摘要
模拟机器人足球比赛(Robot World Cup,RoboCup)作为多智能体系统的一个通用的实验平台,通过它可以来检验各种理论、算法和框架等,已经成为人工智能的研究热点。针对在复杂条件下的使用传统Q学习方法所产生的收敛速度缓慢和泛化能力不强的问题,文中使用人工化能力,缩短了学习的时间。并最终将其运用到仿真组比赛的Keepaway模型中,以此验证了该方法的有效性。
As a representative experimental platform of multi - agent system, RoboCup（Robot World Cup） by which various theories, algorithms and architectures can be evaluated, has become the research center of artificial intelligence. For the converge slowly and time consuming problems arised when using the classic Q- learning method in complicated environment, ttse ANN to represent the Q net and the batch Q learning to process the training data gathered from the environment. By these tactics, improved the generalization capability of the system, and decreased the time cost to learn. It was applied to the experiment of the Keepaway models in the simulation team whose result shows the validity of the method.
出处
《计算机技术与发展》
2009年第7期98-101,共4页
Computer Technology and Development
基金
安徽省自然科学基金(050420204)
安徽省高校拔尖人才基金(05025102)
安徽省自然科学研究项目(2006KJ098B)
