A PyTorch library for all things Reinforcement Learning (RL) for Combinatorial Optimization (CO)

BricksRL: A Platform for Democratizing Robotics and Reinforcement Learning Research and Education with LEGO

This project is a pipeline that connects a Matlab simulation (Simulink) to an OpenAI Gym wrapper for PyTorch Reinforcement Learning using the DQN algorithm.

An adaption of the Flatland environment for TorchRL.

A highly modular and extensible PyTorch-based reinforcement learning library.

Small prototype to show RoboHive usage with TorchRL for visual deep reinforcement learning

Small prototype to show RLBench usage with TorchRL

Clone of the Mars: Mars game using Pygame as a Reinforcement Learning Gym 🚀

Training a PPO agent to play chess with pretraining and self-learning using PyTorch Lightning and TorchRL

🤖 PPO and DQN reinforcement learning algorithms implemented with PyTorch and TorchRL for autonomous driving.

This project aims to improve the efficiency of online reinforcement learning by incorporating bisimulation-based metrics into the experience replay process. Bisimulation Prioritized Experience Replay (BPER) prioritizes experiences based on their behavioral differences.

Application of deep reinforcement learning (DQN and PPO) for automated trading on HPC system, comparing performance across CPU/GPU nodes