Parameter-dependent stability conditions for quasi-LPV Model Predictive Control

This paper extends earlier work on nonlinear predictive control based on a representation of the nonlinear plant as quasi-LPV model. Since the scheduling parameters depend on state variables and inputs, they can be predicted. An efficient predictive scheme with guaranteed stability is proposed that involves solving a sequence of SOCP problems at each sampling period. Compared with previously reported work, the conservatism of the approach is reduced by allowing matrix variables in the stability conditions to be parameter dependent. The efficiency of the proposed method is illustrated with its application in simulation to a model of an arm-driven inverted pendulum. A comparison with other state-of-the-art NMPC methods is made to highlight the benefits of the proposed approach both in terms of closed-loop performance and computation time.