【24h】

Utilizing Automatically Inferred Invariants in Graph Construction and Search

机译:在图的构造和搜索中使用自动推断的不变式

获取原文
获取原文并翻译 | 示例

摘要

In this paper we explore the relative importance of persistent and non-persistent mutex relations in the performance of Graphplan-based planners. We also show the advantages of pre-compiling persistent mutex relations. Using TIM we are able to generate, during a preprocessing analysis, all of the persistent binary mutex relations that would be inferred by Graphplan during graph construction. We show how the efficient storage of, and access to, these pre-processed persistent mu-texes yields a modest improvement in graph construction performance. We further demonstrate that the process by which these persistent mutexes are identified can, in certain kinds of domain, allow the exploitation of binary mutex relations which are inaccessible to Graphplan. We present The Island of Sodor, a simple planning domain characterizing a class of domains in which certain persistent mutexes are present but are not detectable by Graphplan during graph construction. We show that the exploitation of these hidden binary mutexes makes problems in this kind of domain trivially solvable by STAN, where they are intractable for other Graphplan-based planners.
机译:在本文中,我们探讨了基于Graphplan的计划者性能中持久和非持久互斥关系的相对重要性。我们还展示了预编译持久互斥关系的优点。使用TIM,我们能够在预处理分析期间生成所有持久的二进制互斥关系,这些关系将由Graphplan在图构建期间推断出来。我们展示了如何有效存储和访问这些经过预处理的持久性mu-tex,从而适度改善图形构造性能。我们进一步证明,在某些类型的域中,识别这些持久互斥锁的过程可以允许利用Graphplan无法访问的二进制互斥锁关系。我们介绍了Sodor岛,这是一个简单的规划域,其特征是一类域,其中存在某些持久性互斥体,但在图构建过程中Graphplan无法检测到。我们表明,利用这些隐藏的二进制互斥锁使STAN可以轻松解决此类领域中的问题,而对于其他基于Graphplan的计划人员来说,这些问题是很难解决的。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号