R-learning with multiple state-action value tables

Koichiro Ishikawa; Akito Sakurai; Tsutomu Fujinami; Susumu Kunifuji

doi:10.1002/eej.20473

R-learning with multiple state-action value tables

Koichiro Ishikawa, Akito Sakurai, Tsutomu Fujinami, Susumu Kunifuji

Research output: Contribution to journal › Article › peer-review

Abstract

We propose a method to improve the performance of R-learning, a reinforcement learning algorithm, by using multiple state-action value tables. Unlike Q- or Sarsa learning, R-learning learns a policy to maximize undiscounted rewards. Multiple state-action value tables cause substantial explorations as needed and make R-learning work well. Efficiency of the proposed method is verified through experiments in a simulated environment.

Original language	English
Pages (from-to)	34-47
Number of pages	14
Journal	Electrical Engineering in Japan (English translation of Denki Gakkai Ronbunshi)
Volume	159
Issue number	3
DOIs	https://doi.org/10.1002/eej.20473
Publication status	Published - 2007 May
Externally published	Yes

Keywords

Autonomous mobile robot
R-learning
Reinforcement learning

ASJC Scopus subject areas

Energy Engineering and Power Technology
Electrical and Electronic Engineering

Access to Document

10.1002/eej.20473

Cite this

@article{c6113896dbc948b7874c710ff807c779,

title = "R-learning with multiple state-action value tables",

abstract = "We propose a method to improve the performance of R-learning, a reinforcement learning algorithm, by using multiple state-action value tables. Unlike Q- or Sarsa learning, R-learning learns a policy to maximize undiscounted rewards. Multiple state-action value tables cause substantial explorations as needed and make R-learning work well. Efficiency of the proposed method is verified through experiments in a simulated environment.",

keywords = "Autonomous mobile robot, R-learning, Reinforcement learning",

author = "Koichiro Ishikawa and Akito Sakurai and Tsutomu Fujinami and Susumu Kunifuji",

year = "2007",

month = may,

doi = "10.1002/eej.20473",

language = "English",

volume = "159",

pages = "34--47",

journal = "Electrical Engineering in Japan (English translation of Denki Gakkai Ronbunshi)",

issn = "0424-7760",

publisher = "John Wiley and Sons Inc.",

number = "3",

}

TY - JOUR

T1 - R-learning with multiple state-action value tables

AU - Ishikawa, Koichiro

AU - Sakurai, Akito

AU - Fujinami, Tsutomu

AU - Kunifuji, Susumu

PY - 2007/5

Y1 - 2007/5

N2 - We propose a method to improve the performance of R-learning, a reinforcement learning algorithm, by using multiple state-action value tables. Unlike Q- or Sarsa learning, R-learning learns a policy to maximize undiscounted rewards. Multiple state-action value tables cause substantial explorations as needed and make R-learning work well. Efficiency of the proposed method is verified through experiments in a simulated environment.

AB - We propose a method to improve the performance of R-learning, a reinforcement learning algorithm, by using multiple state-action value tables. Unlike Q- or Sarsa learning, R-learning learns a policy to maximize undiscounted rewards. Multiple state-action value tables cause substantial explorations as needed and make R-learning work well. Efficiency of the proposed method is verified through experiments in a simulated environment.

KW - Autonomous mobile robot

KW - R-learning

KW - Reinforcement learning

UR - http://www.scopus.com/inward/record.url?scp=34247179549&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=34247179549&partnerID=8YFLogxK

U2 - 10.1002/eej.20473

DO - 10.1002/eej.20473

M3 - Article

AN - SCOPUS:34247179549

SN - 0424-7760

VL - 159

SP - 34

EP - 47

JO - Electrical Engineering in Japan (English translation of Denki Gakkai Ronbunshi)

JF - Electrical Engineering in Japan (English translation of Denki Gakkai Ronbunshi)

IS - 3

ER -

R-learning with multiple state-action value tables

Abstract

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this