Robust nonlinear Model Predictive Control based on nominal predictions: a zonotopic approach
Model Predictive Control, Nonlinear Control, Robust Control, Constraint Tightening, Invariant Sets, Zonotopes
The main objective of this project is the development, analysis and simulation of new robust model predictive control algorithms for nonlinear systems in the presence of bounded additive disturbances. The proposed controllers satisfy recursive feasibility and stability criteria and are based on existing algorithms for the nominal or linear cases as starting points. Robust constraint satisfaction is reached through nominal predictions coupled with tightened constraints, with the mean-value zonotopic extension being used in order to reduce conservatism in the disturbance propagation.
The problems of regulation without offset in the presence of constant disturbances and tracking of piece-wise constant references were tackled, also considering stochastic disturbance and chance state constraints. The proposed techniques are applied to simulation Buck-Boost and CSTR (Continually Stirred Tank Reactor) benchmark case studies in order to validate and illustrate the proposed approaches.