摘要
文章在简单概述强化学习理论的基础上,对强化学习在实际机器人应用中经常遇到的连续状态-动作空间、信度分配、探索和利用的平衡、不完整信息等关键性问题进行了讨论,给出了一些常用的解决方法,以期为相关的研究和应用提供一个参考。
This paper briefly reviews reinforcement learning theory and its main algorithms firstly,then concentrates on several crux problems we may encounter when applying reinforcement learning in autonomous mobile robots,such as continuous state -action space,credit assignment problem,balance of exploration and exploitation,partially observable Markov decision problem and et al.It not only analyzes these problems but also gives out some methods to solve them.We hope it can be a reference for related work.
出处
《计算机工程与应用》
CSCD
北大核心
2004年第4期69-71,96,共4页
Computer Engineering and Applications
关键词
强化学习
自主机器人
多机器人系统
Reinforcement learning,Autonomous mobile robots,Multi-robots system
