摘要
提出了一种基于改进学习分类器的多机器人强化学习方法。增强学习使机器人能发现一组用于指导其强化学习行为的规则。遗传算法则在现有的规则中淘汰掉较差的,并利用较优的种群规则产生出新的学习规则。规则合并能提高多机器人的并行强化学习效率,使多个机器人自主地学习到相互协作的最优策略。算法的分析和仿真表明,将改进的学习分类器用于多机器人的强化学习是有效的。
This paper proposes a multi-robots reinforcement learning method based on improved learning classifier system.The enhanced learning enables robots to discover a group rules for guiding their reinforcement leaning behavior.Genetic algorithm could eliminate worse ones in the existing rules and produce new learning rules with the superior population rules.The merged rules can increase multi-robots＇ learning efficiency in parallel,thus the multi-robots could learn to collaborate with the best strategy.The algorithm analysis and the simulation indicate that the improved learning classifier system used in the multi-robot reinforcement learning is feasible and effective.
出处
《通信技术》
2010年第4期220-222,共3页
Communications Technology
基金
国家自然科学基金资助项目(批准号:60705020)
面向移动机器人环境感知的主动学习研究
关键词
强化学习
多机器人
改进学习分类器
遗传算法
reinforcement learning
Multi-robot
improved learning classifier system
genetic algorithm
