Farama-Foundation/MAgent2

An engine for high performance multi-agent environments with very large numbers of agents, along with a set of reference environments

What it solves

MAgent2 provides a platform for simulating large numbers of agents in a gridworld environment where they can interact in competitive scenarios, such as battles, to study artificial collective intelligence.

How it works

It is a library that creates pixel-based gridworld environments. It implements reference environments using the PettingZoo API, allowing researchers to benchmark and train agents in multi-agent reinforcement learning settings.

Who it’s for

Researchers and developers working on multi-agent reinforcement learning (MARL) and collective intelligence.

Highlights

  • Pixel-based gridworld environments
  • Support for large numbers of agents
  • Integrated with the PettingZoo API
  • Maintained fork of the original MAgent codebase

Related

  • Project
  • Project
  • Project
  • Project
  • Project