Bartlett! [Then lowering his voice as he sees Sue.]
Mrs. Bartlett is asking to see you, Captain, before you go.
Sue—There! Didn’t I tell you, Pa!
Bartlett—[Struggling with himself—dully.] She’s wantin’ to bound me
again, that be all.
Sue—[Seeing him weakening—grasps his hand persuasively.] Pa! Come
with me. She won’t hound you. How silly you are! Come! [Hesitatingly,
head bowed, he follows her toward the door.]
Bartlett—[As he comes to Drew he stops and looks into the young
man’s angry, accusing face. He mutters half mockingly.] So ye, too, be agin
me, Danny?
Drew—[Unable to restrain his indignation.] What man that’s a real man
wouldn’t be against you, sir?
Sue—[Frightenedly.] Danny! Pa!
Bartlett—[In a sudden rage draws back his fist threateningly. Drew
stares into his eyes unflinchingly—Bartlett controls himself with an effort
and lets his arm fall to his side—scornfully.] Big words from a boy, Danny.
I’ll forget them this time—on account o’ Sue. [He turns to her.] I’m goin’ in
to her to please ye, Sue—but if ye think any words that she kin say’ll
change my mind, ye make a mistake—for I be sailin’ out as I planned I
would in spite o’ all hell! [He walks resolutely into the house. Sue follows
him after exchanging a hopeless glance with Danny.]
Drew—[To himself—with a shudder.] He’s mad, damn him! [He paces
up and down. Horne appears on the ladder from below, followed by
Cates.]
Horne—[Coming forward and addressing Drew.] Is the skipper about?
Drew—[Curtly.] He’s in the house. You can’t speak to him now.
Horne—She’s ready to cast off. I hailed him from below but I ’spect he
didn’t hear. [As Drew makes no comment—impatiently.] If he don’t shake a
leg, we’ll miss the tide. There’s a bit o’ fair breeze, too.
Drew—[Glancing at him resentfully.] Don’t count on his sailing today.
It’s just as likely he’ll change his mind.
Horne—[Angrily.] Change his mind again? After us waitin’ and wastin’
time for weeks! [To Cates in a loud tone so Drew can hear.] What did I tell
ye, Cates? He’s crazy as hell.
Drew—[Sharply.] What’s that?
Horne—I was tellin’ Cates the skipper’s not right in his head [Angrily.]
What man in his senses’d do the way he does?
Drew—[Letting his resentment escape him.] That’s no lie, damn it!
Horne—[Surprised.] Aye, ye’ve seen it, too, have ye? [After a pause.]
Now I axe ye, as a sailor, how’d ye like to be puttin’ out on a vige with a
cracked man for skipper? [Sue comes out of the door, stops with a shudder
of disgust as she sees the two sailors, and stands listening. They do not
notice her presence.]
Drew—It seems to me a crazy voyage all round. What kind of trading is
it you’re to do?
Horne—[Suspiciously.] Ye’ll have to ask the skipper that.
Drew—[With a scornful shrug.] I was forgetting it’s such a dead secret.
That the craziest part, eh? [With sudden interest as if a new idea had come
to him.] But you know all about it, don’t you—what the Captain plans to do
on this voyage—and all that?
Horne—[Dryly.] Aye, as well as himself—but I’m tellin’ no man.
Drew—And I’m not asking. What do you suppose I care about any
sneaking trade deal in the Islands he may have up his sleeve? What I want
to find out is: Do you know enough about this business to make this one
voyage alone and attend to everything—in case the Captain can’t go?
Horne—[Exchanging a quick glance with Cates—trying to hide his
eagerness.] Aye, I could do as well as any man alive. I’ve been sailin’ this
sea for twenty year or more and I know the Island trade inside and out. He
could trust me for it—and I’d make more money for him than he’s likely to
make with his head out o’ gear. [Then scowling.] On’y trouble is, who’d
Captain her if he ain’t goin’?
Drew—[Disappointedly.] Then you don’t know navigation enough for
that?
Horne—I’ve never riz above bo’sun. [Then after a pause in which he
appears to be calculating something—curiously.] Why d’ye ask me them
