LunarLander — Reinforcement Learning
An experimental Reinforcement Learning project focused on training an AI agent to land a rocket safely and accurately using Gymnasium's LunarLander-v3 environment.
The project involved experimenting with reward engineering, developing a custom reward wrapper, and comparing five reward configurations across three random seeds, with one million training steps per configuration.
Through iterative testing and optimization, I refined the reward system to encourage smoother, more accurate landings while avoiding unwanted behaviors such as indefinite hovering.
The final agent achieved a 100% landi
Machine LearningPython
Pièces jointes
G
Gheilen D.
Voir le profil du freelanceCreative Director & Media Maker | Entrepreneur | Technology • Games • Film • Digital Media • Design






