Skip to Main content Skip to Navigation
Conference papers

Multiagent Planning and Learning As MILP

Jilles Dibangoye 1 Olivier Buffet 2 Akshat Kumar 3
1 CHROMA - Robots coopératifs et adaptés à la présence humaine en environnements dynamiques
Inria Grenoble - Rhône-Alpes, CITI - CITI Centre of Innovation in Telecommunications and Integration of services
2 LARSEN - Lifelong Autonomy and interaction skills for Robots in a Sensing ENvironment
Inria Nancy - Grand Est, LORIA - AIS - Department of Complex Systems, Artificial Intelligence & Robotics
Abstract : The decentralized partially observable Markov decisionprocess offers a unified framework for sequential decision-making by multiple collaborating agents but remains in-tractable. Mixed-integer linear formulations proved use-ful for partially observable domains, unfortunately ex-isting applications restrict to domains with one or twoagents. In this paper, we exploit a linearization propertythat allows us to reformulate nonlinear constraints fromn-agent settings into linear ones. We further present plan-ning and learning approaches relying on MILP formula-tions for general and special cases, including network-distributed and transition-independent problems. Experi-ments on standard2-agent benchmarks as well as domainswith a large number of agents provide strong empiricalsupport to the methodology.
Document type :
Conference papers
Complete list of metadatas

https://hal.inria.fr/hal-03081548
Contributor : Olivier Buffet <>
Submitted on : Friday, December 18, 2020 - 11:13:55 AM
Last modification on : Monday, December 21, 2020 - 3:02:02 PM

File

JSD_JFPDA_20.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-03081548, version 1

Citation

Jilles Dibangoye, Olivier Buffet, Akshat Kumar. Multiagent Planning and Learning As MILP. Journées Francophones surla Planification, la Décision et l’Apprentissagepour la conduite de systèmes, Jun 2020, Angers (virtuel), France. ⟨hal-03081548⟩

Share

Metrics

Record views

19

Files downloads

124