Receding horizon control of Markov jump linear systems

Deals with the receding horizon control problem for Markov jump linear systems (MJLS) with associated quadratic costs. In terms of the information available to the controller, two situations are examined: the complete system state is accessible, or only the state of the linear plane is accessible and no information is available on the underlying Markov chain, apart from an initial distribution. Necessary and sufficient conditions of optimality are given for both cases. For the case when the Markov state is not accessible, we restrict the class of controls to linear feedback gains, and obtain the optimal control sequence in a computable form. Closed-loop stability associated with the first element in the control sequence is also addressed