Skip to content

SocialGym 2: A lightweight benchmark and simulator for multi-robot social navigation using ROS and the OpenAI gym.

License

Notifications You must be signed in to change notification settings

ut-amrl/SocialGym2

Repository files navigation

This is the codebase for our multi-agent simulator for real world social navigation. Installation instructions ca be found below. Tutorials are provided in the documentation (currently under active development).

Read the paper

SOCIALGYM 2.0: Simulator for Multi-Agent Social Robot Navigation in Shared Human Spaces, AAAI 2024.

Installation

Prerequisites

You need to have installed Docker and have a GPU.

WARNING: The authors of this project have had numerous issues with Docker and Docker-Compose when installed via snap. We recommend uninstalling docker and docker-compose if you have GPU-related issues with the docker images and installing them via apt.


1.) Clone the repo and checkout main

git clone git@github.com:ut-amrl/SocialGym2.git
git checkout main

2.) Install requirements and run the install script

python3 ./scripts/install_config_runner.py

3.) Run the Config Runner!

python3 config_runner/run.py -c 1_31_23/door/sacadrl.json

This should open an RVIS window (looks like a 2D grid with options on the side-panels). If you do NOT see this window, there's a problem with the docker file. First try running

sudo xhost +

and repeat the command. If it still fails to show windows then try adding --network host to the file {PROJECT_ROOT}/config_runner/run.sh on lines where we are setting docker environment variables. Otherwise, look up stuff like "cannot display window from docker container" -- this is critical for Social Gym to work.

3.) Have fun!

If you use this code, please cite the following


@inproceedings{chandra2024socialgym,
  title={SOCIALGYM 2.0: Simulator for Multi-Robot Learning and Navigation in Shared Human Spaces},
  author={Chandra, Rohan and Sprague, Zayne and Biswas, Joydeep},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={21},
  pages={23778--23780},
  year={2024}
}
@inproceedings{holtz2022socialgym,
  title={Socialgym: A framework for benchmarking social robot navigation},
  author={Holtz, Jarrett and Biswas, Joydeep},
  booktitle={2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={11246--11252},
  year={2022},
  organization={IEEE}
}

License


This project is released under the MIT License. Please review the License file for more details.