Markov games of incomplete information for multi-agent reinforcement learning

Liam Mac Dermed; Charles L. Isbell; Lora G. Weiss

Markov games of incomplete information for multi-agent reinforcement learning

Liam Mac Dermed, Charles L. Isbell, Lora G. Weiss

Office of the Senior Vice President for Research

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

8 Scopus citations

Abstract

Partially observable stochastic games (POSGs) are an attractive model for many multi-agent domains, but are computationally extremely difficult to solve. We present a new model, Markov games of incomplete information (MGII) which imposes a mild restriction on POSGs while overcoming their primary computational bottleneck. Finally we show how to convert a MGII into a continuous but bounded fully observable stochastic game. MGIIs represents the most general tractable model for multi-agent reinforcement learning to date.

Original language	English (US)
Title of host publication	Interactive Decision Theory and Game Theory - Papers from the 2011 AAAI Workshop, Technical Report
Pages	43-51
Number of pages	9
State	Published - Nov 2 2011
Event	2011 AAAI Workshop - San Francisco, CA, United States Duration: Aug 8 2011 → Aug 8 2011

Publication series

Name	AAAI Workshop - Technical Report
Volume	WS-11-13

Conference

Conference	2011 AAAI Workshop
Country/Territory	United States
City	San Francisco, CA
Period	8/8/11 → 8/8/11

All Science Journal Classification (ASJC) codes

Engineering(all)

Cite this

@inproceedings{0ab2b3c6a9c44719b5482372b0963c74,

title = "Markov games of incomplete information for multi-agent reinforcement learning",

abstract = "Partially observable stochastic games (POSGs) are an attractive model for many multi-agent domains, but are computationally extremely difficult to solve. We present a new model, Markov games of incomplete information (MGII) which imposes a mild restriction on POSGs while overcoming their primary computational bottleneck. Finally we show how to convert a MGII into a continuous but bounded fully observable stochastic game. MGIIs represents the most general tractable model for multi-agent reinforcement learning to date.",

author = "{Mac Dermed}, Liam and Isbell, {Charles L.} and Weiss, {Lora G.}",

year = "2011",

month = nov,

day = "2",

language = "English (US)",

isbn = "9781577355298",

series = "AAAI Workshop - Technical Report",

pages = "43--51",

booktitle = "Interactive Decision Theory and Game Theory - Papers from the 2011 AAAI Workshop, Technical Report",

note = "2011 AAAI Workshop ; Conference date: 08-08-2011 Through 08-08-2011",

}

Markov games of incomplete information for multi-agent reinforcement learning. / Mac Dermed, Liam; Isbell, Charles L.; Weiss, Lora G.

Interactive Decision Theory and Game Theory - Papers from the 2011 AAAI Workshop, Technical Report. 2011. p. 43-51 (AAAI Workshop - Technical Report; Vol. WS-11-13).

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

TY - GEN

T1 - Markov games of incomplete information for multi-agent reinforcement learning

AU - Mac Dermed, Liam

AU - Isbell, Charles L.

AU - Weiss, Lora G.

PY - 2011/11/2

Y1 - 2011/11/2

N2 - Partially observable stochastic games (POSGs) are an attractive model for many multi-agent domains, but are computationally extremely difficult to solve. We present a new model, Markov games of incomplete information (MGII) which imposes a mild restriction on POSGs while overcoming their primary computational bottleneck. Finally we show how to convert a MGII into a continuous but bounded fully observable stochastic game. MGIIs represents the most general tractable model for multi-agent reinforcement learning to date.

AB - Partially observable stochastic games (POSGs) are an attractive model for many multi-agent domains, but are computationally extremely difficult to solve. We present a new model, Markov games of incomplete information (MGII) which imposes a mild restriction on POSGs while overcoming their primary computational bottleneck. Finally we show how to convert a MGII into a continuous but bounded fully observable stochastic game. MGIIs represents the most general tractable model for multi-agent reinforcement learning to date.

UR - http://www.scopus.com/inward/record.url?scp=80055062734&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80055062734&partnerID=8YFLogxK

M3 - Conference contribution

AN - SCOPUS:80055062734

SN - 9781577355298

T3 - AAAI Workshop - Technical Report

SP - 43

EP - 51

BT - Interactive Decision Theory and Game Theory - Papers from the 2011 AAAI Workshop, Technical Report

T2 - 2011 AAAI Workshop

Y2 - 8 August 2011 through 8 August 2011

ER -

Markov games of incomplete information for multi-agent reinforcement learning

Abstract

Publication series

Conference

All Science Journal Classification (ASJC) codes

Other files and links

Fingerprint

Cite this