Social Behavior as a Key to Learning-based Multi-Agent Pathfinding Dilemmas

This repository is the official implementation of Social Behavior as a Key to Learning-based Multi-Agent Pathfinding Dilemmas. The paper is currently under review.

Environment Configuration

We provide the MAPF environment configuration we use (based on anaconda) and package it as a MAPF.yml file. You can load it with the following command:

conda env create -f MAPF.yml

Start Training

Once we have confirmed that the environment has been configured, we can activate it with the following command:

conda activate MAPF

Then, we can run the main program to start training:

python driver.py

During the training process, the models and animated images will be stored in /models and /gifs respectively at fixed intervals.

Tracking training

We provide an interface to track training using wandb, you can do this by setting WANDB to True in alg_parameters.py. In addition, you need to modify the following three parameters to the content of your own account correspondingly:

ENTITY = 'full_blank_1'
EXPERIMENT_PROJECT = 'full_blank_2'
EXPERIMENT_NAME = 'full_blank_3'

Testing

If you want to run the test cases on our provided test data, you can run:

python run_the_instances.py

before run it, make sure you have activated the MAPF environment already. Additionally, during the testing phase, we recommend changing the EPISODE_LEN parameter in alg_parameters.py from 256 to 512.

EPISODE_LEN from 256 -> 512

Reference

If this repository is helpful to you, please cite our work by:

@article{he2024social,
  title={Social Behavior as a Key to Learning-based Multi-Agent Pathfinding Dilemmas},
  author={He, Chengyang and Duhan, Tanishq and Tulsyan, Parth and Kim, Patrick and Sartoretti, Guillaume},
  journal={arXiv preprint arXiv:2408.03063},
  year={2024}
}