Output-Feedback Robust MPC for a class of Nonlinear system

Abstract

This paper compares the performance of a robust model predictive control (MPC) with a nominal MPC. The robust scheme is applied to a nonlinear system with input constraints, bounded unknown additive disturbances, and uncertainties on the system states. The design of the robust controller relies on the linearized version of the nonlinear model along its trajectory. It comprises two components, one of which is obtained by a continuous-time MPC approach with tightened constraints for stabilizing the nominal system. The other one is chosen as standard LQR gain, ensuring that the error system trajectories lie in a particular invariant set. The controller is implemented in a digital environment and utilizes a continuous-time model and equations, vastly reducing the computational burden. Furthermore, with some plausible assumptions, the closed-loop system is demonstrated to have regional exponential stability around the origin. Simulation results verify the improvement in disturbance rejection property of the employed controller over nominal MPC.