A Distributed Algorithm Based on Local Centrality for Dynamic Social Network Re-construction in Multi-Agent Systems

In this paper, we focus on a self-adaptive network reconstruction problem in a distributed multi-agent system. In this system, the relationship among agents forms a complex network, in which each agent plays two roles, task execution and communication. As agents move, the topology of the relationship changes, which could reduce the system's performance. To guarantee the performance, we design a local centrality for evaluating the vitality of nodes in the network, and propose a distributed re-construction algorithm (DRA) based on the local centrality for re-constructing the dynamic network. To test the efficiency of DRA, we integrate DRA into our former proposed dynamic task allocation algorithm to form a new dynamic task allocation algorithm. The experiment shows that our new improved algorithm performs more efficiently when compared to other dynamic task allocation algorithms. The results illustrate the efficiency of reconstructed dynamic complex networks, which also indicates that the local centrality preforms efficient. Finally, we replace the local centrality with degree centrality and LocalRank in DRA and execute a series of experiments for performance evaluation. The results demonstrate the high efficiency of LC.