An approach for merging plans hierarchically in decomposable domains

J. Britanik, M. Marefat

Research output: Contribution to journalArticle

1 Scopus citations

Abstract

Recent works in domain-independent plan merging have shown that the optimal plan-merging problem is NP-hard, and heuristic algorithms have been proposed to generate near-optimal solutions. These works have shown heuristic algorithms which assume that the mergeability of two actions can be determined by considering only the actions themselves. In this paper, we show that it is often that case that the surrounding actions in the plan must also be considered when determining the mergeability of two or more actions; therefore, the context in which the actions are used affects their mergeability. To address this problem, we have developed a plan-merging methodology that merges partial-order plans based on the the notion of plan fragments, which encapsulate actions with the context in which they are being utilized. This mediodology applies to a class of planning domains, called decomposable domains, which consist of actions that follow all or some portion of a type sequence. Merging is performed hierarchically by action type. We also investigate the previously unexplored notion of alternative actions in domain-independent plan merging, which can improve the quality of the resulting merged plan. A novel approach is presented for removing cyclic dependencies that may occur during the plan-merging process. A key part of the methodology is the computation of a minimum cost cover of plan fragments. We provide theoretical analyses of two optimal algorithms and a greedy-based algorithm for computing the minimum cost cover. We support the theoretical analysis of these algorithms with experimental data to show that the greedy approach is near-optimal and efficient.

Original languageEnglish (US)
Pages (from-to)358-391
Number of pages34
JournalComputational Intelligence
Volume14
Issue number3
StatePublished - Aug 1998

    Fingerprint

Keywords

  • Plan merging
  • Planning

ASJC Scopus subject areas

  • Computational Mathematics
  • Artificial Intelligence

Cite this