Implementation of the reinforcement learning algorithm Q-Learning.
We use the OpenAI Gym environment MountainCar-v0.
The latter has a continuous state space of 2 features which are the car position and velocity.
Its action space is discrete and is composed of 3 actions: "push left", "no push", "push right".
Watch our video to see the results of the experiment on 5000 episodes.
Côme Cothenet, Kathryn Schutte, Guillaume Fradet