Identification of Switched MIMO ARX models

The paper deals with the identification of a MIMO Switched ARX model from its input-output data. The model is assumed to have an unknown number of submodels of unknown and possibly different orders. This is a challenging problem because of the strong coupling between the unknown discrete state and the unknown model parameters. In our work, we adopt a polynomial decoupling representation for handling switched systems with multiple inputs and multiple outputs. This exact and analytical polynomial representation however comes with an important complexity related to the number of polynomials that need to be estimated. Therefore, an alternative scheme is proposed that operates in two phases. We first classify the data according to the generating submodels and subsequently recover the system parameters.