questions? [Insinuatingly—almost in a whisper.] It can’t be done ’less we
got an officer like you aboard.
Drew—[Angrily.] Eh? What’re you driving at? D’you think I—
Sue—[Who has been listening with aroused interest.] Danny! [She
comes down to him. Horne and Cates bob their heads respectfully and
move back near the platform. Horne watches Sue and Drew out of the
corner of his eye.] Danny, I’ve been listening to what you were saying, but I
don’t understand. What are you thinking of?
Drew [Excitedly.] I was thinking—Listen, Sue! Seems to me from what
I saw your Pa’s out of his right mind, and, being that way, he’s sure bound
to go unless someone or something steps in to stop him. D’you think your
Ma——?
Sue—[Shaking her head—sadly.] No, I’m afraid anything she says will
only make things worse.
Drew—Then you’ve no hope—? No more have I. Something’s got to be
done to keep him home in spite of himself. Even leaving your Ma out of it,
he’s not in any fit state to take a ship to sea; and I was thinking if we could
fix it some way so that fellow Horne could take her out on this voyage—
Sue—But, Danny, Pa’d never give in to that.
Drew—I wasn’t thinking he would. It’d have to be done on the sly. We
—you’d have to give the word—and keep him in the house somehow—and
then when he did come out it’d be too late. The schooner’d be gone.
Sue—[Disturbed, but showing that this plan has caught her mind.] But
—would it be fair?—he’d never forgive—
Drew—When he’s back in his right mind again, he would. [Earnestly.]
I’m not fond of lying and tricks myself, Sue, but this is a case where you
can’t pick and choose. You can’t let him sail, and wreck his ship and
himself in the bargain, likely. Then, there’s your Ma——
Sue—No, no, we can’t let him. [With a glance at Horne and Cates.]
But I don’t trust those men.
Drew—No more do I; but it would be better to chance them than—
[Suddenly interrupting himself—with a shrug of his shoulders.] But there’s
no good talking of that. I was forgetting. None of them can navigate. They
couldn’t take her out.
Sue—But didn’t I hear him say—if they had an officer on board—like
you—
Drew—Yes, but where’ll you find one at a second’s notice?
Sue—[Meaningly.] And you told me, didn’t you, that you’d just got your
master’s papers. Then you’re a captain by rights.
Drew—[Looking at her with stunned astonishment.] Sue! D’you mean
—
Sue—[A light coming over her face.] Oh, Danny, we could trust you!
He’d trust you! And after he’d calmed down I know he wouldn’t mind so
much. Oh, Danny, it’ll break my heart to have you go, to send you away
just after you’ve come back. But I don’t see any other way. I wouldn’t ask
—if it wasn’t for Ma being, this way—and him—Oh, Danny, can’t you see
your way to do it—for my sake?
Drew—[Bewilderedly.] Why, Sue, I—I never thought—[Then as he sees
the look of disappointment which comes over her face at his hesitancy—
resolutely.] Why sure, Sue, I’ll do it—if you want me to. I’ll do it if it can
be done. But we’ve got to hustle. You stand in the door, Sue. You’ve got to
keep him in the house some way if he aims to come out. And I’ll talk to
them. [Sue goes to the doorway. Drew goes over to Horne and Cates.]
Sue—[After listening.] He’s still in with Ma. It’s all right.
Drew—[To Horne, with forced joviality.] How would you like me for
skipper on this one voyage?
Horne—[Craftily.] Ye got your skipper’s papers all reg’lar?
Drew—Yes, that part of it’s all right and square. Listen here. Miss Sue’s
decided her father isn’t in a fit state to Captain this trip. It’d mean danger
for him and the schooner—and for you.
Horne—That’s no lie.
Cates—[To Horne protestingly.] But if we git ketched the Old Man’ll
take it out o’ our hides, not his’n.
Horne—[Savagely.] Shut up, ye fool! [To Drew, craftily.] Cates is right,
jest the same. Ye are as good as his married son and she’s his daughter.
He’d not blame you if things went wrong. He’d take it out on us.
