torchrl

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

Modular reinforcement learning framework for algorithmic trading

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

RouteRL is a multi-agent reinforcement learning framework for modeling and simulating the collective route choices of humans and autonomous vehicles.

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.

Multi-Agent Reinforcement Learning for Level-Based Foraging Using TorchRL.

reinforchy is the implementation of some Reinforcement Learning algorithms

MARL research project, where agents collaboratively self-organize inside dynamically generated geometric patterns.

MARL research project in which rescuer and rescuee agents collaborate to navigate and succeed in complex, obstacle-rich environments.

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

MARL research project, based on the famous board game "Scotland Yard".

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

MARL research project, where multiple agents (ants) interact in a shared 2D environment containing scattered items of different categories.

Using Reinforcement Learning to play Dark Souls III

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.

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

Train an RL agent to catch a descending rocket with mechanical chopstick arms, Mechazilla style.

MARL research project exploring the VMAS Soccer scenario, a continuous-time, physics-based multi-agent environment where two teams of agents compete to score goals.

Inspired by the series "Squid Game", this project requires agents to learn coordinated decision-making and spatial negotiation in a competitive-cooperative setting.