This repository contains two reinforcement learning algorithms, Soft Actor-Critic (SAC) and Proximal Policy Optimization(PPO). We evaluate their performance on a 2D hockey environment, where two agents control paddles to hit a puck into the opponent’s goal. The observation space is 18 dimensional and includes positions, velocities, and angular movements of both players and the puck. The hockey environment supports continuous and discrete action spaces, where agents can either output real-valued forces for movement, rotation, and shooting or select from predefined discrete actions mapped to equiv- alent movements.
