Effect of network structure to the convergence rate of agents in multi-agent systems

Abstract

In this paper, we probe into the influence of network structure on the convergence rate of agents in multi-agent systems, especially for the position and number of nodes of network on it. We discover some results of which kinds of networks are conductive to the faster convergence rate of agents in multi-agent systems. Firstly, if there is no ring in different networks where the number of nodes are same, then the network with fork nodes which are connected to the central position node is in favor of the faster convergence of agents; otherwise, the convergence rate of agents is quicker in the network without fork nodes. Secondly, it is inevitable for agents to converge more ane more slowly with the increasing of node number. One way to improve the convergence rate of agents is to enable the increased nodes to connect to the central position node. Finally, when both the position and the number of nodes of different networks are same, it can save communication cost to remove some edges in network under the condition that the convergence rate of agents is affected little. Therefore, we reveal that if the edges which do not connect to the central position directly are removed, then their impact on the convergence rate is little.