TrackMania RL - Documentation

Welcome to the TrackMania RL project documentation!

This is a fork and extension of the original Linesight project, adapted for reinforcement learning experiments in Trackmania Nations Forever.

The project uses distributional reinforcement learning (IQN - Implicit Quantile Networks) to train an AI agent to drive in TrackMania. The goal is to explore RL algorithms, reward shaping, and training techniques in a complex racing environment.

Key Features:

  • Distributional RL with IQN (Implicit Quantile Network)

  • Modular configuration system for easy experimentation

  • Support for multiple parallel game instances

  • Hot-reloadable training parameters

  • TensorBoard integration for monitoring

  • Virtual checkpoint system for dense progress tracking

All runs produced by this project are Tool Assisted. They must not be submitted to the Official Leaderboards.

User Documentation:

Dev Documentation:

Experiments:

Community tips & tricks