• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2010, Vol. 32 ›› Issue (1): 55-56.doi: 10.3969/j.issn.1007130X.2010.

• 论文 • 上一篇    下一篇

基于模糊聚类的分层强化学习算法

  

  1. (长沙理工大学计算机与通信工程学院,湖南 长沙 410076)
  • 收稿日期:2008-11-19 修回日期:2009-02-25 出版日期:2010-01-18 发布日期:2010-01-18
  • 通讯作者: 410114 湖南省长沙市长沙理工大学(云塘校区)至诚轩3A E-mail:ujn_zhangxin@163.com
  • 作者简介:张欣(1982-),男,山东东营人,硕士生,研究方向为人工智能和强化学习等;戴帅,硕士生,研究方向为机器学习和强化学习等。

A Hierarchical Reinforcement Learning Algorithm Based on Fuzzy Clustering

  • Received:2008-11-19 Revised:2009-02-25 Online:2010-01-18 Published:2010-01-18

摘要:

本文提出了一种新的分层强化学习Option自动生成算法,以Agent在学习初始阶段探测到的状态空间为输入,采用模糊逻辑神经元的网络进行聚类,在聚类后的各状态子集上通过经验回放学习产生内部策略集,生成Option,仿真实验结果表明了该算法的有效性。

关键词: 强化学习, 分层强化学习, 模糊聚类, Option

Abstract:

A new algorithm for the automatic generation of the Option Hierarchical Reinforcement Learning is presented. The algorithm takes the state space detected by the agent as input in the initial learning phase,and clusters the states by employing fuzzy clustering. Based on the clustered state sets, the intrastrategies are learned by an experience replay procedure. As a result, the options are generated. The validity of the algorithm is demonstrated by simulation experiments.

Key words: reinforcement learning;hierarchical reinforcement learning;fuzzy clustering;Option

中图分类号: