roboserg/RoboLeague

A car soccer environment inspired by Rocket League for deep reinforcement learning experiments in an adversarial self-play setting.

What it solves

It provides a set of reinforcement learning (RL) environments for testing vehicle control, specifically focusing on hovering and aerial ball handling.

How it works

The project uses the Unity ML-Agents toolkit to create simulation scenes where agents can be trained to perform specific tasks. It includes pre-defined scenes for basic ML tests, calm hovering, aerial dribbling, and maximizing speed with a constant roll.

Who it’s for

Developers and researchers interested in training machine learning agents for vehicle dynamics and aerial maneuvers within the Unity engine.

Highlights

  • Training environments for aerial ball handling (AirDribble).
  • Dedicated scenes for hovering and speed tests.
  • Support for trained ONNX models.
  • Built using Unity ML-Agents Release 11.

Related

  • Project
  • Project
  • Project
  • Project