Jump Linear Systems in Automatic Control

This book is a monograph on hybrid parameter processes that are characterized by the presence of a discrete parameter and continuous variables. The author considers stochastic models in which the future control trajectories and the present solution do not determine completely the future of the system. The special stochastic processes and systems treated by the author are characterized by random transitions between different regimes, and this randomness primarily occurs through its discrete parameters. The book consists of eight chapters and two appendices. The appendices present brief summaries of basic probability, random processes, optima1 filtering, stochastic stability, stochastic maximum principles, matrix maximum principles, and stochastic dynamic programming. Readers might find it useful to consult references on applied probability and Markov processes before reading the eight chapters of this book. The first chapter introduces the reader to hybrid dynamic models by means of examples from target tracking, manufacturing processes, solar thermal receivers, and fault-tolerant control systems. Chapter 2 examines the global controllability and relative and stochastic stability of hybrid parameter systems. Also included in Chapter 2 are the concepts of Liapunov function and Liapunov exponents, observability, and detectability. Chapter 3 considers control optimization, jump linear quadratic regulators derived from maximum principles and dynamic programming, asymptotic behavior of quadratic regulators, suboptima1 solutions, optima1 switching output feedback, and algorithms for the optimization and evaluation of regulators for jump quadratic systems. The robustness, costs and their distribution, bound costs, and minimax solutions of jump linear systems are treated in Chapter 4, while the jump linear quadratic Gaussian problem is analyzed in some detail with Karman filtering and Poisson impulsive disturbances in Chapter 5. Optimal filtering, Wiener-driven oscillations, filter performance, and point-process observations are considered in Chapter 6. Chapter 7 deals with control under regime uncertainty, stability, control optimization, and regime estimation filters. The final chapter, Chapter 8, considers extensions of hybrid systems, non-Markovian processes, wide-band hybrid models, and extensions of the jump linear systems presented in the previous seven chapters. The book contains many theorems and proofs, is well illustrated with examples, and covers the material in depth. It is relatively free of typographical errors except that pages 206 and 207 have been interchanged.