Air Hockey Agent - Reinforcement Learning Project

As part of the course “Reinforcement Learning” at the University of Tübingen, I have implemented an agent for the Air Hockey environment using Pygame and OpenAI Gym. The environment simulates a simple air hockey game where an agent controls a paddle to hit a puck towards the opponent’s goal while defending its own goal.

I implemented and trained an agent using the TD3 (Twin Delayed Deep Deterministic Policy Gradient) algorithm, which then took part in the final competition of the course against other students’ agents.