基于规划识别的多智能体协作研究及应用

Study on Plan Recognition Based Multi-Agent Collaboration

多智能体系统是规划识别的一个有效应用平台,提出一种基于规划识别多智能体协作算法,对对抗环境和非对抗环境中的基于规划识别的多智能体协作算法进行了分析,实现了对队友和对手行为目的的认识和建模,减少了协作主体间需要通信的时间及难度。该协作算法应用到多智能体的有效测试平台机器人足球赛中,试验结果证明,该算法在通信受限、信息受限或信息延时的系统中可有效预测队友和对手的行为,从而实现智能体间的协作。

Multi-Agent system is an useful application flat for plan recognition. CMBPR（an multi-Agent collaboration algorithm based on plan recognition） is proposed in this article,the coordination algorithm based on plan recognition in adversarial and non-adversarial environment is analyzed,which recognizes and models the behavior of teammates or opponents,then the communication time and difficulty are reduced. This coordination algorithm is applied in Robocup that is the efficient test bed of multi-Agent system,the simulation results indicate this algorithm can predict the behavior of teammates or opponents effectively in local communication,local information or delay information system,then the multi-Agent collaboration is implemented.