This repository provides OpenAI Gym environments for Super Mario Bros. and Super Mario Bros. 2 (Lost Levels) on the NES, targeting reinforcement learning researchers and developers. It enables agents to interact with the classic games through a standardized interface, facilitating the development and benchmarking of AI agents.

How It Works

The environments leverage the nes-py emulator to run the games. They offer a flexible action space, defaulting to the full 256-action NES controller input, but can be constrained to predefined action sets like SIMPLE_MOVEMENT or COMPLEX_MOVEMENT for easier agent control. The reward function is designed to encourage forward progress and survival, combining distance moved, time remaining, and a penalty for dying.

Quick Start & Requirements

• Install via pip: pip install gym-super-mario-bros
• Requires Python.
• Official documentation: https://github.com/Kautenja/gym-super-mario-bros

Highlighted Details

• Supports 32 stages for Super Mario Bros. and Super Mario Bros. 2 (Lost Levels).
• Includes environments for individual stages and random stage selection.
• Provides various ROM modes (standard, downsample, pixel, rectangle) affecting visual fidelity and potentially performance.
• The info dictionary returned by step includes detailed game state like coins, lives, score, and Mario's position.

Maintenance & Community

• Primarily maintained by Christian Kauten.
• No explicit community links (Discord/Slack) are provided in the README.

Licensing & Compatibility

• The repository does not explicitly state a license in the README. This requires further investigation for commercial use or closed-source linking.

Limitations & Caveats

The README does not specify a license, which is a significant caveat for adoption. Random stage selection is currently only available for the original Super Mario Bros., not Lost Levels.