1: \begin{abstract}%

2: In this paper, we investigate the problem of system identification for autonomous switched linear systems with complete state observations.

3: We propose switched least squares method for the identification for switched linear systems, show that this method is strongly consistent, and derive  data-dependent and data-independent rates of convergence. In particular, our data-dependent rate of convergence shows that, almost surely, the system identification error is $\OO\big(\sqrt{\log(T)/T} \big)$  where $T$ is the time horizon. These results show that our method for switched linear systems has the same rate of convergence as least squares method for non-switched linear systems. We compare our results with those in the literature. We present numerical examples to illustrate the performance of the proposed system identification method.

4: \end{abstract}

5: