Skip to content

Latest commit

 

History

History
36 lines (26 loc) · 1.81 KB

README.md

File metadata and controls

36 lines (26 loc) · 1.81 KB

SocNavBench

Welcome to the Social Navigation Benchmark utility (SocNavBench), a codebase for benchmarking robot planning algorithms against various episodes of containing multi-agent environments, accompanying out paper "SocNavBench: A Grounded Simulation Testing Framework for Evaluating Social Navigation".

Fig1

We provide scenarios curated from real world data for social navigation algorithms to be tested and evaluated on. sim2realZara sim2realUniv

We also provide multiple curated maps that closely resemble the environments for the pedestrian datasets. thri-maps

How To's

Install

Guide for installation at docs/install.md

Use

Guide for usage at docs/usage.md

Acknowledgements

This work was funded under grants from the National Science Foundation (NSF IIS-1734361) and the National Institute on Disability, Independent Living, and Rehabilitation Research (NIDILRR 90DPGE0003).

This project is also built upon the Human Active Navigation (HumANav) and Learning-Based Waypoint Navigation (Visual-Navigation) codebases. Special thanks to Varun Tolani for helping us with his projects.

Citation

If you use our work, please cite the corresponding work:

@article{biswas2021socnavbench,
  title={SocNavBench: A Grounded Simulation Testing Framework for Evaluating Social Navigation},
  author={Biswas, Abhijat and Wang, Allan and Silvera, Gustavo and Steinfeld, Aaron and Admoni, Henny},
  journal={arXiv preprint arXiv:2103.00047},
  year={2021}
}