Realization of multi-input/multi-output switched linear systems from Markov parameters

This paper presents a four-stage algorithm for the realization of multi-input/multi-output (MIMO) switched linear systems (SLSs) from Markov parameters. In first stage, time-varying (LTV) that is topologically equivalent to true SLS derived parameters assuming discrete states have common MacMillan degree and mild condition on their dwell times holds. second stationary point set Hankel matrix with fixed dimensions built examined. Splitting this into disjoint intervals complements reveals time-invariant dynamics prevailing these intervals. Clustering over feature space permits recovery up similarity transformations which complete if unimodality assumption holds satisfy residence requirement. third switching sequence estimated by three schemes. The scheme non-iterative in time. based matching system segments. works also same principle, but it optimization/hypothesis testing algorithm. schemes operate different time model structure requirements, requirements are weaker than needed recover states. fourth state estimates brought basis novel transformation necessary predicting outputs prescribed inputs. Robustness amplitude bounded noise studied shown small perturbations may only produce deviations vanishing as diminishes. Time complexities stages studied. A numerical example illustrates results.