Learning action strategies for planning domains using genetic programming

J. Levine, D. Humphreys

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

11 Citations (Scopus)

Abstract

There are many different approaches to solving planning problems, one of which is the use of domain specific control knowledge to help guide a domain independent search algorithm. This paper presents L2Plan which represents this control knowledge as an ordered set of control rules, called a policy, and learns using genetic programming. The genetic program’s crossover and mutation operators are augmented by a simple local search. L2Plan was tested on both the blocks world and briefcase domains. In both domains, L2Plan was able to produce policies that solved all the test problems and which outperformed the hand-coded policies written by the authors.
Original languageEnglish
Title of host publicationProceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group
Number of pages13
Publication statusPublished - 1 Dec 2003

Fingerprint

Genetic programming
Planning

Keywords

  • genetic programming
  • action strategies
  • planning domains

Cite this

Levine, J., & Humphreys, D. (2003). Learning action strategies for planning domains using genetic programming. In Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group
Levine, J. ; Humphreys, D. / Learning action strategies for planning domains using genetic programming. Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group. 2003.
@inproceedings{e6d66ea6431e4b64841a4db07c5f5e3d,
title = "Learning action strategies for planning domains using genetic programming",
abstract = "There are many different approaches to solving planning problems, one of which is the use of domain specific control knowledge to help guide a domain independent search algorithm. This paper presents L2Plan which represents this control knowledge as an ordered set of control rules, called a policy, and learns using genetic programming. The genetic program’s crossover and mutation operators are augmented by a simple local search. L2Plan was tested on both the blocks world and briefcase domains. In both domains, L2Plan was able to produce policies that solved all the test problems and which outperformed the hand-coded policies written by the authors.",
keywords = "genetic programming, action strategies, planning domains",
author = "J. Levine and D. Humphreys",
year = "2003",
month = "12",
day = "1",
language = "English",
booktitle = "Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group",

}

Levine, J & Humphreys, D 2003, Learning action strategies for planning domains using genetic programming. in Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group.

Learning action strategies for planning domains using genetic programming. / Levine, J.; Humphreys, D.

Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group. 2003.

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

TY - GEN

T1 - Learning action strategies for planning domains using genetic programming

AU - Levine, J.

AU - Humphreys, D.

PY - 2003/12/1

Y1 - 2003/12/1

N2 - There are many different approaches to solving planning problems, one of which is the use of domain specific control knowledge to help guide a domain independent search algorithm. This paper presents L2Plan which represents this control knowledge as an ordered set of control rules, called a policy, and learns using genetic programming. The genetic program’s crossover and mutation operators are augmented by a simple local search. L2Plan was tested on both the blocks world and briefcase domains. In both domains, L2Plan was able to produce policies that solved all the test problems and which outperformed the hand-coded policies written by the authors.

AB - There are many different approaches to solving planning problems, one of which is the use of domain specific control knowledge to help guide a domain independent search algorithm. This paper presents L2Plan which represents this control knowledge as an ordered set of control rules, called a policy, and learns using genetic programming. The genetic program’s crossover and mutation operators are augmented by a simple local search. L2Plan was tested on both the blocks world and briefcase domains. In both domains, L2Plan was able to produce policies that solved all the test problems and which outperformed the hand-coded policies written by the authors.

KW - genetic programming

KW - action strategies

KW - planning domains

UR - http://www.cis.strath.ac.uk/cis/research/publications/papers/strath_cis_publication_2182.pdf

M3 - Conference contribution book

BT - Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group

ER -

Levine J, Humphreys D. Learning action strategies for planning domains using genetic programming. In Proceedings of the 22nd Workshop of the UK Planning and Scheduling Special Interest Group. 2003