Subspace-based continuous-time identification

Abstract

The last few years have witnessed a strong interest in system identification using realisation-based algorithms. The use of Markov parameters as suggested by Ho and Kalman [18] Akaike [1], and Kung [28], of a system can be effectively applied to the problem of state-space identification; see Verhaegen et al. [43, 44], van Overschee and de Moor [41], Juang and Pappa [26], Moonen et al. [36], Bayard [3, 4, 33, 34]. Suitable background for the discrete-time theory supporting stochastic subspace model identification is to be found in [1,14,41]. As for model structures and realisation theory, see the important contributions [12, 31]. As these subspace-model identification algorithms deal with the case of fitting a discrete-time model, it remains as an open problem how to extend these methods for continuous-time (CT) systems. A great deal of modelling in natural sciences and technology is made by means of continuous-time models and such models require suitable methods of system identification [19]. To this end, a theoretical framework of continuous-time identification and statistical model validation is needed. In particular, as experimental data are usually provided as time series, it is relevant to provide continuous-time theory and algorithms that permit application to discrete-time data.