Sequential Monte Carlo in probabilistic planning reachability heuristics

Daniel Bryce, Subbarao Kambhampati, David E. Smith

Research output: Chapter in Book/Report/Conference proceedingConference contribution

11 Citations (Scopus)

Abstract

The current best conformant probabilistic planners encode the problem as a bounded length CSP or SAT problem. While these approaches can find optimal solutions for given plan lengths, they often do not scale for large problems or plan lengths. As has been shown in classical planning, heuristic search outperforms CSP/SAT techniques (especially when a plan length is not given a priori). The problem with applying heuristic search in probabilistic planning is that effective heuristics are as yet lacking. In this work, we apply heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning. We evaluate a straight-forward application of these planning graph techniques, which amounts to exactly computing the distribution over reachable relaxed planning graph layers. Computing these distributions is costly, so we apply Sequential Monte Carlo to approximate them. We demonstrate on several domains how our approach enables our planner to far out-scale existing (optimal) probabilistic planners and still find reasonable quality solutions.

Original languageEnglish (US)
Title of host publicationICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling
Pages233-242
Number of pages10
Volume2006
StatePublished - 2006
EventICAPS 2006 - 16th International Conference on Automated Planning and Scheduling - Cumbria, United Kingdom
Duration: Jun 6 2006Jun 10 2006

Other

OtherICAPS 2006 - 16th International Conference on Automated Planning and Scheduling
CountryUnited Kingdom
CityCumbria
Period6/6/066/10/06

Fingerprint

Planning

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Bryce, D., Kambhampati, S., & Smith, D. E. (2006). Sequential Monte Carlo in probabilistic planning reachability heuristics. In ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling (Vol. 2006, pp. 233-242)

Sequential Monte Carlo in probabilistic planning reachability heuristics. / Bryce, Daniel; Kambhampati, Subbarao; Smith, David E.

ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling. Vol. 2006 2006. p. 233-242.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bryce, D, Kambhampati, S & Smith, DE 2006, Sequential Monte Carlo in probabilistic planning reachability heuristics. in ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling. vol. 2006, pp. 233-242, ICAPS 2006 - 16th International Conference on Automated Planning and Scheduling, Cumbria, United Kingdom, 6/6/06.
Bryce D, Kambhampati S, Smith DE. Sequential Monte Carlo in probabilistic planning reachability heuristics. In ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling. Vol. 2006. 2006. p. 233-242
Bryce, Daniel ; Kambhampati, Subbarao ; Smith, David E. / Sequential Monte Carlo in probabilistic planning reachability heuristics. ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling. Vol. 2006 2006. pp. 233-242
@inproceedings{91e3f946164a4c1cb55ac5c7f9e287ab,
title = "Sequential Monte Carlo in probabilistic planning reachability heuristics",
abstract = "The current best conformant probabilistic planners encode the problem as a bounded length CSP or SAT problem. While these approaches can find optimal solutions for given plan lengths, they often do not scale for large problems or plan lengths. As has been shown in classical planning, heuristic search outperforms CSP/SAT techniques (especially when a plan length is not given a priori). The problem with applying heuristic search in probabilistic planning is that effective heuristics are as yet lacking. In this work, we apply heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning. We evaluate a straight-forward application of these planning graph techniques, which amounts to exactly computing the distribution over reachable relaxed planning graph layers. Computing these distributions is costly, so we apply Sequential Monte Carlo to approximate them. We demonstrate on several domains how our approach enables our planner to far out-scale existing (optimal) probabilistic planners and still find reasonable quality solutions.",
author = "Daniel Bryce and Subbarao Kambhampati and Smith, {David E.}",
year = "2006",
language = "English (US)",
volume = "2006",
pages = "233--242",
booktitle = "ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling",

}

TY - GEN

T1 - Sequential Monte Carlo in probabilistic planning reachability heuristics

AU - Bryce, Daniel

AU - Kambhampati, Subbarao

AU - Smith, David E.

PY - 2006

Y1 - 2006

N2 - The current best conformant probabilistic planners encode the problem as a bounded length CSP or SAT problem. While these approaches can find optimal solutions for given plan lengths, they often do not scale for large problems or plan lengths. As has been shown in classical planning, heuristic search outperforms CSP/SAT techniques (especially when a plan length is not given a priori). The problem with applying heuristic search in probabilistic planning is that effective heuristics are as yet lacking. In this work, we apply heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning. We evaluate a straight-forward application of these planning graph techniques, which amounts to exactly computing the distribution over reachable relaxed planning graph layers. Computing these distributions is costly, so we apply Sequential Monte Carlo to approximate them. We demonstrate on several domains how our approach enables our planner to far out-scale existing (optimal) probabilistic planners and still find reasonable quality solutions.

AB - The current best conformant probabilistic planners encode the problem as a bounded length CSP or SAT problem. While these approaches can find optimal solutions for given plan lengths, they often do not scale for large problems or plan lengths. As has been shown in classical planning, heuristic search outperforms CSP/SAT techniques (especially when a plan length is not given a priori). The problem with applying heuristic search in probabilistic planning is that effective heuristics are as yet lacking. In this work, we apply heuristic search to conformant probabilistic planning by adapting planning graph heuristics developed for non-deterministic planning. We evaluate a straight-forward application of these planning graph techniques, which amounts to exactly computing the distribution over reachable relaxed planning graph layers. Computing these distributions is costly, so we apply Sequential Monte Carlo to approximate them. We demonstrate on several domains how our approach enables our planner to far out-scale existing (optimal) probabilistic planners and still find reasonable quality solutions.

UR - http://www.scopus.com/inward/record.url?scp=33746054528&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=33746054528&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:33746054528

VL - 2006

SP - 233

EP - 242

BT - ICAPS 2006 - Proceedings, Sixteenth International Conference on Automated Planning and Scheduling

ER -