Continuous-time Approaches to System Identification

Abstract

At the outset this paper discusses the relevance and importance of continuous-time approaches in systems and control in general and in problems of identification in particular. The various continuous-time approaches to system identification are outlined by drawing attention to some earlier surveys, books and monographs and by highlighting certain recent approaches. Some indirect approaches via nonparametric models and via discrete time techniques are discussed along with some direct approaches. A class of system reference adaptive model (SRAM) methods influenced by and arising out of stable adaptive control techniques which are based on output error (OE) minimization are also outlined. Certain unifying features among some seemingly different methods are exposed. Some of the techniques outlined are then illustrated with the aid of examples and a comparison is made among them. Extensions to MIMO models, and to special problems of identification involving unknown time-varying elements, time delays and nonlinear elements within the broad framework of some of the techniques are then outlined. The paper concludes by discussing possibilities of inclusion of some of the techniques in standard computer program packages for general and real-time applications.