Presentation 2009-07-14
Speeding up Multi-Agent Reinforcement Learning Using a Ring-Type State Recognition Method
Kyohei ONO, Hidehiro NAKANO, Arata MIYAUCHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) Recently, design of actions in complex and large-scale robot networks has been required. However, it is difficult to control learning of multiple agents considering each action of them. Especially, as the number of agents increases, the number of states which the agents observe increases exponentially, and the learning speed decreases significantly. In this study, we propose a ring-type state observation method for large-scale multi-agent reinforcement learning. Various network topologies in observing location information of each agent can be considered. Through numerical experiments, these topologies are evaluated in the viewpoints of reduction of states, learning speed and solution quality.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Multi-agent system / Reinforcement learning / Network topology
Paper # NLP2009-30,NC2009-23
Date of Issue

Conference Information
Committee NLP
Conference Date 2009/7/6(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Nonlinear Problems (NLP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Speeding up Multi-Agent Reinforcement Learning Using a Ring-Type State Recognition Method
Sub Title (in English)
Keyword(1) Multi-agent system
Keyword(2) Reinforcement learning
Keyword(3) Network topology
1st Author's Name Kyohei ONO
1st Author's Affiliation Tokyo City University()
2nd Author's Name Hidehiro NAKANO
2nd Author's Affiliation Tokyo City University
3rd Author's Name Arata MIYAUCHI
3rd Author's Affiliation Tokyo City University
Date 2009-07-14
Paper # NLP2009-30,NC2009-23
Volume (vol) vol.109
Number (no) 124
Page pp.pp.-
#Pages 5
Date of Issue