Robust model predictive control for time-varying systems

This paper extends robust model predictive control to problems involving linear time-varying systems with constraints and subject to persistent, unknown but bounded disturbances. The new controller guarantees constraint satisfaction and feasibility of the optimizations. Robust feasibility is achieved by tightening the constraints in the online optimization, with the advantage that the complexity of the optimization is not changed. This method is applicable to the control of constrained nonlinear systems by linearizing about a reference trajectory. The uncertainty introduced by linearization error can be accommodated in the disturbance model, and constraint satisfaction for the nonlinear system is then guaranteed.