The first probabilistic track of the international planning competition.pdf

下载文档 降价啦

2
0
约8.98万字
约 38页
2017-04-06 发布于江苏
举报
版权申诉
保障服务

The first probabilistic track of the international planning competition.pdf

1、本文档共38页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

The first probabilistic track of the international planning competition

The First Probabilistic Track of the International Planning Competition Ha?kan L. S. Younes; Michael L. Littman; David Weissman; John Asmuth Journal of Artificial Intelligence Research 24: 851–887. c?2005 AI Access Foundation. All rights reserved. /papers/paper1880.html Journal of Artificial Intelligence Research 24 (2005) 851-887 Submitted 08/05; published 12/05 The First Probabilistic Track of the International Planning Competition H?akan L. S. Younes lorens@ Computer Science Department Carnegie Mellon University Pittsburgh, PA 15213 USA Michael L. Littman mlittman@ David Weissman dweisman@ John Asmuth jasmuth@ Department of Computer Science Rutgers University Piscataway, NJ 08854 USA Abstract The 2004 International Planning Competition, IPC-4, included a probabilistic planning track for the first time. We describe the new domain specification language we created for the track, our evaluation methodology, the competition domains we developed, and the results of the participating teams. 1. Background The Fourth International Planning Competition (IPC-4) was held as part of the Interna- tional Conference on Planning and Scheduling (ICAPS’04) in Vancouver, British Columbia in June 2004. By request of the ICAPS’04 organizers, Sven Koenig and Shlomo Zilberstein, we were asked to create the first probabilistic planning track as part of IPC-4. The overriding goal of the first probabilistic planning track was to bring together two communities converging on a similar set of research issues and aid them in creating com- parable tools and evaluation metrics. One community consists of Markov decision process (MDP) researchers interested in developing algorithms that apply to powerfully expressive representations of environments. The other consists of planning researchers incorporating probabilistic and decision theoretic concepts into their planning algorithms. Cross fertil- ization has begun, but the intent of the probabilistic planning track was to create a set of shared be