Flatland is an open-source toolkit for developing and comparing Multi-Agent Reinforcement Learning algorithms in little (or ridiculously large!) gridworlds.
The official website contains full details about the environment and problem statement.
Flatland is tested with Python 3.10, 3.11, and 3.12 on modern versions of macOS, Linux and Windows. You may encounter problems with graphical rendering if you use WSL.
This library was developed specifically for the AIcrowd Flatland challenges in which we strongly encourage you to take part in!
Set up a virtual environment using your preferred method (we suggest the built-in venv) and activate it.
You can use your IDE to do this or by using the command line:
python -m venv .venv
source .venv/bin/activateInstall Flatland using pip:
python -m pip install flatland-rlThis is the preferred method to install Flatland, as it will always install the most recent stable release.
This library was developed by SBB, Deutsche Bahn, SNCF, AIcrowd and numerous contributors from the flatland community.
Please follow the Contribution Guidelines for more details on how you can successfully contribute to the project. We enthusiastically look forward to your contributions!
