SHARP: Shielding-Aware Robust Planning

SHARP

SHARP: Shielding-Aware Robust Planning for Safe and Efficient Human-Robot Interaction

Table of Contents

1. About The Project
2. Dependencies
3. Example
4. Dataset
5. License
6. Contact
7. Paper
8. Acknowledgements

About The Project

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:

Dependencies

Trajectory Optimization

• MPT3 (Toolbox for MPC and parametric optimization)
• MOSEK (Quadratic programming solver. Alternatively, you may consider MATLAB's default quadprog)

Shielding

• Level Set Toolbox (Toolbox for solving HJ PDE)
• helperOC (Toolbox for HJ-based shielding)

Visualization

• Robotics Toolbox for MATLAB (Tools for plotting the vehicles)

Example

In this repository, we provide an example of SHARP applied for human-robot interactive driving scenarios.

Quickstart

1. Clone the repo

   git clone https://github.com/SafeRoboticsLab/SHARP.git

2. Install all dependencies.
3. Under the root directory of Robotics Toolbox for MATLAB, replace plot_vehicle.m with ours.
4. Merge helperOC with ours, which contains the customized dynamics and shielding policy.
5. In MATLAB, run main.m to reproduce our results.
6. (Optional) You may change the problem specifications and planner parameters in here.

Dataset

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.

License

Distributed under the BSD 3-Clause License. See LICENSE for more information.

Contact

Haimin Hu - @HaiminHu - [email protected]

Project Link: https://github.com/SafeRoboticsLab/SHARP

Homepage Link: https://haiminhu.org/research/sharp

Paper

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}
}

Acknowledgements

• This research is supported by the Princeton Project X Program.
• We use the human driver's trajectories from the Waymo Open Motion Dataset.