MushroomRL/mushroom-rl

What it solves

MushroomRL simplifies the process of conducting reinforcement learning (RL) experiments by providing a modular framework that integrates various tensor computation libraries and RL benchmarks.

How it works

It acts as a modular library that allows users to combine different RL algorithms with various backends and environments. It supports both classical RL algorithms (such as Q-Learning, SARSA, and FQI) and deep RL algorithms (including DQN, DDPG, SAC, TD3, TRPO, and PPO). The library is compatible with tensor computation tools like PyTorch and TensorFlow, and integrates with benchmarks such as Gymnasium, PyBullet, and the Deepmind Control Suite.

Who it’s for

Researchers and developers who need a flexible, modular way to implement and test reinforcement learning algorithms across different environments and simulators.

Highlights

• Modular design for easy integration with PyTorch and TensorFlow.
• Support for a wide range of classical and deep RL algorithms.
• Integration with popular RL benchmarks including Gymnasium, PyBullet, and MuJoCo.
• Comprehensive documentation and tutorials available.