Drew—[Impatiently.] I’ll shoulder all that risk, man!
Sue—[Earnestly.] No harm will come to any of you, I promise you. This
is all my plan, and I’ll tell my father I’m alone to blame.
Horne—[In the tone of one clinching a bargain.] Then we’ll chance it.
[Warningly.] But it’s got to be done smart, sir. Ye’d best look lively.
Drew—I’ve got to get my dunnage. I’ll be right back and we’ll tumble
aboard. [He goes to the door.] Hold him, Sue, on some excuse if he’s
coming. Only a second now and it’ll all be safe. [He goes into the house.
She follows him in.]
Cates—[With stupid anger.] This is a hell o’ a mess we’re gettin’ in, if
ye axe me.
Horne—And I tell ye it’s a great stroke o’ luck. It couldn’t o’ come out
better.
Cates—He’ll be aboard to spy on us.
Horne—Let him! What does he know? He thinks we’re goin’ tradin’,
and there’s no one to tell him diff’rent but me.
Cates—He’ll know better afore long. He’ll s’pect—
Horne—’Bout the gold? He ain’t that kind. He’s a soft young swab o’ a
lady steamer’s mate. Leave me to fool him. And when the time comes to git
rid o’ him, I’ll find a means some way or other. But can’t ye see, ye fool,
it’s luck to have him with us till we git clear o’ civilized ports? He kin
navigate and he’s got skipper’s papers that’ll come in handy if there’s any
trouble. And if anythin’ goes wrong at the start and we’re brung back, him
and the girl’ll take the blame.
Cates—[Stupidly.] S’long as he don’t git no share o’ the gold——
Horne—[Contemptuously.] Share, ye dumbhead! I’d see him in hell first
—and send him there myself. [Drew comes out of the house carrying his
bag which he hands to Cates. Sue follows him.]
Drew—Look lively now! Let’s hustle aboard and get her under way.
Horne—Aye—aye, sir. [He and Cates clamber hurriedly down the
ladder.]
Sue—[Throwing her arms around his neck and kissing him.] Good-bye,
Danny. It’s so fine of you to do this for us! I’ll never forget——
Drew—[Tenderly.] Ssssh! It’s nothing, Sue.
Sue—[Tearfully.] Oh, Danny, I hope I’m doing right! I’ll miss you so
dreadfully! But you’ll come back just as soon as you can—
Drew—Of course!
Sue—Danny! Danny! I love you so!
Drew—And I guess you know I love you, don’t you? [Kisses her.] And
we’ll be married when I come back this time sure?
Sue—Yes—yes—Danny—sure!
Drew—I’ve got to run. Good-bye, Sue.
Sue—Good-bye, dear. [They kiss for the last time and he disappears
down the ladder. She stands at the top, sobbing, following him with her
eyes. Nat comes around the house from the rear and goes to the front door.]
Nat—[Seeing his sister.] Sue! He hasn’t gone yet, has he? [She doesn’t
hear him. He hesitates in the doorway for a moment, listening for the sound
of his father’s voice from inside. Then, very careful to make no noise, he
tiptoes carefully into the house. Sue waves her hand to Drew who has
evidently now got aboard the ship. Then she covers her face with her hands,
sobbing. Nat comes out of the house again and goes to his sister. As she
sees him approaching, she dries her eyes hastily, trying to smile.]
Sue—Did you get the doctor, Nat?
Nat—Yes, he’s coming right away, he promised. [Looking at her face.]
What—have you been crying?
Sue—No. [She walks away from the edge of the cliff, drawing him with
her.]
Nat—Yes, you have. Look at your eyes.
Sue—Oh, Nat, everything’s so awful! [She breaks down again.]
Nat—[Trying to comfort her in an absentminded way.] There, don’t get
worked up. Ma’ll be all right as soon as the doctor comes. [Then curiously.]
Pa’s inside with her. They were arguing—have they made it up, d’you
think?
Sue—Oh, Nat, I don’t know. I don’t think so.
Nat—The strain’s been too much for him—waiting and hiding his secret
from all of us. What do you suppose it is, Sue—ambergris?
Sue—[Wildly.] I don’t know and I don’t care! [Noticing the strange
preoccupied look in his eyes—trying to bring him back to earth—
