MAgent: A Many-Agent Reinforcement Learning Platform for Artificial Collective Intelligence

TL;DR

MAgent is a scalable platform enabling research on large-scale multi-agent reinforcement learning, supporting environments with up to one million agents to study collective behaviors and social phenomena.

Contribution

It introduces a highly scalable platform for many-agent reinforcement learning, facilitating the study of emergent social behaviors in AI societies at unprecedented scales.

Findings

Emergence of collective intelligence in large-scale multi-agent environments

Support for environments with up to one million agents on a single GPU

Demonstration of social phenomena like communication and leadership among agents

Abstract

We introduce MAgent, a platform to support research and development of many-agent reinforcement learning. Unlike previous research platforms on single or multi-agent reinforcement learning, MAgent focuses on supporting the tasks and the applications that require hundreds to millions of agents. Within the interactions among a population of agents, it enables not only the study of learning algorithms for agents' optimal polices, but more importantly, the observation and understanding of individual agent's behaviors and social phenomena emerging from the AI society, including communication languages, leaderships, altruism. MAgent is highly scalable and can host up to one million agents on a single GPU server. MAgent also provides flexible configurations for AI researchers to design their customized environments and agents. In this demo, we present three environments designed on MAgent and show emerged collective intelligence by learning from scratch.