Research On The Consensus Of Multi-Agent Systems

In recent years, lots of scientists in diverse disciplines have been attracted by the collective behaviors of multi-agents system. Dynamic network is an effective method of investigating collective behavior which composed by multi-agents. From nature to human society, multi-agents collaboration behavior have been seen everywhere. It is necessary to guarantee the agents to keep and maintain the consistency of information when working in the unpredictable and complex environment. Therefore, consistency is the core issue in the mutli-agent network collaborative research, and consistency of multi-agent network has become an important and challenging frontier issue.Vicsek model which represented a typical model in consistency of multi-agent network research, just using a few simple rules to control the multi-agent that could make a complex emergence phenomenon of the whole multi-agent network. The research of Vicsek model, not only provided theoretical basis for multi-agent cooperation, but also guided the light of multi-agent practical application. By researching the consistency of multi-agent network, on the one hand helped us fully understand the biological intelligence coordination to a consensus process, on the other hand enable us to design the real multi-agent network to complete the desired consistency behavior. Such as unmanned aircraft control, mine search implementation and network congestion control.Under the above background, in this paper, we investigate the consistency of the multi-agent network in the aspect of flocking and rendezvous based the connectivity of multi-agent network . In the overview of the relevant literature, the main work is as follows: Firstly, we investigate some classical models of consistency of multi-agent network, such as Boid model, Vicsek model, the algorithm of location assemble and other basic consistency of multi-agent, and give the algebraic graph theory and matrix theory tools to solve the issue of consistency.Secondly, the influence of the connectivity of the multi-agent network topology to consistence has been studied. Based on the Vicsek model which proposed multi-agent owned the same move rate and different directions moving in the plane, in the situation of ignoring the external interference such as noise, we analyzed the influence of the connectivity of multi- agent network. A special fixed topology networks and dynamic topology network have been studied in this part. On the special fixed topology networks, we gave the original parameter that allows the topology of network to maintain connectivity, ultimately making the agent in the network to reach direction consensus. On dynamic topology networks, we present a perception radius of adaptive variable strategy, making agents to maintain connectivity under the condition of arbitrary initial distribution to ensure the realization of the convergence of the entire network. We simulate this strategy through NetLogo dynamic simulation platform, and verify the feasibility and effectiveness of the proposed strategy .Thirdly, we study the convergence of efficiency of multi-agent network. For accelerating the convergence efficiency of Vicsek model, a new direction consensus rule is presented through studying of the topological structure of the dynamic network. The new improved rule, each individual exploits the median of the maximum deviation direction, replaces the average direction vector updated rule in original Vicsek model. Considering the aspect of connected topology to proved the stability of the multi-agent network under the new improved rule control. When the multi-agent network content the initial location of the distribution topology is connectivity, based on NetLogo simulation platform the results under the new rules model , Vicsek model and linear Vicsek model simulation were compared. Simulation results show that the new improved rule is better than the original Vicsek model in convergence efficiency.Finally, we study the location consistency problem of multi-agent network based on the attraction repulsion functions which proposed by Gazi. Compare with the Gazi's model, a new position consistency model based on the furthest neighbour clustering model is proposed, and we find that our new model can overcome the defect of global interaction , and when there is some weak connectivity it still can achieve the desired results. We also carried out comparing the results by the platform of Matlab simulation, and verify the feasible of the proposed new strategy.