scornfully.] Ambergris! Are you going crazy? Don’t you remember you’ve
always been the first one to laugh at that silly idea?
Nat—Well, there’s something—— [Starts for the platform. Sue does her
best to interpose to hold him back.] Are they all ready on the schooner.
He’ll have to hurry if she’s going to sail on this tide. [With sudden passion.]
Oh, I’ve got to go! I can’t stay here! [Pleadingly.] Don’t you think, Sue, if
you were to ask him for me he’d—You’re the only one he seems to act sane
with or care about any more.
Sue—No! I won’t! I can’t!
Nat—[Angrily.] Haven’t you any sense? Wouldn’t it be better for
everyone if I went in his place?
Sue—No. You know that’s a lie. Ma would lose her mind if you went.
Nat—And I’ll lose mine if I stay! [Half aware of Sue’s intention to keep
him from looking down at the schooner—irritably.] What are you holding
my arm for, Sue? I want to see what they’re doing. [He pushes her aside
and goes to the platform—excitedly.] Hello, they’ve got the fores’l and
mains’l set. They’re setting the stays’l. [In amazement.] Why—they’re
casting off! She’s moving away from the wharf! [More and more excitedly.]
I see four of them on board! Who—who is that, Sue?
Sue—It’s Danny.
Nat—[Furiously.] Danny! What right has he—when I can’t! Sue, call
Pa! They’re sailing, I tell you, you little fool!
Sue—[Trying to calm him—her voice trembling.] Nat! Don’t be such a
donkey! Danny’s only going a little way—just trying the boat to see how
she sails while they’re waiting for Pa.
Nat—[Uncertainly.] Oh. [Then bitterly.] I was never allowed to do even
that—his own son! Look, Sue, that must be Danny at the stern waving.
Sue—[Brokenly.] Yes. [She waves her handkerchief over her head—then
breaks down, sobbing again. There is the noise of Bartlett’s voice from
inside and a moment later he appears in the doorway. He seems terribly
shattered, at the end of his tether. He hesitates uncertainly, looking about
him wildly as if he didn’t know what to do or where to go.]
Sue—[After one look at his face, runs to him and flings her arms about
his neck.] Pa! [She weeps on his shoulder.]
Bartlett—Sue, ye did wrong beggin’ me to see her. I knowed it’d do
no good. Ye promised she’d not hound me——“Confess,” she says—when
they be naught to tell that couldn’t be swore to in any court. “Don’t go on
this vige,” she says, “there be the curse o’ God on it.” [With a note of
baffled anguish.] She kin say that after givin’ the ship her own name! [With
wild, haggard defiance.] But curse or no curse, I be goin’! [He moves
toward the platform, Sue clinging to his arm.]
Sue—[Frightenedly.] Pa! Go back in the house, won’t you?
Bartlett—I be sorry to go agin your will, Sue, but it’s got to be. Ye’ll
know the reason some day—and be glad o’ it. And now good-bye to ye.
[With a sudden strange tenderness he bends and kisses his daughter. Then
as she seems about to protest further, his expression becomes stern and
inflexible.] No more o’ talk, Sue! I be bound out. [He takes her hand off his
arm and strides to the platform. One look down at the harbor and he stands
transfixed—in a hoarse whisper.] What damned trick be this? [He points to
the schooner and turns to Nat bewilderedly.] Ain’t that my schooner, boy—
the Sarah Allen—reachin’ toward the p’int?
Nat—[Surprised.] Yes, certainly. Didn’t you know? Danny’s trying her
to see how she sails while they’re waiting for you.
Bartlett—[With a tremendous sigh of relief.]
Aye. [Then angrily.] He takes a lot o’ rope to himself without askin’
leave o’ me. Don’t he know they’s no time to waste on boy’s foolin’? [Then
with admiration.] She sails smart, don’t she, boy? I knowed she’d show a
pair o’ heels.
Nat—[With enthusiasm.] Yes, she’s a daisy! Say, Danny’s taking her
pretty far out, isn’t he?
Bartlett—[Anxiously.] He’d ought to come about now if he’s to tack
back inside the p’int. [Furiously.] Come about, damn ye! The swab! That’s

