Learning for Layered Safety-Critical Control with Predictive Control Barrier Functions
Abstract
Safety filters leveraging control barrier functions (CBFs) are highly effective for enforcing safe behavior on complex systems. It is often easier to synthesize these for Reduced order Models (RoMs), and then track the resulting safe control input on the Full order Model (FoM)—yet gaps between the RoM and FoM can result in safety violations. This paper introduces \emph{predictive CBFs} to address this gap: leveraging rollouts of the FoM to define a predictive robustness term added to the CBF condition. Theoretically, we prove that this guarantees safety in a layered control implementation. Practically, we learn the predictive robustness term through massive parallel simulation with domain randomization. We demonstrate in simulation that this yields safe behavior on the FoM with minimal conservatism, and experimentally realize predictive CBFs on a 3D hopping robot.
Cite
Text
Compton et al. "Learning for Layered Safety-Critical Control with Predictive Control Barrier Functions." Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, 2025.Markdown
[Compton et al. "Learning for Layered Safety-Critical Control with Predictive Control Barrier Functions." Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, 2025.](https://mlanthology.org/l4dc/2025/compton2025l4dc-learning/)BibTeX
@inproceedings{compton2025l4dc-learning,
title = {{Learning for Layered Safety-Critical Control with Predictive Control Barrier Functions}},
author = {Compton, William D. and Cohen, Max H. and Ames, Aaron D.},
booktitle = {Proceedings of the 7th Annual Learning for Dynamics \& Control Conference},
year = {2025},
pages = {153-165},
volume = {283},
url = {https://mlanthology.org/l4dc/2025/compton2025l4dc-learning/}
}