LiteRL

Open-source software · Python · MIT license

MushroomRL

A Python library that makes reinforcement learning experiments simple, modular and reproducible. Write an experiment in a few lines, swap algorithms and environments freely, and spend your time on the research question rather than on boilerplate.

Install with pip install mushroom-rl

MushroomRL logo

Why MushroomRL

Built for reinforcement learning research

Carlo D'Eramo and Davide Tateo started MushroomRL to give researchers one consistent framework for their experiments. Every agent talks to every environment through the same interface, so the same few lines of code run tabular Q-learning on a grid world or deep actor-critic methods on a MuJoCo robot.

Modular

Agents, policies, approximators and environments are independent building blocks. Combine them, or replace one with your own, without touching the rest.

Batteries included

Classical algorithms such as Q-learning, SARSA and FQI sit next to deep RL methods such as DQN, DDPG, TD3, SAC, TRPO and PPO, plus policy-search methods.

Many benchmarks

Use Gymnasium, Atari, MuJoCo, PyBullet, the DeepMind Control Suite or NVIDIA Isaac Sim, as well as a set of classic control problems built into the library.

Framework-friendly

Bring the tensor library you already use, such as PyTorch, for function approximation. MushroomRL handles the experiment loop around it.

Paper

Cite MushroomRL

If MushroomRL helps your research, please cite our JMLR paper: Carlo D'Eramo, Davide Tateo, Andrea Bonarini, Marcello Restelli and Jan Peters. MushroomRL: Simplifying Reinforcement Learning Research. Journal of Machine Learning Research, 22(131):1–5, 2021.

@article{JMLR:v22:18-056,
  author  = {Carlo D'Eramo and Davide Tateo and Andrea Bonarini and Marcello Restelli and Jan Peters},
  title   = {MushroomRL: Simplifying Reinforcement Learning Research},
  journal = {Journal of Machine Learning Research},
  year    = {2021},
  volume  = {22},
  number  = {131},
  pages   = {1-5},
  url     = {http://jmlr.org/papers/v22/18-056.html}
}

Run your first experiment

The documentation has installation notes, tutorials and the full API reference. The repository has example scripts, including experiments that reproduce published results. Issues and pull requests are welcome.