Safe Autonomous Racing using ILQR and Rollout-based Shielding: A JAX Implementation

Table of Contents

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

About The Project

This repository implements a safe autonomous racing example using ILQR and rollout-based shielding, which relies on JAX for real-time computation performance based on automatic differentiation and just-in-time (JIT) compilation. The repo is primarily developed and maintained by Haimin Hu, a PhD student in the Safe Robotics Lab. Zixu Zhang, Kai-Chieh Hsu and Duy Nguyen also contributed much to the repo (their original repo is here).

Dependencies

This repo depends on the following packages:

1. jax=0.4.19
2. jaxlib=0.4.16
3. matplotlib=3.5.1
4. numpy=1.21.5
5. pyspline=1.5.1
6. python=3.10
7. yaml=0.2.5

Example

Please refer to the Colab Notebook for a demo usage of this repo.

Alternatively, you can directly run the main script:

python3 example_racecar.py

License

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

Contact

Haimin Hu - @HaiminHu - haiminh@princeton.edu

Paper

If you found this repository helpful, please consider citing one of the following papers that turned this repository into research contributions.

• ILQR-based safety filter:

@article{hu2023active,
  title={Active uncertainty reduction for safe and efficient interaction planning: A shielding-aware dual control approach},
  author={Hu, Haimin and Isele, David and Bae, Sangjae and Fisac, Jaime F},
  journal={The International Journal of Robotics Research},
  pages={02783649231215371},
  year={2023},
  publisher={SAGE Publications Sage UK: London, England}
}

• ILQR-based Stackelberg game solver:

@inproceedings{hu2024plays,
  title={Who Plays First? Optimizing the Order of Play in Stackelberg Games with Many Robots},
  author={Hu, Haimin and Dragotto, Gabriele and Zhang, Zixu and Liang, Kaiqu and Stellato, Bartolomeo and Fisac, Jaime F},
  booktitle={Proceedings of Robotics: Science and Systems},
  year={2024}
}

• ILQ-based game solver for data-aware planning:

@inproceedings{lidard2024blending,
  title={Blending Data-Driven Priors in Dynamic Games},
  author={Lidard, Justin and Hu, Haimin and Hancock, Asher and Zhang, Zixu and Contreras, Albert Gim{\'o} and Modi, Vikash and DeCastro, Jonathan and Gopinath, Deepak and Rosman, Guy and Leonard, Naomi and others},
  booktitle={Proceedings of Robotics: Science and Systems},
  year={2024}
}

Acknowledgements

This repo is inspired by the following projects:

• Princeton Race Car
• Ellipsoidal Toolbox (ET)
• ellReach: Ellipsoidal Reachable Set Computation for Linear Time-Varying Systems (under development)