基于博弈论及惩罚机制的多Agent协作控制算法

The Algorithm for Mulit-agents Cooperation Controling Based on Game Theory and Punishment Mechanism

针对协作过程中自利的Agent选择回报更高的任务去执行,导致当前任务不能正常执行的问题,提出一种基于博弈论及惩罚机制的协作控制方法.在形成协作时优先选择信誉值较高的Agent,在协作执行过程中利用惩罚机制来约束退出协作的Agent的行为.仿真结果表明,该算法能有效地避免Agent在协作中随意的退出,确保协作任务的顺利执行,提高协作成功率.

For the problem of currents task cannot be normally performed when selfish agents would select higher utility tasks during collaborative process,a cooperative controlling method based on game theory and punishment mechanism was proposed.In phase of cooperative teamforming,selected agents which have higher credibility,during taskexecuting,used punishment mechanism to constrain the behaviors of agent for exiting.Simulation results showed that the algorithm can effectively avoid agent arbitrary exiting collaboration,ensured the smooth implementation of collaborative tasks and improve the success rate of collaboration.