Peter Stone's Selected Publications

Classified by TopicClassified by Publication TypeSorted by DateSorted by First Author Last NameClassified by Funding Source


Grounded Semantic Networks for Learning Shared Communication Protocols

Matthew Hausknecht and Peter Stone. Grounded Semantic Networks for Learning Shared Communication Protocols. In Deep Reinforcement Learning, NIPS Workshop, December 2016.

Download

[PDF]899.9kB  

Abstract

Cooperative multiagent learning poses the challenge of coordinatingindependent agents. A powerful method to achieve coordination isallowing agents to communicate. We present the Grounded SemanticNetwork, an approach for learning a task-dependent communicationprotocol grounded in the observation space and reward function of thetask. We show that the grounded semantic network effectively learns acommunication protocol that is useful for achieving cooperationbetween agents. Analyzing the messages transmitted between agentsreveals that the agents' policies are highly influenced by thecommunication received from teammates. Further analysis highlights thelimitations of the grounded semantic network, identifying thecharacteristics of domains that it can and cannot solve.

BibTeX Entry

@InProceedings{NIPS-DeepRL16-hausknecht,
  author = {Matthew Hausknecht and Peter Stone},
  title = {Grounded Semantic Networks for Learning Shared Communication Protocols},
  booktitle = {Deep Reinforcement Learning, NIPS Workshop},
  location = {Barcelona, Spain},
  month = {December},
  year = {2016},
  abstract = {
Cooperative multiagent learning poses the challenge of coordinating
independent agents. A powerful method to achieve coordination is
allowing agents to communicate. We present the Grounded Semantic
Network, an approach for learning a task-dependent communication
protocol grounded in the observation space and reward function of the
task. We show that the grounded semantic network effectively learns a
communication protocol that is useful for achieving cooperation
between agents. Analyzing the messages transmitted between agents
reveals that the agents' policies are highly influenced by the
communication received from teammates. Further analysis highlights the
limitations of the grounded semantic network, identifying the
characteristics of domains that it can and cannot solve.
  },
}

Generated by bib2html.pl (written by Patrick Riley ) on Wed Nov 25, 2020 21:31:01