A class of subspace model identification algorithms to identify periodically and arbitrarily time-varying systems

Abstract

Subspace model identification algorithms that allow the identification of a linear, time-varying (LTV) state space model from an ensemble set of input-output measurements are presented in this paper. Each pair of input and output sequences in this ensemble is recorded when the underlying system to be identified undergoes the same time-varying behavior. The algorithms operate directly on the available ensemble of input-output data and are a generalization of the recently proposed Multivariable Output Error State sPace (MOESP) class of algorithms to this ensemble type of identification problems. A special case is considered in this paper, where the repetition of this time-varying behavior is intrinsic, namely in periodically time-varying systems. An example of identifying a multirate sampled data system from a recorded input and output sequence demonstrates some of the capabilities of the presented subspace model identification algorithms.