Network decomposition for distributed control through community detection in input–output bipartite graphs

This paper addresses the decomposition of network systems for distributed control. We construct a novel weighted input–output bipartite graph representation of control systems, in which the input–output edge weights capture topological connectivity and short-time response sensitivities. We then introduce community detection as a network-theoretic tool to generate a decomposition with strong intra-subsystem interactions and weak inter-subsystem interactions. A modularity-based graph bisection procedure is applied recursively to determine the optimal decomposition. The proposed method is applied to a chemical process network example.