Incremental Clustering and Expansion for Faster Optimal Planning in Decentralized POMDPs

Frans A. Oliehoek, Matthijs T. J. Spaan, Christopher Amato, Shimon Whiteson

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

This article presents the state-of-the-art in optimal solution methods for decentralized partially observable Markov decision processes (Dec-POMDPs), which are general models for collaborative multiagent planning under uncertainty. Building off the generalized multiagent A* (GMAA*) algorithm, which reduces the problem to a tree of one-shot collaborative Bayesian games (CBGs), we describe several advances that greatly expand the range of Dec-POMDPs that can be solved optimally. First, we introduce lossless incremental clustering of the CBGs solved by GMAA*, which achieves exponential speedups without sacrificing optimality. Second, we introduce incremental expansion of nodes in the GMAA* search tree, which avoids the need to expand all children, the number of which is in the worst case doubly exponential in the node's depth. This is particularly beneficial when little clustering is possible. In addition, we introduce new hybrid heuristic representations that are more compact and thereby enable the solution of larger Dec-POMDPs. We provide theoretical guarantees that, when a suitable heuristic is used, both incremental clustering and incremental expansion yield algorithms that are both complete and search equivalent. Finally, we present extensive empirical results demonstrating that GMAA*-ICE, an algorithm that synthesizes these advances, can optimally solve Dec-POMDPs of unprecedented size.
Original languageEnglish
Pages (from-to)449-509
Number of pages61
JournalJournal of Artificial Intelligence Research
Volume46
DOIs
Publication statusPublished - 2013

Cite this

Oliehoek, Frans A. ; Spaan, Matthijs T. J. ; Amato, Christopher ; Whiteson, Shimon. / Incremental Clustering and Expansion for Faster Optimal Planning in Decentralized POMDPs. In: Journal of Artificial Intelligence Research. 2013 ; Vol. 46. pp. 449-509.
@article{9ad816ccbeae4f4ba431fa41cf3d0d6e,
title = "Incremental Clustering and Expansion for Faster Optimal Planning in Decentralized POMDPs",
abstract = "This article presents the state-of-the-art in optimal solution methods for decentralized partially observable Markov decision processes (Dec-POMDPs), which are general models for collaborative multiagent planning under uncertainty. Building off the generalized multiagent A* (GMAA*) algorithm, which reduces the problem to a tree of one-shot collaborative Bayesian games (CBGs), we describe several advances that greatly expand the range of Dec-POMDPs that can be solved optimally. First, we introduce lossless incremental clustering of the CBGs solved by GMAA*, which achieves exponential speedups without sacrificing optimality. Second, we introduce incremental expansion of nodes in the GMAA* search tree, which avoids the need to expand all children, the number of which is in the worst case doubly exponential in the node's depth. This is particularly beneficial when little clustering is possible. In addition, we introduce new hybrid heuristic representations that are more compact and thereby enable the solution of larger Dec-POMDPs. We provide theoretical guarantees that, when a suitable heuristic is used, both incremental clustering and incremental expansion yield algorithms that are both complete and search equivalent. Finally, we present extensive empirical results demonstrating that GMAA*-ICE, an algorithm that synthesizes these advances, can optimally solve Dec-POMDPs of unprecedented size.",
author = "Oliehoek, {Frans A.} and Spaan, {Matthijs T. J.} and Christopher Amato and Shimon Whiteson",
year = "2013",
doi = "10.1613/jair.3804",
language = "English",
volume = "46",
pages = "449--509",
journal = "Journal of Artificial Intelligence Research",
issn = "1076-9757",
publisher = "AI Access Foundation",

}

Incremental Clustering and Expansion for Faster Optimal Planning in Decentralized POMDPs. / Oliehoek, Frans A.; Spaan, Matthijs T. J.; Amato, Christopher; Whiteson, Shimon.

In: Journal of Artificial Intelligence Research, Vol. 46, 2013, p. 449-509.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Incremental Clustering and Expansion for Faster Optimal Planning in Decentralized POMDPs

AU - Oliehoek, Frans A.

AU - Spaan, Matthijs T. J.

AU - Amato, Christopher

AU - Whiteson, Shimon

PY - 2013

Y1 - 2013

N2 - This article presents the state-of-the-art in optimal solution methods for decentralized partially observable Markov decision processes (Dec-POMDPs), which are general models for collaborative multiagent planning under uncertainty. Building off the generalized multiagent A* (GMAA*) algorithm, which reduces the problem to a tree of one-shot collaborative Bayesian games (CBGs), we describe several advances that greatly expand the range of Dec-POMDPs that can be solved optimally. First, we introduce lossless incremental clustering of the CBGs solved by GMAA*, which achieves exponential speedups without sacrificing optimality. Second, we introduce incremental expansion of nodes in the GMAA* search tree, which avoids the need to expand all children, the number of which is in the worst case doubly exponential in the node's depth. This is particularly beneficial when little clustering is possible. In addition, we introduce new hybrid heuristic representations that are more compact and thereby enable the solution of larger Dec-POMDPs. We provide theoretical guarantees that, when a suitable heuristic is used, both incremental clustering and incremental expansion yield algorithms that are both complete and search equivalent. Finally, we present extensive empirical results demonstrating that GMAA*-ICE, an algorithm that synthesizes these advances, can optimally solve Dec-POMDPs of unprecedented size.

AB - This article presents the state-of-the-art in optimal solution methods for decentralized partially observable Markov decision processes (Dec-POMDPs), which are general models for collaborative multiagent planning under uncertainty. Building off the generalized multiagent A* (GMAA*) algorithm, which reduces the problem to a tree of one-shot collaborative Bayesian games (CBGs), we describe several advances that greatly expand the range of Dec-POMDPs that can be solved optimally. First, we introduce lossless incremental clustering of the CBGs solved by GMAA*, which achieves exponential speedups without sacrificing optimality. Second, we introduce incremental expansion of nodes in the GMAA* search tree, which avoids the need to expand all children, the number of which is in the worst case doubly exponential in the node's depth. This is particularly beneficial when little clustering is possible. In addition, we introduce new hybrid heuristic representations that are more compact and thereby enable the solution of larger Dec-POMDPs. We provide theoretical guarantees that, when a suitable heuristic is used, both incremental clustering and incremental expansion yield algorithms that are both complete and search equivalent. Finally, we present extensive empirical results demonstrating that GMAA*-ICE, an algorithm that synthesizes these advances, can optimally solve Dec-POMDPs of unprecedented size.

U2 - 10.1613/jair.3804

DO - 10.1613/jair.3804

M3 - Article

VL - 46

SP - 449

EP - 509

JO - Journal of Artificial Intelligence Research

JF - Journal of Artificial Intelligence Research

SN - 1076-9757

ER -