SOCIALGYM 2.0: Simulator for Multi-Agent Social Robot Navigation in Shared Human Spaces

Zayne Sprague, Rohan Chandra, Jarrett Holtz, Joydeep Biswas

Abstract

We present SocialGym 2.0, a multi-agent navigation simulator for social robot research. Our simulator models multiple autonomous agents, replicating real-world dynamics in complex environments, including doorways, hallways, intersections, and roundabouts. Unlike traditional simulators that concentrate on single robots with basic kinematic constraints in open spaces, SocialGym 2.0 employs multi-agent reinforcement learning (MARL) to develop optimal navigation policies for multiple robots with diverse, dynamic constraints in complex environments. Built on the PettingZoo MARL library and Stable Baselines3 API, SocialGym 2.0 offers an accessible python interface that integrates with a navigation stack through ROS messaging. SocialGym 2.0 can be easily installed and is packaged in a docker container, and it provides the capability to swap and evaluate different MARL algorithms, as well as customize observation and reward functions. We also provide scripts to allow users to create their own environments and have conducted benchmarks using various social navigation algorithms, reporting a broad range of social navigation metrics.

Project

Paper. This paper is submitted to RSS 2023.
Project. GitHub repository
Documentation

If you use our work, please cite the following

@article{sprague2023socialgym,
  title={SOCIALGYM 2.0: Simulator for Multi-Agent Social Robot Navigation in Shared Human Spaces},
  author={Sprague, Zayne and Chandra, Rohan and Holtz, Jarrett and Biswas, Joydeep},
  journal={arXiv preprint arXiv:2303.05584},
  year={2023}
}

@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}
}