Solving Hierarchical Information-Sharing Dec-POMDPs: An Extensive-Form Game Approach

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Peralez, Johan, Delage, Aurélien, Buffet, Olivier, Dibangoye, Jilles S.
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912175395700736
author Peralez, Johan
Delage, Aurélien
Buffet, Olivier
Dibangoye, Jilles S.
author_facet Peralez, Johan
Delage, Aurélien
Buffet, Olivier
Dibangoye, Jilles S.
contents A recent theory shows that a multi-player decentralized partially observable Markov decision process can be transformed into an equivalent single-player game, enabling the application of \citeauthor{bellman}'s principle of optimality to solve the single-player game by breaking it down into single-stage subgames. However, this approach entangles the decision variables of all players at each single-stage subgame, resulting in backups with a double-exponential complexity. This paper demonstrates how to disentangle these decision variables while maintaining optimality under hierarchical information sharing, a prominent management style in our society. To achieve this, we apply the principle of optimality to solve any single-stage subgame by breaking it down further into smaller subgames, enabling us to make single-player decisions at a time. Our approach reveals that extensive-form games always exist with solutions to a single-stage subgame, significantly reducing time complexity. Our experimental results show that the algorithms leveraging these findings can scale up to much larger multi-player games without compromising optimality.
format Preprint
id arxiv_https___arxiv_org_abs_2402_02954
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Solving Hierarchical Information-Sharing Dec-POMDPs: An Extensive-Form Game Approach
Peralez, Johan
Delage, Aurélien
Buffet, Olivier
Dibangoye, Jilles S.
Computer Science and Game Theory
Machine Learning
A recent theory shows that a multi-player decentralized partially observable Markov decision process can be transformed into an equivalent single-player game, enabling the application of \citeauthor{bellman}'s principle of optimality to solve the single-player game by breaking it down into single-stage subgames. However, this approach entangles the decision variables of all players at each single-stage subgame, resulting in backups with a double-exponential complexity. This paper demonstrates how to disentangle these decision variables while maintaining optimality under hierarchical information sharing, a prominent management style in our society. To achieve this, we apply the principle of optimality to solve any single-stage subgame by breaking it down further into smaller subgames, enabling us to make single-player decisions at a time. Our approach reveals that extensive-form games always exist with solutions to a single-stage subgame, significantly reducing time complexity. Our experimental results show that the algorithms leveraging these findings can scale up to much larger multi-player games without compromising optimality.
title Solving Hierarchical Information-Sharing Dec-POMDPs: An Extensive-Form Game Approach
topic Computer Science and Game Theory
Machine Learning
url https://arxiv.org/abs/2402.02954