Strategy for learning cooperative behavior with local information for multi-agent systems

Fumito Uwano; Keiki Takadama

doi:10.1007/978-3-030-03098-8_54

Strategy for learning cooperative behavior with local information for multi-agent systems

Fumito Uwano, Keiki Takadama

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

1 Citation (Scopus)

Abstract

Toward learning cooperative behavior for any number of agents, this paper proposes a multi-agent reinforcement learning method without communication, called PMRL-based Learning for Any number of Agents (PLAA). PLAA prevents from agents reaching the purpose for spending too many times, and to promote the local multi-agent cooperation without communication by PMRL as a previous method. To guarantee the effectiveness of PLAA, this paper compares PLAA with Q-learning, and two previous methods in 10 kinds of the maze for the 2 and 3 agents. From the experimental result, we revealed those things: (a) PLAA is the most effective method for cooperation among 2 and 3 agents; (b) PLAA enable the agents to cooperate with each other in small iterations.

Original language	English
Title of host publication	PRIMA 2018
Subtitle of host publication	Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings
Editors	Nir Oren, Yuko Sakurai, Itsuki Noda, Tran Cao Son, Tim Miller, Bastin Tony Savarimuthu
Publisher	Springer Verlag
Pages	663-670
Number of pages	8
ISBN (Print)	9783030030971
DOIs	https://doi.org/10.1007/978-3-030-03098-8_54
Publication status	Published - 2018
Externally published	Yes
Event	21st International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2018 - Tokyo, Japan Duration: Oct 29 2018 → Nov 2 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11224 LNAI
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	21st International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2018
Country	Japan
City	Tokyo
Period	10/29/18 → 11/2/18

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-030-03098-8_54

Fingerprint Dive into the research topics of 'Strategy for learning cooperative behavior with local information for multi-agent systems'. Together they form a unique fingerprint.

Cite this

Uwano, F., & Takadama, K. (2018). Strategy for learning cooperative behavior with local information for multi-agent systems. In N. Oren, Y. Sakurai, I. Noda, T. Cao Son, T. Miller, & B. T. Savarimuthu (Eds.), PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings (pp. 663-670). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11224 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-030-03098-8_54

Strategy for learning cooperative behavior with local information for multi-agent systems. / Uwano, Fumito; Takadama, Keiki.

PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings. ed. / Nir Oren; Yuko Sakurai; Itsuki Noda; Tran Cao Son; Tim Miller; Bastin Tony Savarimuthu. Springer Verlag, 2018. p. 663-670 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11224 LNAI).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Uwano, F & Takadama, K 2018, Strategy for learning cooperative behavior with local information for multi-agent systems. in N Oren, Y Sakurai, I Noda, T Cao Son, T Miller & BT Savarimuthu (eds), PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11224 LNAI, Springer Verlag, pp. 663-670, 21st International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2018, Tokyo, Japan, 10/29/18. https://doi.org/10.1007/978-3-030-03098-8_54

Uwano F, Takadama K. Strategy for learning cooperative behavior with local information for multi-agent systems. In Oren N, Sakurai Y, Noda I, Cao Son T, Miller T, Savarimuthu BT, editors, PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings. Springer Verlag. 2018. p. 663-670. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-03098-8_54

Uwano, Fumito ; Takadama, Keiki. / Strategy for learning cooperative behavior with local information for multi-agent systems. PRIMA 2018: Principles and Practice of Multi-Agent Systems - 21st International Conference, 2018, Proceedings. editor / Nir Oren ; Yuko Sakurai ; Itsuki Noda ; Tran Cao Son ; Tim Miller ; Bastin Tony Savarimuthu. Springer Verlag, 2018. pp. 663-670 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{4c15688524244020bb96275fa86789b7,

title = "Strategy for learning cooperative behavior with local information for multi-agent systems",

abstract = "Toward learning cooperative behavior for any number of agents, this paper proposes a multi-agent reinforcement learning method without communication, called PMRL-based Learning for Any number of Agents (PLAA). PLAA prevents from agents reaching the purpose for spending too many times, and to promote the local multi-agent cooperation without communication by PMRL as a previous method. To guarantee the effectiveness of PLAA, this paper compares PLAA with Q-learning, and two previous methods in 10 kinds of the maze for the 2 and 3 agents. From the experimental result, we revealed those things: (a) PLAA is the most effective method for cooperation among 2 and 3 agents; (b) PLAA enable the agents to cooperate with each other in small iterations.",

author = "Fumito Uwano and Keiki Takadama",

note = "Funding Information: This work was supported by JSPS KAKENHI Grant Number JP17J08724.; 21st International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2018 ; Conference date: 29-10-2018 Through 02-11-2018",

year = "2018",

doi = "10.1007/978-3-030-03098-8_54",

language = "English",

isbn = "9783030030971",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

pages = "663--670",

editor = "Nir Oren and Yuko Sakurai and Itsuki Noda and {Cao Son}, Tran and Tim Miller and Savarimuthu, {Bastin Tony}",

booktitle = "PRIMA 2018",

}

TY - GEN

T1 - Strategy for learning cooperative behavior with local information for multi-agent systems

AU - Uwano, Fumito

AU - Takadama, Keiki

N1 - Funding Information: This work was supported by JSPS KAKENHI Grant Number JP17J08724.

PY - 2018

Y1 - 2018

N2 - Toward learning cooperative behavior for any number of agents, this paper proposes a multi-agent reinforcement learning method without communication, called PMRL-based Learning for Any number of Agents (PLAA). PLAA prevents from agents reaching the purpose for spending too many times, and to promote the local multi-agent cooperation without communication by PMRL as a previous method. To guarantee the effectiveness of PLAA, this paper compares PLAA with Q-learning, and two previous methods in 10 kinds of the maze for the 2 and 3 agents. From the experimental result, we revealed those things: (a) PLAA is the most effective method for cooperation among 2 and 3 agents; (b) PLAA enable the agents to cooperate with each other in small iterations.

AB - Toward learning cooperative behavior for any number of agents, this paper proposes a multi-agent reinforcement learning method without communication, called PMRL-based Learning for Any number of Agents (PLAA). PLAA prevents from agents reaching the purpose for spending too many times, and to promote the local multi-agent cooperation without communication by PMRL as a previous method. To guarantee the effectiveness of PLAA, this paper compares PLAA with Q-learning, and two previous methods in 10 kinds of the maze for the 2 and 3 agents. From the experimental result, we revealed those things: (a) PLAA is the most effective method for cooperation among 2 and 3 agents; (b) PLAA enable the agents to cooperate with each other in small iterations.

UR - http://www.scopus.com/inward/record.url?scp=85056485163&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85056485163&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-03098-8_54

DO - 10.1007/978-3-030-03098-8_54

M3 - Conference contribution

AN - SCOPUS:85056485163

SN - 9783030030971

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 663

EP - 670

BT - PRIMA 2018

A2 - Oren, Nir

A2 - Sakurai, Yuko

A2 - Noda, Itsuki

A2 - Cao Son, Tran

A2 - Miller, Tim

A2 - Savarimuthu, Bastin Tony

PB - Springer Verlag

T2 - 21st International Conference on Principles and Practice of Multi-Agent Systems, PRIMA 2018

Y2 - 29 October 2018 through 2 November 2018

ER -

Strategy for learning cooperative behavior with local information for multi-agent systems

Abstract

Publication series

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Cite this