A variety of techniques have been devised over the years for system identification. In general the identification techniques are derived from the optimization and estimation theories. The purpose of this book, in contrast to most other books that include a multitude of methods in a single volume, is to focus on the least squares method as a basic solution to the system identification problem. Since the least squares method is a classical method frequently practiced among scientists in various fields, this book can appeal to a large audience. The other motivation for focus- ing on the least squares method is that other popular identification meth- ods, such as cross-correlation, maximum likelihood, Kalman filtering, instrumental variables and stochastic approximation, can be easily related to the least squares algorithms. Therefore, System Identification provides a basis of some degree of integration and unification of many system identification methodologies.
Throughout the book, only systems in open-loop configuration are considered. The basic results can be applied to identify closed-loop sys- tems. In terms of system models, the emphasis is placed on the input- Output characterizations (difference, equation and weighting sequence) rather than characterizations by state variables.
The mathematical treatment here is moderate so that the widest possi- ble group of scientists and engineers can participate as readers. However, selected references are also included to allow interested readers to pursue the theoretical developments further. Therefore, the book is suitable as a text for graduate students studying system engineering at the universities.