abmarl 0.1.3

Agent Based Simulation and MultiAgent Reinforcement Learning

Abmarl

Abmarl is a package for developing Agent-Based Simulations and training them with MultiAgent Reinforcement Learning (MARL). We provide an intuitive command line interface for engaging with the full workflow of MARL experimentation: training, visualizing, and analyzing agent behavior. We define an Agent-Based Simulation Interface and Simulation Manager, which control which agents interact with the simulation at each step. We support integration with popular reinforcement learning simulation interfaces, including gym.Env and MultiAgentEnv.

Abmarl leverages RLlib’s framework for reinforcement learning and extends it to more easily support custom simulations, algorithms, and policies. We enable researchers to rapidly prototype MARL experiments and simulation design and lower the barrier for pre-existing projects to prototype RL as a potential solution.

Quickstart

To use Abmarl, install via pip: pip install abmarl

To develop Abmarl, clone the repository and install via pip's development mode. Note: Abmarl requires python3.7 or python3.8.

git clone [e-mail]:LLNL/Abmarl.git
cd abmarl
pip install -r requirements.txt
pip install -e . --no-deps

Train agents in a multicorridor simulation:

abmarl train examples/multi_corridor_example.py

Visualize trained behavior:

abmarl visualize ~/abmarl_results/MultiCorridor-2020-08-25_09-30/ -n 5 --record

Note: If you install with conda, then you must also include ffmpeg in your virtual environment.

Documentation

You can find the latest Abmarl documentation, on our ReadTheDocs page.

Community

Reporting Issues

Please use our issue tracker to report any bugs or submit feature requests. Great bug reports tend to have:

• A quick summary and/or background
• Steps to reproduce, sample code is best.
• What you expected would happen
• What actually happens

Contributing

Please submit contributions via pull requests from a forked repository. Find out more about this process here. All contributions are under the BSD 3 License that covers the project.

Additional support

• Edward Rusu, [e-mail]
• Ruben Glatt, [e-mail]

Release

LLNL-CODE-815883