Skip to Main content Skip to Navigation

Learning to Act in Decentralized Partially Observable MDPs

Abstract : We address a long-standing open problem of reinforcement learning in decentralized partially observable Markov decision processes. Previous attempts focussed on different forms of generalized policy iteration, which at best led to local optima. In this paper, we restrict attention to plans, which are simpler to store and update than policies. We derive, under certain conditions, the first near-optimal cooperative multi-agent reinforcement learning algorithm. To achieve significant scalability gains, we replace the greedy maximization by mixed-integer linear programming. Experiments show our approach can learn to act near-optimally in many finite domains from the literature.
Complete list of metadata

Cited literature [44 references]  Display  Hide  Download
Contributor : Jilles Steeve Dibangoye Connect in order to contact the contributor
Submitted on : Thursday, June 7, 2018 - 11:55:32 AM
Last modification on : Wednesday, November 3, 2021 - 7:57:48 AM
Long-term archiving on: : Saturday, September 8, 2018 - 1:50:50 PM


Files produced by the author(s)


  • HAL Id : hal-01809897, version 1


Jilles Dibangoye, Olivier Buffet. Learning to Act in Decentralized Partially Observable MDPs. [Research Report] RR-9179, INRIA Grenoble - Rhone-Alpes - CHROMA Team; INRIA Nancy, équipe LARSEN. 2018. ⟨hal-01809897⟩



Les métriques sont temporairement indisponibles