This paper proposes a new method for the identification of switched linear systems. The proposed method includes two main steps of mapping and clustering. At the first step, a mapping is developed from the space of input-output data into the parameter space. A lot of linear equation sets, composed of equal number of equations and unknowns, are solved in this part. At the next step, submodel parameters are derived by clustering the parameters obtained in previous step, into several groups. Since the clustering step is carried out in the parameter space, the proposed method makes no distinction between the identification of switched linear systems and piecewise linear systems and the identification of submodel parameters is done independently of the estimation of switching signal. Numerical examples show the effectiveness of the proposed method in the identification of switched linear systems.

Keywords: System identification, Switched linear systems, Piecewise affine systems, System of linear equations

Full-Text [PDF 330 kb]   (2932 Downloads)