Dynamic Output Disturbance Models for Robust Constraint Satisfaction

Abstract

Robust control methods such as tube-based robust model predictive control (MPC) schemes, developed to provide robust constraint satisfaction guarantees, require an uncertain model of the controlled plant. In this paper, we present a method to identify such models, along with a robust MPC scheme with reduced conservativeness tailored to employ them. We consider input-output models in which uncertainty is modeled as an additive disturbance on the output. Reduction of conservativeness is achieved by identifying the dynamics generating the disturbance. Standard linear system identification methods are used in the model development procedure, with residuals from the identification process extracted to characterize uncertainty in a set-membership setting. The effectiveness of a using dynamic output disturbance models is demonstrated through simulations.