| 摘要: |
| 许多不确定环境下的自主机器人规划任务都可以用部分可观察的马氏决策过程(partially observableMarkov decision process,简称POMDP)建模.尽管研究者们在近似求解技术的设计方面已经取得了显著的进展,开发高效的POMDP 规划算法依然是一个具有挑战性的问题.以前的研究结果表明:在线规划方法能够高效地处理大规模的POMDP 问题,因而是一类具有研究前景的近似求解方法.这归因于它们采取的是“按需”作决策而不是预前对整个状态空间作决策的方式.旨在通过设计一个新颖的杂合启发式函数来进一步加速POMDP 在线规划过程,该函数能够充分利用现有算法里一些被忽略掉的启发式信息.实现了一个新的杂合启发式在线规划(hybrid heuristiconline planning,简称HHOP)算法.在一组POMDP 基准问题上,HHOP 有明显优于现有在线启发式搜索算法的实验性能. |
| 关键词: 部分可观察的马氏决策过程 在线规划 杂合启发法 |
| DOI：10.3724/SP.J.1001.2013.04318 |
| 分类号: |
| 基金项目:国家自然科学基金(60745002, 61175057); 国家高技术研究发展计划(863)(2008AA01Z150) |
|
| Hybrid Heuristic Online Planning for POMDPs |
|
ZHANG Zong-Zhang, CHEN Xiao-Ping
|
|
School of Computer Science and Technology, University of Science and Technology of China, Hefei 230027, China
|
| Abstract: |
| Lots of planning tasks of autonomous robots under uncertain environments can be modeled as a partially observable Markov decision processes (POMDPs). Although researchers have made impressive progress in designing approximation techniques, developing an efficient planning algorithm for POMDPs is still considered as a challenging problem. Previous research has indicated that online planning approaches are promising approximate methods for handling large-scale POMDP domains efficiently as they make decisions “on demand”, instead of proactively for the entire state space. This paper aims to further speed up the POMDP online planning process by designing a novel hybrid heuristic function, which provides a feasible way to take full advantage of some ignored heuristics in current algorithms. The research implements a new method called hybrid heuristic online planning (HHOP). HHOP substantially outperformes state-of-the-art online heuristic search approaches on a suite of POMDP benchmark problems. |
| Key words: partially observable Markov decision process (POMDP) online planning hybrid heuristics |
