Learning macro-actions genetically from plans

M. A. H. Newton, J. Levine, M. Fox, D. Long

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

Abstract

Despite recent progress in planning, many complex domains and even larger problems in simple domains remain hard and challenging. One way to achieve further improvement is to utilise knowledge acquired for the planner from the domain. Macro-actions are a promising means by which to convey such knowledge. A macro-action, or macro in short, is a group of actions selected for application as a single choice. Most existing works on macros utilise knowledge explicitly specific to the planners and the domains. But presumably any
particular properties are not likely to be common with different planners or wider range of domains. Therefore, a macro learning system that does not exploit any structural knowledge about planners and domains explicitly is of immense interest. This paper presents an offline system capable of learning macros genetically from plans. Given a planner, a domain, and necessary problems, our system generates macros, lifting from plans of some problems, under guidance from a genetic algorithm. It represents macros like regular actions, evaluates them individually by solving other problems, and suggests the best macro to be added to the domain permanently. Genetic algorithms are automatic learning methods that can learn properties of a system using no explicit knowledge
about it. Our system thus does not strive to discover or utilise any structural properties specific to a planner or a domain.
Original languageEnglish
Title of host publicationProceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006)
Number of pages8
Publication statusPublished - 1 Dec 2006

Fingerprint

Macros
Genetic algorithms
Learning systems
Structural properties
Planning

Keywords

  • macros
  • planning
  • macro-actions
  • genetic algorithms

Cite this

Newton, M. A. H., Levine, J., Fox, M., & Long, D. (2006). Learning macro-actions genetically from plans. In Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006)
Newton, M. A. H. ; Levine, J. ; Fox, M. ; Long, D. / Learning macro-actions genetically from plans. Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006). 2006.
@inproceedings{d37927ba39e245b997c7978e6711cba0,
title = "Learning macro-actions genetically from plans",
abstract = "Despite recent progress in planning, many complex domains and even larger problems in simple domains remain hard and challenging. One way to achieve further improvement is to utilise knowledge acquired for the planner from the domain. Macro-actions are a promising means by which to convey such knowledge. A macro-action, or macro in short, is a group of actions selected for application as a single choice. Most existing works on macros utilise knowledge explicitly specific to the planners and the domains. But presumably any particular properties are not likely to be common with different planners or wider range of domains. Therefore, a macro learning system that does not exploit any structural knowledge about planners and domains explicitly is of immense interest. This paper presents an offline system capable of learning macros genetically from plans. Given a planner, a domain, and necessary problems, our system generates macros, lifting from plans of some problems, under guidance from a genetic algorithm. It represents macros like regular actions, evaluates them individually by solving other problems, and suggests the best macro to be added to the domain permanently. Genetic algorithms are automatic learning methods that can learn properties of a system using no explicit knowledge about it. Our system thus does not strive to discover or utilise any structural properties specific to a planner or a domain.",
keywords = "macros, planning, macro-actions, genetic algorithms",
author = "Newton, {M. A. H.} and J. Levine and M. Fox and D. Long",
year = "2006",
month = "12",
day = "1",
language = "English",
booktitle = "Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006)",

}

Newton, MAH, Levine, J, Fox, M & Long, D 2006, Learning macro-actions genetically from plans. in Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006).

Learning macro-actions genetically from plans. / Newton, M. A. H.; Levine, J.; Fox, M.; Long, D.

Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006). 2006.

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

TY - GEN

T1 - Learning macro-actions genetically from plans

AU - Newton, M. A. H.

AU - Levine, J.

AU - Fox, M.

AU - Long, D.

PY - 2006/12/1

Y1 - 2006/12/1

N2 - Despite recent progress in planning, many complex domains and even larger problems in simple domains remain hard and challenging. One way to achieve further improvement is to utilise knowledge acquired for the planner from the domain. Macro-actions are a promising means by which to convey such knowledge. A macro-action, or macro in short, is a group of actions selected for application as a single choice. Most existing works on macros utilise knowledge explicitly specific to the planners and the domains. But presumably any particular properties are not likely to be common with different planners or wider range of domains. Therefore, a macro learning system that does not exploit any structural knowledge about planners and domains explicitly is of immense interest. This paper presents an offline system capable of learning macros genetically from plans. Given a planner, a domain, and necessary problems, our system generates macros, lifting from plans of some problems, under guidance from a genetic algorithm. It represents macros like regular actions, evaluates them individually by solving other problems, and suggests the best macro to be added to the domain permanently. Genetic algorithms are automatic learning methods that can learn properties of a system using no explicit knowledge about it. Our system thus does not strive to discover or utilise any structural properties specific to a planner or a domain.

AB - Despite recent progress in planning, many complex domains and even larger problems in simple domains remain hard and challenging. One way to achieve further improvement is to utilise knowledge acquired for the planner from the domain. Macro-actions are a promising means by which to convey such knowledge. A macro-action, or macro in short, is a group of actions selected for application as a single choice. Most existing works on macros utilise knowledge explicitly specific to the planners and the domains. But presumably any particular properties are not likely to be common with different planners or wider range of domains. Therefore, a macro learning system that does not exploit any structural knowledge about planners and domains explicitly is of immense interest. This paper presents an offline system capable of learning macros genetically from plans. Given a planner, a domain, and necessary problems, our system generates macros, lifting from plans of some problems, under guidance from a genetic algorithm. It represents macros like regular actions, evaluates them individually by solving other problems, and suggests the best macro to be added to the domain permanently. Genetic algorithms are automatic learning methods that can learn properties of a system using no explicit knowledge about it. Our system thus does not strive to discover or utilise any structural properties specific to a planner or a domain.

KW - macros

KW - planning

KW - macro-actions

KW - genetic algorithms

UR - http://www.cis.strath.ac.uk/cis/research/publications/papers/strath_cis_publication_2167.pdf

M3 - Conference contribution book

BT - Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006)

ER -

Newton MAH, Levine J, Fox M, Long D. Learning macro-actions genetically from plans. In Proceedings of the 25th Workshop of the UK Planning and Scheduling Special Interest Group (PlanSIG 2006). 2006