Hierarchically decomposing multi-agent system to accelerate group consensus

Cooperative control of multi-agent system has been extensively discussed for groups with single-layer topology. This paper is unfolded from the problem of consensus and discusses how to decompose a single-layer topology into layers of connected subgraphs to improve the cooperative performance of group. A hierarchical decomposition algorithm is developed to split the topology into hierarchy. On the premise that each agent can only interact with its neighbors restricted by the original topology of system, performing consensus within the hierarchical structure obtained by our algorithm will yield the fastest convergence when compared with running consensus on original single-layer topology or other hierarchical structures. This paper also presents a connection between topological structure of group and the global performance in cooperative control.