The Research Of Multi-agent Task For Forwarding Task Based On Evolutionary Game Theory

With the development of Multi-Agent in terms of emergency relief, assigningtasks issue is receiving more and more attention from researchers. Since Agent mayperform tasks in a relatively harsh environment, it is critical for them to achievecooperation in the intelligence system.Aiming at the decision of the Agent forwarding task that influencing other agentto get task, we hope each agent forwarding tasks they received as far as possible inorder to share resources. Considering the moment agent forwarding task with itslimited resources, Multi-Agent would not forward the tasks they received. And thenthey failed to send task to their neighboring agents accurately. So the reflectionmechanism is imported into the Multi-Agent forwarding task model based on themethod of probability and random process. The model is also based on the traditionalreplicator dynamic model of evolutionary game theory that reflects individualstrategy of Multi-Agent adjustment in the process of evolution. In order to select themost qualified candidates of the Multi-Agent for this task, we assume whether theMulti-Agent accept the task, they have two action including forwarding the task ornot forwarding. Based on the dynamics of evolutionary game analysis, it was derivedto Multi-Agent system forwarding task evolutionary stable condition. Experimentsshow the analysis the factors of the reward system and the communication on theevolutionary stable state. We assume that regardless of whether or not to do the task, agent could forwardthe task or not. Take the evolutionary game dynamics as the basis for analysis,deduced making Multi-Agent system to forward the task evolutionarily stablecondition and the sum of the maximum income for a period of time for each Agent.Experiments verify the derivation of the conclusion and the influence of confoundingfactors such as incentive systems and communications for the evolution of a stablestate.The article summarized as follows:1. Considering the agent’s own selfishness and affected communicationsenvironment, the article gives the grounds of contradictions with game theory andother Agent forwarding the task. Then the text educes research background,significance and contents;2. By introducing Agent reflection mechanism, incentive mechanism, and basedon the probability and stochastic process approach to construct forwarding agent taskevolution model that replicated the traditional evolutionary game dynamics model.Finally give the theorem that the process of evolution reflects the individual agentadjustment strategies;3. Draw the conclusions and make the analysis of experimental results.