Multiple low rank matrix approach to switched autoregressive exogenous system identification

This paper deals with the identification problem of switched autoregressive exogenous (SARX) systems with unknown model orders, which is a problem of estimating simultaneously multiple linear subspaces and their corresponding non-overlapping regions. This problem is important and has lots of applications because it can describe approximations of a wide variety of nonlinear systems. In this paper, the SARX identification problem is formulated as a kind of matrix rank minimization problem, and a practical algorithm is proposed based on the iterative partial matrix shrinkage (IPMS) algorithm. Numerical examples show that the proposed algorithm identifies SARX systems effectively.