Farama-Foundation/Stable-Retro

A fork of gym-retro with additional games, emulators and supported platforms

What it solves

Stable Retro transforms classic video games into standardized environments for reinforcement learning (RL). It allows researchers and developers to use retro games as benchmarks or training grounds for AI agents, providing a consistent interface (Gymnasium) to interact with the games.

How it works

The project is a maintained fork of OpenAI's gym-retro, extending it with more emulators, platforms, and games. It works by wrapping emulated games in a Gymnasium interface, allowing an AI agent to receive game state (via memory locations for in-game variables) and send inputs. It includes an integration tool to help users add new games by defining reward functions, episode end conditions, and memory locations for game variables.

Who it’s for

Reinforcement learning researchers, AI developers, and enthusiasts who want to use classic video games as complex environments for training and testing AI agents.

Highlights

  • Broad Platform Support: Supports a wide array of systems including Atari 2600, NES, SNES, N64, Gameboy, Sega Genesis, Sega Saturn, Dreamcast, and Arcade machines.
  • Huge Game Library: Integrated support for over 1,000 games across various genres like platformers, fighters, sports, and RPGs.
  • Gymnasium Integration: Provides a standardized interface for RL training, making it compatible with modern RL libraries like Stable Baselines3.
  • Game Integration Tool: Includes tools to facilitate the adding of new ROMs and states to the environment.

Related

  • Dispatch
    OpenAI Gym Retro Release

    OpenAI has released the full version of Gym Retro, a reinforcement learning platform that expands the available game library to over 1,000 titles across multiple classic consoles to enable research into agent generalization.

  • Dispatch
    OpenAI Retro Contest and Gym Retro Release

    OpenAI launched the Retro Contest to measure reinforcement learning generalization across unseen video game levels and released Gym Retro, a platform integrating 30 SEGA Genesis and 62 Atari 2600 games into Gym.

  • Project
    Farama-Foundation/Gymnasium

    A standard Python API and environment suite for developing and comparing reinforcement learning algorithms.

  • Project
    Farama-Foundation/Arcade-Learning-Environment

    A framework for developing AI agents for Atari 2600 games, providing a standardized interface between the Stella emulator and agent design.

  • Project
    PKU-Alignment/safety-gymnasium

    A scalable and customizable Safe Reinforcement Learning benchmark library providing a standardized set of environments for evaluating safety-constrained AI agents.