SHARP: Shielding-Aware Robust Planning for Safe and Efficient Human-Robot Interaction
SHARP: Shielding-Aware Robust Planning is a general framework for safe and efficient human-robot interaction. We provide a MATLAB implementation of SHARP for autonomous driving applications, which can be found here.
The Python implementation is being actively developed. An iLQR-based shielding example can be found here.
Click to watch our spotlight video:
MPT3
(Toolbox for MPC and parametric optimization)MOSEK
(Quadratic programming solver. Alternatively, you may consider MATLAB's defaultquadprog
)
Level Set Toolbox
(Toolbox for solving HJ PDE)helperOC
(Toolbox for HJ-based shielding)
Robotics Toolbox for MATLAB
(Tools for plotting the vehicles)
In this repository, we provide an example of SHARP applied for human-robot interactive driving scenarios.
Quickstart
- Clone the repo
git clone https://github.com/SafeRoboticsLab/SHARP.git
- Install all dependencies.
- Under the root directory of
Robotics Toolbox for MATLAB
, replaceplot_vehicle.m
with ours. - Merge
helperOC
with ours, which contains the customized dynamics and shielding policy. - In MATLAB, run
main.m
to reproduce our results. - (Optional) You may change the problem specifications and planner parameters in here.
We use the human driver's trajectories from the Waymo Open Motion Dataset. In particular, we filtered out 50 representative highway overtaking scenarios from the original dataset. Raw data with filtered trajectories in npy
format can be found here. Trajectories converted into MATLAB's cell
format can be found here.
Distributed under the BSD 3-Clause License. See LICENSE
for more information.
Haimin Hu - @HaiminHu - haiminh@princeton.edu
Project Link: https://github.com/SafeRoboticsLab/SHARP
Homepage Link: https://haiminhu.org/research/sharp
IEEE Xplore: https://ieeexplore.ieee.org/document/9723544
arXiv: https://arxiv.org/abs/2110.00843
@article{hu2022sharp,
author={Hu, Haimin and Nakamura, Kensuke and Fisac, Jaime F.},
journal={IEEE Robotics and Automation Letters},
title={SHARP: Shielding-Aware Robust Planning for Safe and Efficient Human-Robot Interaction},
year={2022},
volume={7},
number={2},
pages={5591-5598},
doi={10.1109/LRA.2022.3155229}
}
Our follow-up paper:
Available on arXiv: https://arxiv.org/abs/2202.07720
@inproceedings{hu2023active,
title={Active Uncertainty Reduction for Human-Robot Interaction: An Implicit Dual Control Approach},
author={Hu, Haimin and Fisac, Jaime F},
booktitle={Algorithmic Foundations of Robotics XV},
pages={385--401},
year={2023},
publisher={Springer International Publishing}
}
- This research is supported by the Princeton Project X Program.
- We use the human driver's trajectories from the Waymo Open Motion Dataset.