基于协作协进化的多Agent协作机制的研究  

A study of multi-agent cooperation mechanism basedon cooperative co-evolution

作　　者：汤琼[1] 杨东勇[1] 

机构地区：[1]浙江工业大学信息工程学院,浙江杭州310032

出　　处：《浙江工业大学学报》2004年第5期526-530,577,共6页Journal of Zhejiang University of Technology

基　　金：浙江省自然科学基金(601078)

摘　　要：协作问题是自主多智能体机器人系统研究的关键问题之一。基于多智能体机器人系统的CCP协作协议所生成的各Agent的任务序列依赖于目标的初始顺序,难以得到优化解。提出了一种利用协作协进化实现多Agent协作的机制,采用多种群协作生成多Agent系统的任务执行序列,在给定的任务分解所产生的所有可行解中寻找优化解,从而达到优化系统的性能的目的。利用该机制,对3个智能体协作搬运8个物体进行计算机模拟,结果表明,该方法在优化任务执行序列方面作用明显,从而能有效提高多智能体机器人系统的协作性能。Cooperation is one of the key problems of the autonomous multi-agent robotic systems. The task sequences created by CCP of the multi-agent robotic systems are dependent on the initiated sequence of the objects, it is difficult to create the optimized solutions. In this paper, a cooperative mechanism is proposed, which realize the cooperation of multi-agent by cooperative co-evolution. The multiple population cooperation is used to create the task implementing sequences of the multi-agent system, and optimized solutions are searched from all feasible solutions based on task decompose. So that the performance of the system is optimized. Experiment that three agents move eight objects cooperatively has been done. Results show that the proposed cooperative mechanism can optimize the implementing sequence of the task evidently, and improve the performance of the system effectively.

关 键 词：多智能体机器人系统 多Agent系统 协作机制 任务分解 多种群 可行解 协作性 任务执行 优化解 关键问题 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] TP18[自动化与计算机技术—控制科学与工程]