A parallel distributed supervision strategy for multi-agent networked systems

This paper introduces a novel parallel distributed supervision strategy for interconnected linear systems subject to pointwise-in-time set-membership coordination constraints. The scheme is based on the Feed Forward Command Governor paradigm. It is assumed that the systems are locally regulated, possibly dynamically coupled, and connected via data links. The supervisor is in charge of enforcing coordination constraints by modifying, whenever necessary, the nominal reference signals so that the global coordination goal is achieved. The proposed method extends in a substantial way the sequential Distributed Command Governor previously developed by the authors by allowing each agent to modify its local reference in parallel with the others. The main idea consists in bounding each subsystem with a set of local artificial constraints able to ensure global constraints satisfaction. These virtual constraints are dynamically generated in a distributed way on the basis of the common information available to all agents. The proposed strategy is described and its properties analyzed. The performance of the obtained scheme is compared with previous solutions in the final example.