POGEMA Benchmark

Welcome to the official repository for the POGEMA Benchmark. This is an umbrella repository that contains links and information about all the tools and algorithms related to the POGEMA Benchmark.

Repository Overview

This repository is organized into several key areas:

• Raw Data Directories: Contains evaluation results for baseline approaches. Details include:
  • raw_data_LMAPF: LMAPF challenges, including configurations and maps.
  • raw_data_MAPF: MAPF challenges, including configurations and maps.
  • Both directories feature YAML configuration files detailing the evaluation settings (number of agents, maps, seeds, episode length) and a maps.yaml file listing all the maps used in the evaluations.

Installation

POGEMA Environment

Install the POGEMA environment using pip:

pip3 install pogema

For more details, visit the POGEMA GitHub repository.

POGEMA Toolbox

Install additional tools for POGEMA:

pip3 install pogema-toolbox

Algorithms

Explore integrated algorithms located in the algorithms directory:

• Navigate to the directory using:

  cd algorithms

• Install necessary dependencies:

  pip3 install -r docker/requirements.txt

• Optionally, build a Docker image to containerize the environment:

  cd docker && sh build.sh

  MAMBA baseline requires a separate Docker image:

  cd docker/mamba && sh build.sh

The following table contains links to the original repositories of all the integrated approaches:

Approach	Link
DCC	https://github.com/ZiyuanMa/DCC
Follower	https://github.com/AIRI-Institute/learn-to-follow
LaCAM	https://github.com/Kei18/lacam3
MATS-LP	https://github.com/AIRI-Institute/mats-lp
RHCR	https://github.com/Jiaoyang-Li/RHCR
SCRIMP	https://github.com/marmotlab/SCRIMP
MAMBA	https://github.com/jbr-ai-labs/mamba

Evaluation

Execute the evaluation script:

python eval.py

Contents at a Glance

.
├── algorithms
│   ├── Multiple algorithms for benchmarking
│   ├── Docker configuration for container setup
│   ├── eval.py for running evaluations
├── raw_data_LMAPF
│   ├── Data categorized by map types: Random, Mazes, Warehouse, etc.
├── raw_data_MAPF
│   ├── Similar categorization with specific map evaluations
└── README.md

Citation

If you use this repository in your research or wish to cite it, please make a reference to our paper:

@misc{skrynnik2024pogema,
      title={POGEMA: A Benchmark Platform for Cooperative Multi-Agent Navigation}, 
      author={Alexey Skrynnik and Anton Andreychuk and Anatolii Borzilov and Alexander Chernyavskiy and Konstantin Yakovlev and Aleksandr Panov},
      year={2024},
      eprint={2407.14931},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2407.14931}, 
}