Patrolling Zoo

This repository contains the policy/training code for the paper, "Graph Neural Network-based Multi-agent Reinforcement Learning for Resilient Distributed Coordination of Multi-Robot Systems", by Anthony Goeckner, Yueyuan Sui, Nicolas Martinet, Xinliang Li, and Qi Zhu of Northwestern University in Evanston, Illinois.

If you use this code, please cite our paper as:

@INPROCEEDINGS{10802510,
  author={Goeckner, Anthony and Sui, Yueyuan and Martinet, Nicolas and Li, Xinliang and Zhu, Qi},
  booktitle={2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)}, 
  title={Graph Neural Network-based Multi-agent Reinforcement Learning for Resilient Distributed Coordination of Multi-Robot Systems}, 
  year={2024},
  volume={},
  number={},
  pages={5732-5739},
  doi={10.1109/IROS58592.2024.10802510}}

Package Description

Packages are as follows:

• onpolicy: Contains the algorithm code.
• patrolling_zoo: Contains the environment code.

Installation

1. Clone the patrolling_zoo repository:

   git clone --recurse git@github.com:NU-IDEAS-Lab/patrolling_zoo.git

2. Create a Conda environment with required packages:

   cd ./patrolling_zoo
   conda env create -n patrolling_zoo -f ./environment.yml
   conda activate patrolling_zoo

3. Install PyTorch to the new patrolling_zoo conda environment using the steps outlined on the PyTorch website.

4. Install the onpolicy and patrolling_zoo packages:

   pip install -e .
   

Operation

You may run the example in onpolicy/scripts/train_patrolling_scripts/mappo.ipynb.