cleanrl

De novo cyclic protein polypeptide design using reinforcement learning.

🤖 MuJoCo-based 6-DOF reaching task - Deep RL for UR5 robotic arm control (Comparison of PPO, SAC, TD3, and DDPG parameter tuning and learning performance)

Reinforcement learning agent for Prince of Persia (1989) using PPO, FiGAR, and SDLPoP ENV.

Trained a custom hexapod robot to crawl using reinforcement learning

Polyvision is a reinforcement learning platform for training and benchmarking Polytopia agents through a headless Java engine connected to Python via Py4J. Its first phase targets economy-first play maximizing Stars Per Turn in 10-turn Bardur Drylands games, with a roadmap toward combat, live-game computer vision, and an open community leaderboard.

Experiments with hyper parameters

Important document for research using PPO. Includes working version with neuron-astrocyte model. These are modified files from the cleanrl repository.