Model Predictive Control Theory Computation and Design 2Nd Edition Rawlings James B Full Chapter

Uploaded by

Copyright:

Available Formats

You might also like

Model Predictive Control Theory Computation and Design 2Nd Edition Rawlings James B Full Chapter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Model Predictive Control Theory Computation and Design 2Nd Edition Rawlings James B Full Chapter

Uploaded by

Copyright:

Available Formats

Model Predictive Control: Theory,

Computation, and Design, 2nd Edition

Santa Barbara, California

Cover design by Cheryl M. and James B. Rawlings.

Copyright c 2019 by Nob Hill Publishing, LLC

All rights reserved.

Nob Hill Publishing, LLC

No part of this book may be reproduced, in any form or by any means,

Library of Congress Control Number: 2017909542

Printed in the United States of America.

for their love, encouragement, and patience.

reasons for this removal; first, we wanted to maintain a manageable

JBR DQM MMD

Our goal in this text is to provide a comprehensive and foundational

covers extended and unscented Kalman filtering, and particle filtering.

methods and their presentation during the last 20 years.

1 Getting Started with Model Predictive Control 1

2 Model Predictive Control—Regulation 89

2.4.2 Stabilizing Conditions . . . . . . . . . . . . . . . . . 114

3 Robust and Stochastic Model Predictive Control 193

3.4 Robust Min-Max MPC . . . . . . . . . . . . . . . . . . . . . . 220

4 State Estimation 269

4.7 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

5 Output Model Predictive Control 339

6 Distributed Model Predictive Control 369

6.5.3 Distributed Nonlinear Cooperative Control . . . . 425

7 Explicit Control Laws for Constrained Linear Systems 451

8 Numerical Optimal Control 491

8.3.4 Affine Invariance . . . . . . . . . . . . . . . . . . . . . 519

Author Index 601

Citation Index 608

Subject Index 614

A Mathematical Background 624

B Stability Theory 693

B.11 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724

1.1 System with input u, output y, and transfer function ma-

2.1 Example of MPC. . . . . . . . . . . . . . . . . . . . . . . . . . . 101

2.5 Closed-loop economic MPC versus tracking MPC starting

3.1 Open-loop and feedback trajectories. . . . . . . . . . . . . . . 198

3.8 Concentration versus time for the ancillary model predic-

4.1 MHE arrival cost Ẑk (p), underbounding prior weighting

7.1 The sets Z, X, and U(x). . . . . . . . . . . . . . . . . . . . . . 454

8.1 Feasible set and reduced objective ψ(u(0)) of the non-

A.1 The four fundamental subspaces of matrix A . . . . . . . . 626

B.1 Stability of the origin. . . . . . . . . . . . . . . . . . . . . . . . 697

C.1 Routing problem. . . . . . . . . . . . . . . . . . . . . . . . . . . 730

1.1 Example: Sum of quadratic functions . . . . . . . . . . . . . 15

2.1 Proposition: Continuity of system solution . . . . . . . . . . 94

2.20 Definition: Exponential stability . . . . . . . . . . . . . . . . . 120

2.52 Proposition: Asymptotic average performance . . . . . . . . 155

3.1 Definition: Robust global asymptotic stability . . . . . . . . 207

4.1 Definition: i-IOSS . . . . . . . . . . . . . . . . . . . . . . . . . . 272

4.37 Proposition: Locally Lipschitz upper bound . . . . . . . . . . 302

5.1 Definition: Positive invariance; robust positive invariance . 345

6.1 Algorithm: Suboptimal MPC (simplified) . . . . . . . . . . . . 375

6.14 Lemma: Global asymptotic stability and exponential con-

7.1 Definition: Polytopic (polyhedral) partition . . . . . . . . . . 456

8.1 Example: Nonlinear MPC . . . . . . . . . . . . . . . . . . . . . 495