RL Racing

2021 // Solo Project @HSLU // JS


Play Back

RL (Reinforcement Learning) Racing was created as part of the fifth semester module "Computational Perception extended". Several agents attempt to drive their "car" as far as possible within a given time. The agents sense the track limits with 3 ray casts and respond with a set of reactions consisting of steering, acceleration and breaking. After each generation, the best driver is selected and slightly mutated. The next generations is then populated with these mutations.

The source code can be found on GitHub.