MIMO-SLS Identification from Input-Output Data

In this paper, we propose a framework to iden-tify discrete-time, multi-input/multi-output (MIMO), switched-linear systems (SLSs) from input-output data. The key step is an observer-based transformation to a switched auto-regressive with exogenous input (SARX) model. This transformation has a nontrivial kernel complicating identification, but converts the state-space identification problem to SARX model identification. Issues resulting from this transformation are carefully addressed by considering equivalance classes.