Active mode estimation via clustering algorithm for switched linear systems

The work presented in this paper deals with the active mode identification problem for switched linear systems based on a measurement data set. This problem is an issue closely related to the classification problem of input-output measure data. Indeed, each data group associated with their most appropriate sub-model presents a mode (discrete state) of operation. Therefore, we propose a method for discrete state estimation based on a clustering algorithm combined with a decision mechanism. The clustering algorithm provides the class centers which will be exploited by the decision mechanism in order to identify the discrete state. Simulation results are presented to illustrate the performance of the proposed method.