tmrl-test

Self-driving default benchmark of the TMRL open-source project.

TMRL is a Deep Reinforcement Learning framework developed at Polytechnique Montreal. It facilitates training Artificial Intelligences in real-time applications (robots, video-games...).

In particular, researchers and AI enthusiasts over the world use TMRL to train self-racing AIs in TrackMania 2020, because the framework comes with a training pipeline for this video game that is readily implemented and easy to use.

TMRL hosts the TrackMania Roborace League , an AI competition currently in open beta-test, where self-racing policies are evaluated on the tmrl-test track.

You may also have seen Gwen and Affi, winners of the TM-Cup 2022, race against (i.e.,destroy 😛 ) one of our baseline AIs on the tmrl-test track in the Underscore_ french talk-show.