RoboLeague provides a car soccer environment for deep reinforcement learning research, specifically targeting adversarial self-play. It is designed for researchers and practitioners in RL who need a realistic simulation inspired by Rocket League for training agents.

How It Works

The environment is built using the Unity ML Agents Toolkit, leveraging its physics and agent control capabilities. It aims to replicate key Rocket League mechanics like ball physics, car aerial control, and ground acceleration, with plans to integrate dodges and a more refined lateral friction model.

Quick Start & Requirements

• Installation requires the Unity ML-Agents Release 11. Specific installation steps are detailed within the Unity ML-Agents repository.
• Prerequisites include the Unity game engine and the ML-Agents Toolkit.
• Further details and integration guidance can be found in the Unity ML-Agents Documentation.

Highlighted Details

• Mimics core Rocket League physics and car control.
• Designed for adversarial self-play experiments.
• Planned OpenAI Gym interface integration.

Maintenance & Community

Information regarding maintainers, community channels, or roadmaps is not explicitly provided in the README.

Licensing & Compatibility

The environment is licensed under Attribution-NonCommercial 4.0 International (CC BY-NC 4.0). This license prohibits commercial use and requires attribution.

Limitations & Caveats

The project is still under development, with features like dodges and a refined lateral friction model planned for future implementation. The current version relies on a specific release of the Unity ML-Agents Toolkit, which may impact compatibility with newer versions.