340
Book reviews / Automatica 40 (2004) 337 – 342
motion and trajectory following). Landmark Recognition in Indoor Navigation by Fuzzy Maps and CBR (Chapter 9) aims of using techniques from Arti5cial Intelligence to design recognition algorithms with the objective of rendering a robot able to navigate autonomously in an unknown indoor environment. The localization problem for a mobile robot is addressed in Sensor Fusion for Robot Localization (Chapter 10), using data coming from different sensors, which are properly combined (fused). The techniques presented in the three chapters are validated by experimental results, which are carried out by using their own prototypes (the SuperMARIO mobile robot) and commercial devices (the nomad mobile robots 150 and 200). In conclusion, the book is written from an engineering perspective (with emphasis on control theory and mechanics), with a good balance between theory and practice for some advanced topics in robotics: the formulas and properties are illustrated and commented properly. An extensive bibliography (435 references) provides pointers to more complete papers and books, where the reader can 5nd details of the presented material. There is a strong emphasis on an intuitive understanding of the topics treated. The advantages and limitations of the various techniques considered
in the book are described in detail for the bene5t of the potential user. Antonio TornambTe Dipartimento di Informatica; Sistemi e Produzione; Universitu@e di Roma Tor Vergata; via del Politecnico; 1; 00133; Rome; Italy E-mail address:
[email protected] Antonio Tornamb e is full professor of Industrial Robotics and Control Theory at UniversitTa di Roma Tor Vergata since 2001 and before he held various research and teaching positions at Fondazione Ugo Bordoni, UniversitTa di Roma Tor Vergata, Politecnico di Torino, UniversitTa di Roma Tre, UniversitTa di Siena. His research interests are in the area of Control Theory and in its application for the control of robotic manipulators and mechanical systems; his research results have been reported in about 200 papers published in international journals and presented at international conferences. He is the author of “Discrete-Event System Theory: an introduction,” World Scienti5c Publishing, Singapore, 1995, and co-author of “Mathematical Methods for System Theory,” World Scienti5c Publishing, Singapore, 1998. He is the co-editor of “Modelling and Control of Mechanisms and Robots,” World Scienti5c Publishing, Singapore, 1996, “Modelling and Control of Mechanical Systems,” Imperial College Press, London, 1997, “Theory and Practice of Control and Systems,” World Scienti5c Publishing, Singapore, 1998, “Modern Control Theory,” Esculapio, Roma, 1999.
doi:10.1016/j.automatica.2003.09.007
Industrial control systems design Michael J. Grimble; Wiley, New York, 2001, ISBN 0-471-49225-6 There are applications in practice, for which satisfactory control can be achieved without much eEort, only with an appreciation of the physical behavior of the process and the right hardware. However, there is often the case that, due to process complexity or to varying, demanding industrial requirements, advanced control systems are needed for eEective solutions. Published in 2001, Industrial Control Systems Design by Grimble illustrates the development and implementation of optimal techniques to a range of industrial applications. The book is structured in three parts: the 5rst part refers to Polynomial System Descriptions, the second one provides State Space and Frequency Response Descriptions and the last one comprises Industrial Applications. A summary of each chapter is given below. The 5rst chapter does not belong to any of the categories mentioned above, it is a stand-alone chapter meant to give a broad(er) description of the methodologies appearing in advanced industrial control. With this introductory chapter, the book has 13 chapters. As the main bene5t of advanced process control is increased economic yield of the plant, which can be often obtained by operating the process closer to the boundaries, issues such as robust control, fault tolerant control
and fault-monitoring become very important. These topics are covered in this 5rst chapter in short, concentrated descriptions, together with overviews of intelligent control techniques and systems integration. Starting with the second chapter the system is described in a polynomial form or frequency domain form. This approach oEers a quick overview of system properties in terms of pole-zero behavior, noise and disturbance rejection, and stability. Chapter 2 introduces the scalar solution of H2 -optimal control problem based on KucUera approach to frequency domain optimal control. The solution is derived 5rst for the standard unity feedback loop. Taking into account that feedforward control combined with conventional feedback can be a powerful control tool, feedforward action is included in the loop, but the way the problem was stated does not allow independent tuning of the two controllers. A more general feedforward optimal control problem is considered in the second part of this chapter, which provides freedom in tuning the feedback and feedforward controllers, and improved performance when disturbance models are not entirely known. Chapter 3 is devoted to predictive optimal control. Long-range predictive control methods were used successfully by many companies over the years, due to major bene5ts obtained by introducing optimization in the control loop. Among LRPC algorithms, the very well-known generalized predictive control (GPC) is presented and its simple solution is derived. However, GPC algorithms provide sometimes
Book reviews / Automatica 40 (2004) 337 – 342
suboptimal results due to inconsistency appearing in the theoretical aspects employed. Suboptimal results can be avoided if linear quadratic Gaussian predictive control (LQGPC) algorithm, described further on in this chapter, is used. This algorithm leads to a LQG type of solution in the same simple matrix representation as the usual GPC. In Chapter 4, the solution to the H2 optimal control is extended to cover the multivariable case. Practical aspects are considered and potential fault conditions are estimated, the LQG optimal control solution involving polynomial matrices and additional diophantine equations. The last part of the chapter investigates the possibility of deriving observer forms of estimators and controllers, investigation leading to dual solutions, depending on the order in which the control and the estimation problems are solved. The H∞ control synthesis problem is introduced in Chapter 5. All the solutions to diEerent H∞ problems stated in this chapter, are derived as special cases of weighted H2 optimization problems. First, the scalar H∞ feedback/feedforward control is presented, then SIMO optimal control is considered for both H2 and H∞ and, lastly, the H∞ equivalent of the LQGPC control law is discussed, the broad range of problems approached illustrating the Dexibility of the H∞ polynomial approach. H2 and H∞ estimation problems, alternatives