基于状态预测的多智能体动态协作算法

针对复杂动态环境下的多智能体协作问题,提出基于信息处理和状态预测的优化动态协作算法.充分考虑其它智能体对环境的影响,采用重要度函数和信息处理方法,对协作所需信息进行筛选和处理.通过引入状态预测算法,在多智能体动态协作过程中对智能体的行为和系统的状态进行预测,以实现协作结构的在线调整,使得多智能体能在内部以新的控制任务或新的平衡状态为目标,进行联合行动的动态协作.通过在典型的复杂动态MAS研究平台--机器人救援仿真比赛系统中应用,验证了该算法的有效性.

Dynamic Cooperation Algorithm in Multi-Agent System Based on State Prediction

A new technique was proposed for the multi-agent cooperation problem in a complicated dynamic environment. The technique,which was an optimized cooperation algorithm,was based on the information processing and state predicting. The proposed algorithm,which taken the effect of other agents into account fully,adopted importance function and information processing method to filter and process the acquired information. A state prediction method was used to predict the behavior of agents and the trend of the system states during the cooperation process. Consequently,the multi-agent dynamic cooperation model was established according with the new task and the new balanced state. This technique is successfully used in the RoboCupRescue simulation team and validates the generalization and validity of the cooperation algorithm.