An Automatically Configurable Portfolio-based Planner with Macro-actions

PbP

Alfonso E. Gerevini, Alessandro Saetti, Mauro Vallati

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

31 Citations (Scopus)

Abstract

While several powerful domain-independent planners have recently been developed, no one of these clearly outperforms all the others in every known benchmark domain. We present PbP, a multi-planner which automatically configures a portfolio of planners by (i) computing some sets of macro-actions for every planner in the portfolio, (ii) selecting a promising combination of planners in the portfolio and relative useful macro-actions, and (iii) defining some running time slots for their round-robin scheduling during planning. The configuration relies on some knowledge about the performance of the planners in the portfolio and relative macro-actions which is automatically generated from a training problem set. PbP entered the learning track of IPC-2008 and was the overall winner of this competition track. An experimental study confirms the effectiveness of PbP, and shows that the learned configuration knowledge is useful for PbP.

Original languageEnglish
Title of host publicationICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling
PublisherAAAI press
Pages350-353
Number of pages4
ISBN (Print)9781577354062
Publication statusPublished - 16 Oct 2009
Externally publishedYes
Event19th International Conference on Automated Planning and Scheduling - Thessaloniki, Greece
Duration: 19 Sep 200923 Sep 2009
Conference number: 19
http://icaps09.uom.gr/ (Link to Conference Website)

Conference

Conference19th International Conference on Automated Planning and Scheduling
Abbreviated titleICAPS 2009
CountryGreece
CityThessaloniki
Period19/09/0923/09/09
Internet address

Fingerprint

Experimental study
Benchmark
Planning

Cite this

Gerevini, A. E., Saetti, A., & Vallati, M. (2009). An Automatically Configurable Portfolio-based Planner with Macro-actions: PbP. In ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling (pp. 350-353). AAAI press.
Gerevini, Alfonso E. ; Saetti, Alessandro ; Vallati, Mauro. / An Automatically Configurable Portfolio-based Planner with Macro-actions : PbP. ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling. AAAI press, 2009. pp. 350-353
@inproceedings{3c1e4f1bb80649d1a1dc0902439ad35a,
title = "An Automatically Configurable Portfolio-based Planner with Macro-actions: PbP",
abstract = "While several powerful domain-independent planners have recently been developed, no one of these clearly outperforms all the others in every known benchmark domain. We present PbP, a multi-planner which automatically configures a portfolio of planners by (i) computing some sets of macro-actions for every planner in the portfolio, (ii) selecting a promising combination of planners in the portfolio and relative useful macro-actions, and (iii) defining some running time slots for their round-robin scheduling during planning. The configuration relies on some knowledge about the performance of the planners in the portfolio and relative macro-actions which is automatically generated from a training problem set. PbP entered the learning track of IPC-2008 and was the overall winner of this competition track. An experimental study confirms the effectiveness of PbP, and shows that the learned configuration knowledge is useful for PbP.",
keywords = "Automated multi-planner configuratio, Learning and macro actions for planning, Portfolio-based planning",
author = "Gerevini, {Alfonso E.} and Alessandro Saetti and Mauro Vallati",
year = "2009",
month = "10",
day = "16",
language = "English",
isbn = "9781577354062",
pages = "350--353",
booktitle = "ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling",
publisher = "AAAI press",

}

Gerevini, AE, Saetti, A & Vallati, M 2009, An Automatically Configurable Portfolio-based Planner with Macro-actions: PbP. in ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling. AAAI press, pp. 350-353, 19th International Conference on Automated Planning and Scheduling, Thessaloniki, Greece, 19/09/09.

An Automatically Configurable Portfolio-based Planner with Macro-actions : PbP. / Gerevini, Alfonso E.; Saetti, Alessandro; Vallati, Mauro.

ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling. AAAI press, 2009. p. 350-353.

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

TY - GEN

T1 - An Automatically Configurable Portfolio-based Planner with Macro-actions

T2 - PbP

AU - Gerevini, Alfonso E.

AU - Saetti, Alessandro

AU - Vallati, Mauro

PY - 2009/10/16

Y1 - 2009/10/16

N2 - While several powerful domain-independent planners have recently been developed, no one of these clearly outperforms all the others in every known benchmark domain. We present PbP, a multi-planner which automatically configures a portfolio of planners by (i) computing some sets of macro-actions for every planner in the portfolio, (ii) selecting a promising combination of planners in the portfolio and relative useful macro-actions, and (iii) defining some running time slots for their round-robin scheduling during planning. The configuration relies on some knowledge about the performance of the planners in the portfolio and relative macro-actions which is automatically generated from a training problem set. PbP entered the learning track of IPC-2008 and was the overall winner of this competition track. An experimental study confirms the effectiveness of PbP, and shows that the learned configuration knowledge is useful for PbP.

AB - While several powerful domain-independent planners have recently been developed, no one of these clearly outperforms all the others in every known benchmark domain. We present PbP, a multi-planner which automatically configures a portfolio of planners by (i) computing some sets of macro-actions for every planner in the portfolio, (ii) selecting a promising combination of planners in the portfolio and relative useful macro-actions, and (iii) defining some running time slots for their round-robin scheduling during planning. The configuration relies on some knowledge about the performance of the planners in the portfolio and relative macro-actions which is automatically generated from a training problem set. PbP entered the learning track of IPC-2008 and was the overall winner of this competition track. An experimental study confirms the effectiveness of PbP, and shows that the learned configuration knowledge is useful for PbP.

KW - Automated multi-planner configuratio

KW - Learning and macro actions for planning

KW - Portfolio-based planning

UR - http://www.scopus.com/inward/record.url?scp=78650587990&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781577354062

SP - 350

EP - 353

BT - ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling

PB - AAAI press

ER -

Gerevini AE, Saetti A, Vallati M. An Automatically Configurable Portfolio-based Planner with Macro-actions: PbP. In ICAPS 2009 - Proceedings of the 19th International Conference on Automated Planning and Scheduling. AAAI press. 2009. p. 350-353