An Identification Procedure for Linear Continuous Time Systems with Jump Parameters

We formulate an identification procedure for multi-input/ multi-output linear continuous time systems with jump parameters in the absence of observation noise. We assume that the parameters of the system depend on a Markov chain whose matrix of transition rates is unknown. We propose discrete time parameter estimation methods for the continuous time system to be identified. This includes a procedure for estimating the statistics of the Markov chain driving the parameter jumps.