Learning multiagent communication with backpropagation

Sainbayar Sukhbaatar, Arthur Szlam, Robert Fergus

Research output: Contribution to journalArticle

Abstract

Many tasks in AI require the collaboration of multiple agents. Typically, the communication protocol between agents is manually specified and not altered during training. In this paper we explore a simple neural model, called CommNet, that uses continuous communication for fully cooperative tasks. The model consists of multiple agents and the communication between them is learned alongside their policy. We apply this model to a diverse set of tasks, demonstrating the ability of the agents to learn to communicate amongst themselves, yielding improved performance over non-communicative agents and baselines. In some cases, it is possible to interpret the language devised by the agents, revealing simple but effective strategies for solving the task at hand.

Original languageEnglish (US)
Pages (from-to)2252-2260
Number of pages9
JournalAdvances in Neural Information Processing Systems
StatePublished - 2016

Fingerprint

Backpropagation
Communication
Network protocols

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Learning multiagent communication with backpropagation. / Sukhbaatar, Sainbayar; Szlam, Arthur; Fergus, Robert.

In: Advances in Neural Information Processing Systems, 2016, p. 2252-2260.

Research output: Contribution to journalArticle

@article{f06f5406d92e49d2b10ed77d9cbd8f8c,
title = "Learning multiagent communication with backpropagation",
abstract = "Many tasks in AI require the collaboration of multiple agents. Typically, the communication protocol between agents is manually specified and not altered during training. In this paper we explore a simple neural model, called CommNet, that uses continuous communication for fully cooperative tasks. The model consists of multiple agents and the communication between them is learned alongside their policy. We apply this model to a diverse set of tasks, demonstrating the ability of the agents to learn to communicate amongst themselves, yielding improved performance over non-communicative agents and baselines. In some cases, it is possible to interpret the language devised by the agents, revealing simple but effective strategies for solving the task at hand.",
author = "Sainbayar Sukhbaatar and Arthur Szlam and Robert Fergus",
year = "2016",
language = "English (US)",
pages = "2252--2260",
journal = "Advances in Neural Information Processing Systems",
issn = "1049-5258",

}

TY - JOUR

T1 - Learning multiagent communication with backpropagation

AU - Sukhbaatar, Sainbayar

AU - Szlam, Arthur

AU - Fergus, Robert

PY - 2016

Y1 - 2016

N2 - Many tasks in AI require the collaboration of multiple agents. Typically, the communication protocol between agents is manually specified and not altered during training. In this paper we explore a simple neural model, called CommNet, that uses continuous communication for fully cooperative tasks. The model consists of multiple agents and the communication between them is learned alongside their policy. We apply this model to a diverse set of tasks, demonstrating the ability of the agents to learn to communicate amongst themselves, yielding improved performance over non-communicative agents and baselines. In some cases, it is possible to interpret the language devised by the agents, revealing simple but effective strategies for solving the task at hand.

AB - Many tasks in AI require the collaboration of multiple agents. Typically, the communication protocol between agents is manually specified and not altered during training. In this paper we explore a simple neural model, called CommNet, that uses continuous communication for fully cooperative tasks. The model consists of multiple agents and the communication between them is learned alongside their policy. We apply this model to a diverse set of tasks, demonstrating the ability of the agents to learn to communicate amongst themselves, yielding improved performance over non-communicative agents and baselines. In some cases, it is possible to interpret the language devised by the agents, revealing simple but effective strategies for solving the task at hand.

UR - http://www.scopus.com/inward/record.url?scp=85018860957&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85018860957&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:85018860957

SP - 2252

EP - 2260

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

SN - 1049-5258

ER -