Consensus on hierarchically decomposed topology to accelerate convergence and to improve delay robustness

Understanding the information flow structure is one of fundamental problems in multi-agent system, because it implies the mode and degree of information sharing among members, and thus affects the cooperative performance of group. Unfolded from consensus problem, this paper discusses the performance of multi-agent system from the view of topological optimization. A topology decomposition algorithm is developed to split the original single-layer topology into multi-layer structures. Consider the following constraints that 1) each layer establishes the connection with other layers via a sole leader in the layer, which guarantees the execution of hierarchical consensus; 2) each agent can at most interact with its neighbors restricted by the original topology of system. Performing consensus within the hierarchical structure obtained by our proposed algorithm will yield not only a faster convergence but also a better cooperative robustness with respective to delay than consensus on the original topology or other hierarchical structures.

1 Introduction

In the last decades cooperative control of agent networks has fascinated more and more researchers. It is the common subject on how to understand the cooperative motion patterns and the resulting performance of a large group of agents. One may note that the network connectivity ensures information sharing via multi-hop communication paths between members [1]. Furthermore, the structure of information flow topology implies the mode and degree of information sharing among members, and thus affects the cooperative performance of group.