Application of Episodic Q-Learning to a Multi-agent Cooperative Task

Authors:
Akira Ito
Affiliations:
-
Venue:
PRICAI '02 Proceedings of the 7th Pacific Rim International Conference on Artificial Intelligence: Trends in Artificial Intelligence
Year:
2002

Citing 5
Cited 0

Technical Note: \cal Q-Learning

Machine Learning
Introduction to Reinforcement Learning

Introduction to Reinforcement Learning
Learning to Predict by the Methods of Temporal Differences

Machine Learning
Credit Assignment in Rule Discovery Systems Based on Genetic Algorithms

Machine Learning
Multi-Agent Reinforcement Learning for Planning and Scheduling Multiple Goals

ICMAS '00 Proceedings of the Fourth International Conference on MultiAgent Systems (ICMAS-2000)

Quantified Score

Hi-index	0.00

Visualization

Abstract

Episodic Q-learning is successfully applied to a multi-agent cooperative task, which is strongly non-Markovian and for which Q-learning is believed to have poor performance. The 3-hunter game, which is a modified version of the pursuit problem, is employed and the time necessary for hunters to capture the escapee is measured. By restricting the amount of the history used for learning, a significant increase in the speed of learning is realized. The success is not accidental, but based on the mind-reading algorithm we have proposed.