A Connectivity Reduction Strategy for Multi-agent Systems

This paper considers the connectivity reduction of multi-agent systems which are represented with directed graphs. A simple distributed algorithm is presented for each agent to independently remove some of its incoming links based on only the local information of its neighbors. The algorithm results in an information graph with sparser connections. The goal is to reduce computational effort associated with communication while still maintaining overall system performance. The main contribution of this paper is a distributed algorithm that can, under certain conditions, find and remove redundant edges in a directed graph using only local information.