Stochastic output-feedback model predictive control for systems with multiplicative and additive uncertainty

This paper designs output-feedback model predictive control for discrete-time stochastic linear systems. An augmented system of both estimated state and estimation error is formulated. A series of polyhedrons is designed for uncertainty propagation of the augmented system and handling probabilistic constraints. In particular, the polyhedron of estimation error is maintained over the repeated optimization such that the real system state is properly estimated. The resultant optimization problem of the controller is quadratic programming and is solved efficiently. A numerical example is given to illustrate performance of the controller.