Learning action descriptions of opponent behaviour in the robocup 2D simulation environment

Authors:
A. Illobre;J. Gonzalez;R. Otero;J. Santos
Affiliations:
Computer Science Department, University of Corunna, Spain;Computer Science Department, University of Corunna, Spain;Computer Science Department, University of Corunna, Spain;Computer Science Department, University of Corunna, Spain
Venue:
ILP'10 Proceedings of the 20th international conference on Inductive logic programming
Year:
2010

Citing 6
Cited 1

Learning to Take Actions

Machine Learning
Inductive Verification and Validation of the KULRoT RoboCup Team

RoboCup-98: Robot Soccer World Cup II
Learning Programs in the Event Calculus

ILP '97 Proceedings of the 7th International Workshop on Inductive Logic Programming
Learning Action Models for Reactive Autonomous Agents

Learning Action Models for Reactive Autonomous Agents
Adapting behavior by inductive prediction in soccer agents

PRICAI'00 Proceedings of the 6th Pacific Rim international conference on Artificial intelligence
Skill acquisition via transfer learning and advice taking

ECML'06 Proceedings of the 17th European conference on Machine Learning

An overview on opponent modeling in RoboCup soccer simulation 2D

Robot Soccer World Cup XV

Quantified Score

Hi-index	0.00

Visualization

Abstract

The Robocup 2D simulation competition [13] proposes a dynamic environment where two opponent teams are confronted in a simplified soccer game. All major teams use a fixed algorithm to control its players. An unexpected opponent strategy, not previously considered by the developers, might result in winning all matches. To improve this we use ILP to learn action descriptions of opponent players; for learning on dynamic domains, we have to deal with the frame problem. The induced descriptions can be used to plan for desired field states. To show this we start with a simplified scenario where we learn the behaviour of a goalkeeper based on the actions of a shooter player. This description is used to plan for states where a goal can be scored. This result can directly be extended to a multiplayer environment.