Mobile version of Hanspub

文章引用说明 更多>> (返回到该文章)

Holland, J.H. (1986) A mathematical frame work for studying learning in classifier systems. In: Farmer, D., Lapedes, A., Packard, N. and Wendroff, B., Eds., Evolution, Games and Learning, North-Holland, Amsterdam, 307-317.

被以下文章引用:

  • 标题: 狭隘环境下一种多机器人路径规划方法A Multi-Robot Path Planning Method under Narrow Environments

    作者: 邵杰, 于景茹

    关键字: 路径规划, 多机器人, 学习分类器, 遗传算法, Q学习Path Planning, Multi-Robot, Learning Classifier System, Genetic Algorithm, Q Learning

    期刊名称: 《Artificial Intelligence and Robotics Research》, Vol.4 No.2, 2015-05-29

    摘要: 狭隘环境下多机器人路径规划使用共享资源时,极易产生冲突,优先顺序化是解决共享资源冲突的一个重要技术。本文提出了一种基于学习分类器的动态分配优先权的方法,提高机器人团队的性能。首先机器人通过XCS优化各自的行为,然后引入和训练高水平的机器人管理者来分配优先权解决冲突。本方法适用于部分可知的Markov环境,仿真实验结果表明本文所提方法用于解决多机器人的路径规划冲突是有效的,提高了多机器人系统解决路径规划冲突的能力。 Under narrow environments, conflict easily occurs when multi-robot path planning uses shared resources, and prioritisation is an important technology to solve this problem. This paper pre- sents a dynamic allocation priority method based on learning classifier to improve the perfor-mance of the robot team. Firstly robots optimize their behaviors by XCS, and then high-level robot managers are introduced and trained to resolve conflicts by assigning priority. The novel approach is designed for partially observable Markov decision process environments. Simulation results show that the method presented is effective to solve the conflict in multi-robot path planning and improves the capacity of multi-robot path planning.