Safe Learning in the Real World via Adaptive Shielding with Hamilton-Jacobi Reachability

Michael Lu, Jashanraj Gosain, Luna Sang, Mo Chen
Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, PMLR 283:1257-1270, 2025.

Abstract

We present a robust shielding framework using Hamilton-Jacobi Reachability that can be combined with any off-policy Reinforcement Learning to enable safer learning. Using an approximate model of a system dynamics, our method can capture the local model mismatch from a safety perspective. This leads to a more conservative safety filter that can adapt to model mismatch. Using a Turtlebot 2, we demonstrate that our method can allow for safe learning in the real-world with minimal human intervention.

Cite this Paper

BibTeX
@InProceedings{pmlr-v283-lu25a, title = {Safe Learning in the Real World via Adaptive Shielding with Hamilton-Jacobi Reachability}, author = {Lu, Michael and Gosain, Jashanraj and Sang, Luna and Chen, Mo}, booktitle = {Proceedings of the 7th Annual Learning for Dynamics \& Control Conference}, pages = {1257--1270}, year = {2025}, editor = {Ozay, Necmiye and Balzano, Laura and Panagou, Dimitra and Abate, Alessandro}, volume = {283}, series = {Proceedings of Machine Learning Research}, month = {04--06 Jun}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v283/main/assets/lu25a/lu25a.pdf}, url = {https://proceedings.mlr.press/v283/lu25a.html}, abstract = {We present a robust shielding framework using Hamilton-Jacobi Reachability that can be combined with any off-policy Reinforcement Learning to enable safer learning. Using an approximate model of a system dynamics, our method can capture the local model mismatch from a safety perspective. This leads to a more conservative safety filter that can adapt to model mismatch. Using a Turtlebot 2, we demonstrate that our method can allow for safe learning in the real-world with minimal human intervention.} }
Endnote
%0 Conference Paper %T Safe Learning in the Real World via Adaptive Shielding with Hamilton-Jacobi Reachability %A Michael Lu %A Jashanraj Gosain %A Luna Sang %A Mo Chen %B Proceedings of the 7th Annual Learning for Dynamics \& Control Conference %C Proceedings of Machine Learning Research %D 2025 %E Necmiye Ozay %E Laura Balzano %E Dimitra Panagou %E Alessandro Abate %F pmlr-v283-lu25a %I PMLR %P 1257--1270 %U https://proceedings.mlr.press/v283/lu25a.html %V 283 %X We present a robust shielding framework using Hamilton-Jacobi Reachability that can be combined with any off-policy Reinforcement Learning to enable safer learning. Using an approximate model of a system dynamics, our method can capture the local model mismatch from a safety perspective. This leads to a more conservative safety filter that can adapt to model mismatch. Using a Turtlebot 2, we demonstrate that our method can allow for safe learning in the real-world with minimal human intervention.
APA
Lu, M., Gosain, J., Sang, L. & Chen, M.. (2025). Safe Learning in the Real World via Adaptive Shielding with Hamilton-Jacobi Reachability. Proceedings of the 7th Annual Learning for Dynamics \& Control Conference, in Proceedings of Machine Learning Research 283:1257-1270 Available from https://proceedings.mlr.press/v283/lu25a.html.

Related Material