Research Area:  Machine Learning

Abstract:

Over recent years, deep reinforcement learning has shown strong successes in complex single-agent tasks, and more recently this approach has also been applied to multi-agent domains. In this paper, we propose a novel approach, called MAGnet, to multi-agent reinforcement learning (MARL) that utilizes a relevance graph representation of the environment obtained by a self-attention mechanism, and a message-generation technique inspired by the NerveNet architecture. We applied our MAGnet approach to the Pommerman game and the results show that it significantly outperforms state-of-the-art MARL solutions, including DQN, MADDPG, and MCTS.

Keywords:  
Multi-Agent Reinforcement Learning
Graphs
Deep Learning

Author(s) Name:  Aleksandra Malysheva, Tegg Taekyong Sung, Chae-Bong Sohn, Daniel Kudenko, Aleksei Shpilman

Publisher name:  arXiv:1811.12557

DOI:  10.48550/arXiv.1811.12557

Volume Information:  

Paper Link:   https://arxiv.org/abs/1811.12557