Identification of switched linear systems using self-adaptive SVR algorithm

摘要

We consider the problem of switched linear system identification from input-output data set. This set may be a mixte set whose data are generated from a different switching affine subsystems so that one does not know a priori or a switching dynamics is unavailable. To overcome this main challenge, we develop an identification approach which consists in determining simultaneously a linear regression function which represents each submodel and a switching signal estimation via a self-adaptive clustering algorithm. The regression function is identified based on the Support Regression Vector (SVR) approach. However, the switching signal is provided by an unsupervised classification algorithm with self-adaptive capacities.