Matrix-wise approach for identification of multi-mode Switched ARX models with noise

In this paper we propose a method to improve the algebraic geometric solution for identification of Switched ARX models. Despite the advantages of the AG solution, it fails to deliver promising results when the noise level is noticeable. Therefore the proposed method aims to improve the SARX parameter estimation results. In this method matrix differential calculus and Kronecker product are used to form a nonlinear algebraic equation for calculating the parameters of SARX model. This algebraic equation will be solved by combining Newton iterative method and particle swarm optimization algorithm. Simulation results will illustrate the improvements achieved by the proposed approach.