Coalitional Model Predictive Control Scheme: Constraint Satisfaction and Application

This work presents a novel coalitional model predictive control scheme for linear systems composed of disturbed and constrained subsystems. These subsystems are dynamically interconnected and controlled by a group of local model predictive controllers or agents connected by a network. A supervisor periodically activates or deactivates the network links to achieve the best balance between control performance and communication cost. Each coalition, which is a group of agents connected by activated links, then employs a decentralized model predictive control strategy to coordinate their control actions. To guarantee the satisfaction of the system’s constraints, we extend the concept of a control invariant set for non-switched constraint systems to switched constraint systems, referred to as a ”common control invariant set.” This set is adopted as the terminal constraint of the collective coalitional model predictive control optimization problem. A three-reaches irrigation canal system is considered for validating the performance of the proposed control strategy.