| 分层强化学习原理研究 |
| |
| 引用本文： | 柯文德,陈珂,余凤燕.分层强化学习原理研究[J].茂名学院学报,2013(4):30-33,52. |
| |
| 作者姓名： | 柯文德 陈珂 余凤燕 |
| |
| 作者单位： | 1. 广东石油化工学院计算机科学与技术系,广东茂名,525000
2. 茂名职业技术学院机电信息系,广东茂名,525000 |
| |
| 基金项目： | 国家自然科学基金项目(61272382)；广东省自然科学基金项目(8152500002000003， S2012010009963)；广东省高等学校科技创新项目(2012KJCX0077)；广东高校石化装备故障诊断与信息化控制工程中心项目 |
| |
| 摘 要： | 首先介绍了强化学习基本原理，分析了马尔科夫决策过程与半马尔科夫决策过程的理论基础及其在强化学习中的应用，其次阐述了分层强化学习中分层与抽象的思想，分析了HAM、Options与MaxQ等方法，并从分层与抽象角度进行了比较，最后指出了分层强化学习的研究发展方向。
|
| 关 键 词： | 分层强化学习 半马尔科夫决策过程 抽象 收敛 学习 |
A Study on the Principle of Hierarchical Reinforcement Learning |
| |
| KE Wen-de
,
CHEN Ke
,
YU Feng-yan.A Study on the Principle of Hierarchical Reinforcement Learning[J].Journal of Maoming College,2013(4):30-33,52. |
| |
| Authors: | KE Wen-de CHEN Ke YU Feng-yan |
| |
| Institution: | 1 Department of Computer Science,Guangdong University of Petrochemical Technology,Maoming 525000,China;2 Department of Mechanical and Electronic Information,Maoming Polytechnic,Maoming 525011,China) |
| |
| Abstract: | Firstly, the principle of RL （reinforcement learning） was introduced and the theories and applications of MDP （Markov decision process） and SMDP for RL were analyzed. Secondly, the concepts of layer division and abstraction were demonstrated and the three HRL methods, including HAM, Options, MaxQ, were analyzed and compared from the aspects of layer division and abstraction. Finally the de- veloping directions of HRL were given. |
| |
| Keywords: | hierarchical reinforcement learning (HRL) semi Markov decision process (SMDP) abstraction convergence study |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
