In practical control systems, the plant states are not always measurable, so state estimation becomes essential before the state feedback control is applied. In this paper, we consider output feedback model predictive control (MPC) for linear parameter varying (LPV) systems with input constraints. We proposed two approaches to obtain the observer gain, that is to compute the gain in the dynamic optimization at each time instant (on-line), and to compute the gain in advance (off-line), respectively. By applying both approaches, the state estimation error goes to zero asymptotically, meanwhile, the state feedback gain is optimized. In fact, the on-line approach can help enlarge the feasibility region and improve the control performance. It has been shown that feasibility of both approaches can be maintained for the closed-loop control systems even in the presence of state estimation error. Finally, the proposed output-feedback MPC strategies are applied to an angular positioning control system and the control of a transcritical CO2 vapor compression refrigeration system.
Actuator fault tolerance evaluation approach of nonlinear model predictive control systems using viability theory
