New algorithm for identification of discrete-time switched linear systems

This paper addresses the problem of the identification of discrete-time switched linear MIMO state space systems. The methodology proposed explores the use of subspace identification techniques, clustering and classification of data to estimate the quantity, a matrix representation of the sub-models and an estimate of the instants of switching of the system. An important feature of the algorithm is that the matrices of the different sub-models can be directly combined because they are obtained for the same base states. The efficiency of the developed method is illustrated using a numerical simulation.