Tube-based anticipative model predictive control for linear parameter-varying systems

Currently available model predictive control methods for linear parameter-varying systems assume that the future behavior of the scheduling trajectory is unknown over the prediction horizon. In this paper, an anticipative tube MPC algorithm for polytopic linear parameter-varying systems under full state feedback is developed. In contrast to existing approaches, the method explicitly takes into account expected future variations in the scheduling variable: its current value is measured exactly, while the future values over the prediction horizon are assumed to belong to a sequence of sets describing expected deviations from a nominal trajectory. Through this mechanism, the controller “anticipates” upon future changes in the system dynamics. The algorithm constructs a tube homothetic to a terminal set and employs gain scheduled vertex control laws. A worst-case cost is minimized: the corresponding optimization problem is a single linear program with complexity linear in the prediction horizon. Numerical examples show the validity of the approach.