Research On Cooperative Evolution Of Multi-agent In Complex Networks

Multi-sensor information fusion makes collaborative work more and more complex.Multi-agent can coordinate with each other to complete different tasks,which has got much attention from many researchers.Cooperation among agents is the core to ensure that they can work together.Cooperation can not only improve the performance of intelligent systems formed by agents,but also enhance their ability to solve practical problems.Therefore,it is an inevitable requirement for developing multi-agent systems to study the cooperation between agents.In recent years,It has produced profound influence for agent cooperation study through the game theory experiments.For instance,The Max Planck Institute Millin cooperation game theory experiment,the university of Zurich and university of St.Gallen in Switzerland on public goods game experiment,and Rand D G the repeated prisoner's dilemma of experiment and so on.These studies have certain contribution to evolutionary dynamics,cooperative evolution,and resource allocation within groups.This paper simulates and analyses the cooperative evolution process among multi-agents based on the above research on the evolution of multi-agent cooperation,which takes the game dynamics of complex networks as the core,introduces the dynamic game mechanism in the complex networks,and connects the intelligent individuals in the way of three-dimensional connection.The main work of this paper is as follows:1.Firstly,the effect of behavioral consistency in multi-agent on the cooperative evolution of prisoner's dilemma game is studied.The research shows that introducing the behavior consistency mechanism into the intelligent group can reduce the variation range of the group cooperation rate and make the multi-agent have higher cooperation density in the cooperation evolution.2.Because we need to consider the influence of reputation mechanism on the evolution of cooperation,we need to simulate and analyze the public goods game experiment with a certain degree of punishment and reward on the network.Research shows that moral constraints can promote group cohesion and restrain betrayal in intelligent groups.And in the experiment on Snowdraft Game in BA Network,considering the strong reciprocal behavior among groups when the number of cooperating individuals is unchanged during initialization,it can promote the cooperating to a certain extent between the intelligent agents who participate Snowdraft Game.In the follow-up study,Cooperative Evolution of Multi-Agents influenced by the behavior of punishing high reputation individuals is studied.3.In the simulation experiment of prisoner's dilemma game,the influence of strong reciprocity mechanism and behavior consistency mechanism on the evolution of group cooperation was analyzed.In BA scale-free community network,when there is a behavior consistency mechanism,Multi-Agents produce a certain degree of cooperation.Under the influence of strong reciprocal individuals,evolutionary game can have higher cooperation density when it reaches evolutionary equilibrium.However,higher cooperation density needs a certain gain of individual expenditure to maintain.All of the above are based on complex networks to connect intelligent individuals in a three-dimensional way.In addition,game evolution simulation experiments were carried out among intelligent groups,and the internal influence of strong reciprocity mechanism and behavior consistency mechanism on multi-agent cooperation was studied by analyzing the simulated data.It has certain significance for the development of multi-agent system in the future.