| 部分可观察马尔可夫决策过程研究进展 |
| |
| 引用本文： | 仵博,吴敏. 部分可观察马尔可夫决策过程研究进展[J]. 计算机工程与设计, 2007, 28(9): 2116-2119,2126 |
| |
| 作者姓名： | 仵博 吴敏 |
| |
| 作者单位： | 深圳职业技术学院,计算机应用工程系,广东,深圳,518055;中南大学信息科学与工程学院,湖南,长沙,410083 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划) |
| |
| 摘 要： | 部分可观察马尔可夫决策过程是通过引入信念状态空间将非马尔可夫链问题转化为马尔可夫链问题来求解,其描述真实世界的特性使它成为研究随机决策过程的重要分支.介绍了部分可观察马尔可夫决策过程的基本原理和决策过程,然后介绍了3种典型的算法,它们分别是Littman等人的Witness算法、hcremental Pruning算法和Pineau等人的基于点的值迭代算法,对这3种算法进行了分析比较.讲述部分可观察马尔可夫决策过程的应用.
|
| 关 键 词： | 部分可观察马尔可夫 决策算法 智能体 马尔可夫链 值迭代算法 |
| 文章编号： | 1000-7024（2007）09-2116-04 |
| 修稿时间： | 2006-04-25 |
Research progress of partially observable Markov decision processes |
| |
| WU Bo,WU Min. Research progress of partially observable Markov decision processes[J]. Computer Engineering and Design, 2007, 28(9): 2116-2119,2126 |
| |
| Authors: | WU Bo WU Min |
| |
| Affiliation: | 1. Department of Computer Application Engineering, Shenzhen Polytechnic, Shenzhen 518055, China; 2. School of Information Science and Engineering, Central South University, Changsha 410083, China |
| |
| Abstract: | Partially observable Markov decision processes(POMDP) changes the non Markovian into Markovian over the belief state space.It has been an important branch of stochastic decision processes for its characteristics of describing the real world.At first,the principles and decision processes of POMDP is described,then three typical algorithms is presented,including Littman,et al's witness algorithm,incremental pruning algorithm and Pineau,et al's point-based value iteration algorithm,and the results from each of algorithms are analyzed and compared.At last,some applications are introduced using POMDP. |
| |
| Keywords: | POMDP decision algorithm agent Markovian Iteration algorithm |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
