Observer-based Output Feedback Robust Model Predictive Control

Model predictive control(MPC),also known as receding horizon control,is a kind of control techniques that solves constrained optimal problems based on predicted system future behaviors.Compared with the traditional MPC,robust model predictive control(RMPC)can take system uncertainties and physical constraints into account in receding horizon optimal control problems and ensure the robust stability of a control system.RMPC has received much attention in international academia and industry.However,in many practical processes,system states are unknown,and only the system output information can be directly obtained.Therefore,output feedback RMPC is more closely related to practical engineering applications and has been widely studied in recent years.This thesis focuses on the research on the observer-based output feedback RMPC approaches for linear parameter-varying(LPV)systems with bounded disturbances and noises.The research expands the existing output feedback RMPC approaches for uncertain systems with parametric uncertainties,bounded disturbances and noises.The main contributions are summarized as the following aspects:1.For LPV systems with bounded disturbances and noises,a synthesis approach to an interval observer-based RMPC is investigated.Currently,the studies on interval observer systems mainly focus on the design of interval observer systems and interval observer-based robust control approaches.Nevertheless,most of the results about the robust control approaches based on interval observer systems have not considered physical constraints in controlled systems.In this thesis,for LPV systems with bounded disturbances and noises,an interval observer is designed to obtain the upper and lower bounds of true states,which ensures the cooperativity and stability of the estimation error systems.In the on-line interval observerbased RMPC optimization control problem,the interval observer gain guarantees that the nominal estimation error system converges to the origin.Then,the optimized controller gains ensure that the nominal closed-loop interval observer system converges to the origin.When bounded disturbances and noises are considered,the estimation error system and closed-loop observer system respectively bounded within time-varying robust positively invariant(RPI)sets and robust control invariant(RCI)sets are steered to a region in the neighborhood of the origin such that the robust stability of the controlled systems is guaranteed.2.For LPV systems with bounded disturbances and noises,an off-line approach to observer-based output feedback RMPC via the zonotope set-membership state estimation is investigated.The off-line approach reduces the on-line computational burden of the output feedback RMPC with polyhedral estimation error sets.In order to reduce the on-line computational burden,we proposes a method to move the on-line computation to off-line optimizations based on a look-up table.In the off-line stage,the look-up table is constructed to store offline optimized controller gains and the corresponding regions of attraction.In the on-line stage,the regions of attraction with the closest containment of real-time estimated states and the corresponding controller laws are searched in the look-up table according to the real-time estimated states and the bounds of the estimation error sets.