Awesome Open Source
Awesome Open Source


Continuous Integration Documentation Status Maintainability Test Coverage

MushroomRL: Reinforcement Learning Python library.

What is MushroomRL

MushroomRL is a Python Reinforcement Learning (RL) library whose modularity allows to easily use well-known Python libraries for tensor computation (e.g. PyTorch, Tensorflow) and RL benchmarks (e.g. OpenAI Gym, PyBullet, Deepmind Control Suite). It allows to perform RL experiments in a simple way providing classical RL algorithms (e.g. Q-Learning, SARSA, FQI), and deep RL algorithms (e.g. DQN, DDPG, SAC, TD3, TRPO, PPO).

Full documentation and tutorials available here.


You can do a minimal installation of MushroomRL with:

pip3 install mushroom_rl

Installing everything

To install the whole set of features, you will need additional packages installed. You can install everything by running:

pip3 install mushroom_rl[all]

This will install every dependency of MushroomRL, except MuJoCo and Plots dependencies. For ubuntu>20.04, you may need to install pygame and gym dependencies:

sudo apt -y install libsdl-image1.2-dev libsdl-mixer1.2-dev libsdl-ttf2.0-dev \
                 libsdl1.2-dev libsmpeg-dev libportmidi-dev ffmpeg libswscale-dev \
                 libavformat-dev libavcodec-dev swig

To use the mujoco-py MushroomRL interface you can run the command:

pip3 install mushroom_rl[mujoco]

Below is the code that you need to run to install the Plots dependencies:

sudo apt -y install python3-pyqt5
pip3 install mushroom_rl[plots]

You might need to install external dependencies first. For more information about mujoco-py installation follow the instructions on the project page

To use dm_control MushroomRL interface, install dm_control following the instruction that can be found here

Using Habitat and iGibson with MushroomRL

Habitat and iGibson are simulation platforms providing realistic and sensory-rich learning environments. In MushroomRL, the agent's default observations are RGB images, but RGBD, agent sensory data, and other information can also be used.

If you have previous versions of iGibson or Habitat already installed, we recommend to remove them and do clean installs.

iGibson Installation

Follow the official guide and install its assets and datasets.

For <MUSHROOM_RL PATH>/mushroom-rl/examples/ you need to run

python -m igibson.utils.assets_utils --download_assets
python -m igibson.utils.assets_utils --download_demo_data
python -m igibson.utils.assets_utils --download_ig_dataset

You can also use third party datasets.

The scene details are defined in a YAML file, that needs to be passed to the agent. See <IGIBSON PATH>/igibson/test/test_house.YAML for an example.

Habitat Installation

Follow the official guide and do a full install with habitat_baselines. Then you can download interactive datasets following this and this. If you need to download other datasets, you can use this utility.

Basic Usage of Habitat

When you create a Habitat environment, you need to pass a wrapper name and two YAML files: Habitat(wrapper, config_file, base_config_file).

  • The wrapper has to be among the ones defined in <MUSHROOM_RL PATH>/mushroom-rl/environments/, and takes care of converting actions and observations in a gym-like format. If your task / robot requires it, you may need to define new wrappers.

  • The YAML files define every detail: the Habitat environment, the scene, the sensors available to the robot, the rewards, the action discretization, and any additional information you may need. The second YAML file is optional, and overwrites whatever was already defined in the first YAML.

    If you use YAMLs from habitat-lab, check if they define a YAML for BASE_TASK_CONFIG_PATH. If they do, you need to pass it as base_config_file to Habitat(). habitat-lab YAMLs, in fact, use relative paths, and calling them from outside its root folder will cause errors.

  • If you use a dataset, be sure that the path defined in the YAML file is correct, especially if you use relative paths. habitat-lab YAMLs use relative paths, so be careful with that. By default, the path defined in the YAML file will be relative to where you launched the python code. See the navigation example below for more details.

Rearrange Task Example

  • Download assets and the ReplicaCAD datasets
python -m habitat_sim.utils.datasets_download --uids habitat_test_pointnav_dataset --data-path data
python -m habitat_sim.utils.datasets_download --uids habitat_test_scenes --data-path data
  • For this task we use <HABITAT_LAB PATH>/habitat_baselines/config/rearrange/ddppo_rearrangepick.yaml. This YAML defines BASE_TASK_CONFIG_PATH: configs/tasks/rearrangepick_replica_cad.yaml, and since this is a relative path we need to overwrite it by passing its absolute path as base_config_file argument to Habitat().
  • Then, rearrangepick_replica_cad.yaml defines the dataset to be used, and this is in <HABITAT_LAB PATH>. However, since the path defined is relative to where we launch our code, we need to make a link to the data folder. If you launch from its example folder, run
ln -s <HABITAT_LAB PATH>/data/ <MUSHROOM_RL PATH>/mushroom-rl/examples/habitat
  • Finally, you can launch python

Navigation Task Example

  • Download and extract Replica scenes

    WARNING! The dataset is very large!

sudo apt-get install pigz
git clone
cd Replica-Dataset
./ replica-path
  • For this task we only use the custom YAML file pointnav_apartment-0.yaml.

  • DATA_PATH: "replica_{split}_apartment-0.json.gz" defines the JSON file with some scene details, such as the agent's initial position and orientation. The {split} value is defined in the SPLIT key.

    If you want to try new positions, you can sample some from the set of the scene's navigable points. After initializing a habitat environment, for example mdp = Habitat(...), run mdp.env._env._sim.sample_navigable_point().

  • SCENES_DIR: "Replica-Dataset/replica-path/apartment_0" defines the scene. As said before, this path is relative to where you launch the script, thus we need to link the Replica folder. If you launch from its example folder, run

ln -s <PATH TO>/Replica-Dataset/ <MUSHROOM_RL PATH>/mushroom-rl/examples/habitat
  • Finally, you can launch python

Editable Installation

You can also perform a local editable installation by using:

pip install --no-use-pep517 -e .

To install also optional dependencies:

pip install --no-use-pep517 -e .[all]

How to set and run and experiment

To run experiments, MushroomRL requires a script file that provides the necessary information for the experiment. Follow the scripts in the "examples" folder to have an idea of how an experiment can be run.

For instance, to run a quick experiment with one of the provided example scripts, run:

python3 examples/

Cite MushroomRL

If you are using MushroomRL for your scientific publications, please cite:

    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     = {}

How to contact us

For any question, drop an e-mail at [email protected].

Follow us on Twitter @Mushroom_RL!

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
Python (1,118,862
Deep Learning (22,978
Pytorch (11,194
Reinforcement Learning (3,862
Deep Reinforcement Learning (1,042
Openai Gym (540
Dqn (439
Rl (233
Atari (206
Ddpg (196
Related Projects