Robustness Issues Related to the Application of Distributed Model Predictive Control Strategies

Abstract In this paper, a methodology is proposed to address robustness aspects related to the application of distributed model predictive control. Two problems are studied: the decomposition problem and the coordination problem in the presence of model errors. Three different MPC strategies are considered: centralized, fully decentralized, and Nash equilibrium based MPC. The methodology requires the computation of closed-loop system's variability via the solution of generalized eigenvalue problem which is formulated as a finite set of linear matrix inequalities. To select the best model decomposition or control strategy based on robust performance, the worst variability for each candidate is minimized by manipulating the input weights of the controller. Two case studies are presented to illustrate the application of the methodology.