Policy gradient approach for learning of soccer player agents: Pass selection of midfielders

Harukazu Igarashi, Hitoshi Fukuoka, Seiji Ishihara

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

Abstract

This research develops a learning method for the pass selection problem of midfielders in RoboCup Soccer Simulation games. A policy gradient method is applied as a learning method to solve this problem because it can easily represent the various heuristics of pass selection in a policy function. We implement the learning function in the midfielders' programs of a well-known team, UvA Trilearn Base 2003. Experimental results show that our method effectively achieves clever pass selection by midfielders in full games. Moreover, in this method's framework, dribbling is learned as a pass technique, in essence to and from the passer itself. It is also shown that the improvement in pass selection by our learning helps to make a team much stronger.

Original languageEnglish
Title of host publicationLecture Notes in Electrical Engineering
Pages137-148
Number of pages12
Volume70 LNEE
DOIs
Publication statusPublished - 2011
EventInternational Conference on Advances in Intelligent Control and Computer Engineering - Hong Kong
Duration: 2010 Mar 172010 Mar 19

Publication series

NameLecture Notes in Electrical Engineering
Volume70 LNEE
ISSN (Print)18761100
ISSN (Electronic)18761119

Other

OtherInternational Conference on Advances in Intelligent Control and Computer Engineering
CityHong Kong
Period10/3/1710/3/19

Fingerprint

Gradient methods

Keywords

  • Multi-agent system
  • Pass selection
  • Policy gradient method
  • Reinforcement learning
  • RoboCup

ASJC Scopus subject areas

  • Industrial and Manufacturing Engineering

Cite this

Igarashi, H., Fukuoka, H., & Ishihara, S. (2011). Policy gradient approach for learning of soccer player agents: Pass selection of midfielders. In Lecture Notes in Electrical Engineering (Vol. 70 LNEE, pp. 137-148). (Lecture Notes in Electrical Engineering; Vol. 70 LNEE). https://doi.org/10.1007/978-94-007-0286-8_12

Policy gradient approach for learning of soccer player agents : Pass selection of midfielders. / Igarashi, Harukazu; Fukuoka, Hitoshi; Ishihara, Seiji.

Lecture Notes in Electrical Engineering. Vol. 70 LNEE 2011. p. 137-148 (Lecture Notes in Electrical Engineering; Vol. 70 LNEE).

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

Igarashi, H, Fukuoka, H & Ishihara, S 2011, Policy gradient approach for learning of soccer player agents: Pass selection of midfielders. in Lecture Notes in Electrical Engineering. vol. 70 LNEE, Lecture Notes in Electrical Engineering, vol. 70 LNEE, pp. 137-148, International Conference on Advances in Intelligent Control and Computer Engineering, Hong Kong, 10/3/17. https://doi.org/10.1007/978-94-007-0286-8_12
Igarashi H, Fukuoka H, Ishihara S. Policy gradient approach for learning of soccer player agents: Pass selection of midfielders. In Lecture Notes in Electrical Engineering. Vol. 70 LNEE. 2011. p. 137-148. (Lecture Notes in Electrical Engineering). https://doi.org/10.1007/978-94-007-0286-8_12
Igarashi, Harukazu ; Fukuoka, Hitoshi ; Ishihara, Seiji. / Policy gradient approach for learning of soccer player agents : Pass selection of midfielders. Lecture Notes in Electrical Engineering. Vol. 70 LNEE 2011. pp. 137-148 (Lecture Notes in Electrical Engineering).
@inproceedings{ced1814418c04297a132e6b183744d1e,
title = "Policy gradient approach for learning of soccer player agents: Pass selection of midfielders",
abstract = "This research develops a learning method for the pass selection problem of midfielders in RoboCup Soccer Simulation games. A policy gradient method is applied as a learning method to solve this problem because it can easily represent the various heuristics of pass selection in a policy function. We implement the learning function in the midfielders' programs of a well-known team, UvA Trilearn Base 2003. Experimental results show that our method effectively achieves clever pass selection by midfielders in full games. Moreover, in this method's framework, dribbling is learned as a pass technique, in essence to and from the passer itself. It is also shown that the improvement in pass selection by our learning helps to make a team much stronger.",
keywords = "Multi-agent system, Pass selection, Policy gradient method, Reinforcement learning, RoboCup",
author = "Harukazu Igarashi and Hitoshi Fukuoka and Seiji Ishihara",
year = "2011",
doi = "10.1007/978-94-007-0286-8_12",
language = "English",
isbn = "9789400702851",
volume = "70 LNEE",
series = "Lecture Notes in Electrical Engineering",
pages = "137--148",
booktitle = "Lecture Notes in Electrical Engineering",

}

TY - GEN

T1 - Policy gradient approach for learning of soccer player agents

T2 - Pass selection of midfielders

AU - Igarashi, Harukazu

AU - Fukuoka, Hitoshi

AU - Ishihara, Seiji

PY - 2011

Y1 - 2011

N2 - This research develops a learning method for the pass selection problem of midfielders in RoboCup Soccer Simulation games. A policy gradient method is applied as a learning method to solve this problem because it can easily represent the various heuristics of pass selection in a policy function. We implement the learning function in the midfielders' programs of a well-known team, UvA Trilearn Base 2003. Experimental results show that our method effectively achieves clever pass selection by midfielders in full games. Moreover, in this method's framework, dribbling is learned as a pass technique, in essence to and from the passer itself. It is also shown that the improvement in pass selection by our learning helps to make a team much stronger.

AB - This research develops a learning method for the pass selection problem of midfielders in RoboCup Soccer Simulation games. A policy gradient method is applied as a learning method to solve this problem because it can easily represent the various heuristics of pass selection in a policy function. We implement the learning function in the midfielders' programs of a well-known team, UvA Trilearn Base 2003. Experimental results show that our method effectively achieves clever pass selection by midfielders in full games. Moreover, in this method's framework, dribbling is learned as a pass technique, in essence to and from the passer itself. It is also shown that the improvement in pass selection by our learning helps to make a team much stronger.

KW - Multi-agent system

KW - Pass selection

KW - Policy gradient method

KW - Reinforcement learning

KW - RoboCup

UR - http://www.scopus.com/inward/record.url?scp=78651546538&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78651546538&partnerID=8YFLogxK

U2 - 10.1007/978-94-007-0286-8_12

DO - 10.1007/978-94-007-0286-8_12

M3 - Conference contribution

AN - SCOPUS:78651546538

SN - 9789400702851

VL - 70 LNEE

T3 - Lecture Notes in Electrical Engineering

SP - 137

EP - 148

BT - Lecture Notes in Electrical Engineering

ER -