Subspace Methods for System Identification

An in-depth introduction to subspace methods for system identification in discrete-time linear systems thoroughly augmented with advanced and novel results, this text is structured into three parts.

Part I deals with the mathematical preliminaries: numerical linear algebra; system theory; stochastic processes; and Kalman filtering. Part II explains realization theory as applied to subspace identification. Stochastic realization results based on spectral factorization and Riccati equations, and on canonical correlation analysis for stationary processes are included. Part III demonstrates the closed-loop application of subspace identification methods.

Subspace Methods for System Identification is an excellent reference for researchers and a useful text for tutors and graduate students involved in control and signal processing courses. It can be used for self-study and will be of interest to applied scientists or engineers wishing to use advanced methods in modeling and identification of complex systems.

The introductory material combined with novel research results and proofs will make this an all-in-one reference of subspace identification methods for control and signal processing researchers Develops new results in stochastic realization theory incorporating the presence of exogenous inputs Chapter-by-chapter examples, exercises and solutions will guide the graduate student through learning basic and abstruse aspects of subspace methods and system identification Tutors and supervisors are provided with ready checks on their students' progress with exercise solutions and pre-written Matlab® programs reducing preparation effort to a minimum Includes supplementary material: sn.pub/extras

Auteur

Tohru Katayama received B.E., M.E. and Ph.D. degrees in applied mathematics and physics, from Kyoto University, in 1964, 1966 and 1969, respectively. Since 1986, he has been Professor at the Department of Applied Mathematics and Physics, Kyoto University, and had visiting positions at UCLA and the University of Padova. His main research interests include statistical estimation theory, Kalman filtering, spectral factorization, stochastic realization, system identification, and modeling and control of industrial processes, in which areas he has published over 100 papers, six books in Japanese, and edited a book on control and signal processing.

Professor Katayama has been an Associate Editor of IEEE Transactions on Automatic Control from 1996 to 1998, and a Subject Editor of Journal of Nonlinear and Robust Control for the last 10 years. He is a Fellow of the Society of Instrumentation and Control Engineers, Japan, is a past Chair of the IFAC Technical Committee on Stochastic Systems and is now the Chair of the IFAC Coordinating Committee on Systems and Signals for 2002-2005.

Texte du rabat

System identification provides methods for the sensible approximation of real systems using a model set based on experimental input and output data. Tohru Katayama sets out an in-depth introduction to subspace methods for system identification in discrete-time linear systems thoroughly augmented with advanced and novel results. The text is structured into three parts.

First, the mathematical preliminaries are dealt with: numerical linear algebra; system theory; stochastic processes; and Kalman filtering. The second part explains realization theory, particularly that based on the decomposition of Hankel matrices, as it is applied to subspace identification methods. Two stochastic realization results are included, one based on spectral factorization and Riccati equations, the other on canonical correlation analysis (CCA) for stationary processes. Part III uses the development of stochastic realization results, in the presence of exogenous inputs, to demonstrate the closed-loop application of subspace identification methods CCA and ORT (based on orthogonal decomposition).

The addition of tutorial problems with solutions and Matlab® programs which demonstrate various aspects of the methods propounded to introductory and research material makes Subspace Methods for System Identification not only an excellent reference for researchers but also a very useful text for tutors and graduate students involved with courses in control and signal processing. The book can be used for self-study and will be of much interest to the applied scientist or engineer wishing to use advanced methods in modeling and identification of complex systems.

Contenu
Preliminaries.- Linear Algebra and Preliminaries.- Discrete-Time Linear Systems.- Stochastic Processes.- Kalman Filter.- Realization Theory.- Realization of Deterministic Systems.- Stochastic Realization Theory (1).- Stochastic Realization Theory (2).- Subspace Identification.- Subspace Identification (1) ORT.- Subspace Identification (2) CCA.- Identification of Closed-loop System.