to Kalman 5ltering, are solved in Chapter 6. H2 approach is to be used when stochastic problems are dominating, while H∞ approach provides better results when robustness is the main issue. This chapter introduces also a generalization of H2 and H∞ estimation problem that can be used to solve problems where no transfer function or polynomial-based solution is available. The 5rst part of the book, the widest one, ends with the sixth chapter. The second part begins with the presentation of H2 and H∞ control and 5ltering, in Chapter 7. A Wiener– Hopf approach is considered and a more generalized LQG problem, which involves transport delay elements on input and output channels is solved using the separation principle. Next, the relationship between H2 and H∞ is emphasized by showing that the H∞ optimal controller satis5es also a type of separation principle result, which allows the H∞ solution to be obtained as the solution of separate state-feedback optimal control and estimation problems. Chapter 8 introduces the state space approach to predictive optimal control, approach that provides more insight information on how the future reference or setpoint can be used. The GPC and LQGPC control laws, that are derived using a multi-step criterion, represent a 5rst alternative. A diEerent method of using the future reference signal, based on a single-step criterion, is described in the second part of the chapter. The resulting LQ or LQG controllers are easily calculated using a standard Riccati equation procedure. Although this procedure may result in a more complicated controller structure, there are clear advantages that motivate its usage. As the emphasis in the previous chapters was mainly on synthesis techniques, attention is focused in Chapter 9 on
341
frequency domain design, namely on quantitative feedback theory (QFT), approach that provides valuable insights into the robustness issues under the condition that classical design expertise is available. Combining the simplicity of optimal techniques with the insights and robustness of QFT solutions, a design philosophy is developed which can be used also for systems with sophisticated dynamics. The third part of the book presents practical applications of the theory issues elaborated in the previous two parts. Four industrial applications are considered. Chapter 10 is devoted to power generation and transmission, more specifically to the voltage control of the generators accomplished by automatic voltage regulators (AVRs). H2 and H∞ SIMO control problems in the polynomial approach are solved to design AVRs and additional output feedbacks are used to improve the stability. Chapter 11 discusses the design of control for metal processing. First an overview of the existing methodologies for multivariable hot strip mill control is provided, then a detailed description of problems that can occur is presented and 5nally MIMO state space H∞ optimal technique is applied to design the controller. The main bene5t over the classical techniques is the natural way the multivariable nature of the problem is treated. Marine control systems are introduced in Chapter 12. Two problems are considered: the 5n roll stabilization and the robust control of ship positioning systems. A SISO H∞ problem is formulated and solved in the polynomial approach described in Chapter 5, for each of the above-mentioned control problems. The derived solutions provide superior performance compared to the standard LQG Kalman method. In the last part of the chapter the solution of MIMO state space H∞ ship positioning control is proposed as a compromise between robustness and stochastic requirements. The last chapter of the book, Chapter 13, deals with the aero-engine and Dight control design. Two applications of H∞ robust multivariable control systems design in the state space approach are considered: the control of a twin spool reheated turbofan engine and the control of a generic canard-delta aircraft. In both cases the classical designs give reasonable results, the bene5ts of the H∞ design being that it copes with the multivariable nature of the process more directly and provides an automated design procedure. As the title suggests, the book is focused on illustrating advanced control methodologies on industrial control applications. The third part of the book is entirely devoted to this purpose and four practical control problems are presented in separate chapters. Each chapter contains a detailed description of the process and control objectives, together with an overview of the existing methodologies, proving the great deal of experience the author has with these industrial applications. The 5rst two parts present the theoretical background on which Part 3 relies. It can be noticed that a preference exists for the polynomial approach, the 5rst part being wider in presentation and content than the second one. Either in polynomial or state–space form, each chapter begins by
342
Book reviews / Automatica 40 (2004) 337 – 342
presenting a speci5c problem occurring in real-life process control, the theory developed further providing a solution to the stated problem. Not all theorem proves are fully derived, instead indications about the derivation or references are given, underlying the main goal of the book, that of providing a framework for solving practical control applications rather than giving rigorous theoretical demonstrations. Examples are also given throughout the text to 5ll the gap between theory and practice. Two surprising remarks can be made: (1) While the 5rst two parts contain mostly discrete-time formulations of problems and solutions, the industrial applications presented in the third part implement only continuous-time solutions. (2) Only H2 and H∞ techniques are used in the last four chapters, although theoretical background is established also for optimal predictive control, 5ltering, estimation and QFT. Generally speaking, the book cannot be used as a teaching reference or a textbook, since it is not self-contained, the doi: 10.1016/j.automatica.2003.09.006
basic notions are brieDy presented in the beginning of each chapter and then rapidly extended to generalized structures suited to deal with more complex problems appearing in practice. From this point of view, this book is a good reference for practitioners and researchers who are confronted with practical problems for which classical control techniques do not provide the most appropriate solutions. Mihaela Sbarciog Department of Electrical Energy; Systems & Automation (EeSA); UGent—Ghent University; Technologiepark 913, Ghent 9052; Belgium E-mail address:
[email protected] About the Reviewer Mihaela Sbarciog was born in Romania in 1976. In 1999, she received her engineer degree in Control Engineering and Industrial Informatics and in 2000 her master degree in Arti5cial Intelligence Systems in Process Control from the Galati University, Romania. From February 2002, she is a research assistant at Ghent University. Her research interests include system control theory, adaptive control, predictive control of nonlinear systems, system identi5cation.
