BRNE: Bayesian Recursive Nash Equilibrium for Social Navigation

This repository contains relevant code for BRNE (Bayesian Recursive Nash Equilibrum), an algorithm that computes mixed strategy Nash equilibrium for real-time human-aware navigation in dense crowds. It contains PyTorch-based tutorial, a fully ROS integrated social navigation framework powered by BRNE, implemented in both PyTorch and C++, as well as implementations for simulated benchmark studies.

Video demonstrations of the algorithm deployed on an Unitree Go1 quadruped can be found at the project website and Katie's portfolio post. You can also find more details on the project from our invited presentation at NVIDIA GTC 2024.

Details of the algorithm can be found in our IJRR paper. To cite this work:

@article{doi:10.1177/02783649241302342,
author = {Max Muchen Sun and Francesca Baldini and Katie Hughes and Peter Trautman and Todd Murphey},
title ={Mixed strategy Nash equilibrium for crowd navigation},
journal = {The International Journal of Robotics Research},
year = {2024},
doi = {10.1177/02783649241302342},
URL = {https://doi.org/10.1177/02783649241302342},
eprint = {https://doi.org/10.1177/02783649241302342}}

The structure of the repo is as follow:

• tutorials contains Python tutorials to illustrate the principle of the BRNE algorithm and intuitive PyTorch implementation for simple hallway navigation scenarios.
• brne_nav contains the fully integrated ROS social navigation framework. If you are interested in deploying BRNE, you should copy this directory as a ROS package.
• socnavbench contains the implementation and installation instructions for the SocNavBench benchmark study.

Contact: Please contact Max ([email protected]) if you have any question.