Output feedback receding horizon control for constrained linear systems

To overcome the situations where we cannot access the full information of the plant states, advanced output feedback receding horizon control for constrained linear systems are studied in this thesis. We focus on three types of constrained linear systems: LTI systems without disturbances, LTI systems with energy-bounded disturbances and uncertain linear systems with magnitude-bounded disturbances. A stable observer introduced to estimate the plant states ensures that the errors between the original plant states and the estimated states will approach zero with time going on. Our studies are based on these estimated states instead of the original. The output feedback controllers are developed on the basis of measured output signals. The synthesis conditions are solved by LMI, including initial condition, stability condition and constrained input/output conditions.;For constrained LTI systems without disturbances, an offline output feedback controller is proposed by minimizing a quadratic performance of the controlled output signals. This work is extended to online infinite-horizon RHC by recalculate the same synthesis conditions at each time step. The resulted performance index is monotonically decreasing. The cost of online finite-horizon RHC consists of performance over finite horizon and a terminal cost. A terminal set determined from the offline output feedback control, is used as a terminal constraint in the finite-horizon RHC to ensure the states go into this region over finite steps. The finite-horizon RHC is solved by treating the predicted states and control forces as decision variables instead of controller parameters in the offline output feedback control and the infinite-horizon RHC.;For constrained LTI systems with energy-bounded disturbances, an offline robust output feedback controller is achieved by optimizing closed-loop H infinity performance index from disturbances to controlled output. An dissipation constraint is added in the corresponding robust output feedback infinite-horizon RHC to guarantee the moving horizon stability. The dissipation terms are defined by a recursive equation. In robust output feedback finite-horizon RHC, we allow the system has different controller parameters at each step. Then the synthesis conditions are calculated by choosing controller parameters as decision variables. Thus the finite-horizon RHC is less conservative than the infinite-horizon RHC and achieves a better performance index.;For constrained uncertain linear systems with magnitude-bounded disturbances, the systems matrices of the uncertain linear systems belongs to a convex hull. The synthesis conditions become a set of similar conditions in which the system parameters are chosen as vertex points of the convex hull. Our goal to design an offline robust output feedback controller for uncertain linear systems is to find a stability region such that once the plant states go into this region, they will not leave this region any more. Thus the offline robust output feedback control has a different stability condition instead of H infinity or H2 criteria. The corresponding online robust output feedback infinite-horizon RHC obtains a optimal stability region at each time step. Based on this stability region, an online robust output feedback finite-horizon RHC is proposed by assuming that the cost will be zero once the plant states go into the stability region, the cost performance index is a quadratic function of controlled output signals without terminal cost. The finite-horizon RHC algorithm have two procedures. The first is to solve the robust output feedback infinite-horizon RHC problem to achieve a stability region. The second is to compute the robust output feedback finite-horizon RHC problem to obtain control force.;All of proposed output feedback control algorithms are demonstrated by simulations of single microcantilever system, two-mass/sping/damper system and single, non-isothermal continuous stirred-tank reactor. All control algorithms obtain satisfied dynamic response and performance index.