对抗性多智能体系统中的规划识别方案

A plan recognition approach for agent in adversarial multi-agent system

针对当前规划识别限于协作多智能体领域的问题,提出了一种适用于对抗环境下智能体的基于Bayesian网络模型的规划识别方案以识别其他智能体的动作目标.研究了机器人足球的几个典型规划,并利用机器人足球比赛仿真系统进行了测试,仿真结果表明该方案有效地增强了智能体间的协作,提高了球队的获胜率.

The existing plan recognition models are efficient only in collaborative field, so a plan recognition approach based on Bayesian belief network model is proposed to suit the adversarial multi-agent environment. Several typical rules of robotic soccer were studied and experiments were done by using the robot soccer simulation game. The simulation results show that the collaborations of the whole team and the ratio of winning game were both improved remarkably.