Risk-Sensitive Safety Filters for Reinforcement Learning With Probabilistic Guarantees
IEEE Transactions on Automatic Control · 2025-11 · 70 (11) · 7414-7429
DOI: 10.1109/tac.2025.3575565 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Automatic Control, 2025-11. Verfasst von Armin Lederer und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 48 Literaturangaben. Gefördert von: European Research Council, Safe data-driven control for human-centric systems, ONR.
Abstract (Originalsprache)
Humans have the ability to deviate from their natural behavior when necessary, which is a cognitive process called response inhibition. Similar approaches have independently received increasing attention in recent years for ensuring the safety of control. Realized using control barrier functions or predictive safety filters, these approaches can effectively ensure the satisfaction of state constraints through an online adaptation of nominal control laws, e.g., obtained through reinforcement learning. While the focus of these realizations of inhibitory control has been on risk-neutral formulations, human studies have shown a tight link between response inhibition and risk attitude. Inspired by this insight, we propose a flexible, risk-sensitive method for inhibitory control. Our method is based on a risk-aware condition for value functions, which guarantees the satisfaction of state constraints. We propose a method for learning these value functions using common techniques from reinforcement learning and derive sufficient conditions for its success. By enforcing the derived safety conditions online using the learned value function, risk-sensitive inhibitory control is effectively achieved. The effectiveness of the developed control scheme is demonstrated in simulations.
Förderung
- European Research Council
- Safe data-driven control for human-centric systems
- ONR
- Clark Foundation Distinguished Fellowship
- Army Research Laboratory
- Swiss National Science Foundation as a part of NCCR Automation
- NCCR Catalysis
Zitieren
APA
Lederer, A., Noorani, E., Baras, J. S., & Hirche, S. (2025). Risk-Sensitive Safety Filters for Reinforcement Learning With Probabilistic Guarantees. IEEE Transactions on Automatic Control, 70(11), 7414-7429. https://doi.org/10.1109/tac.2025.3575565
Vancouver
Lederer A, Noorani E, Baras JS, Hirche S. Risk-Sensitive Safety Filters for Reinforcement Learning With Probabilistic Guarantees. IEEE Transactions on Automatic Control. 2025;70(11):7414-7429. doi:10.1109/tac.2025.3575565
BibTeX
@article{Lederer2025,
author = {Lederer, Armin and Noorani, Erfaun and Baras, John S. and Hirche, Sandra},
title = {Risk-Sensitive Safety Filters for Reinforcement Learning With Probabilistic Guarantees},
journal = {IEEE Transactions on Automatic Control},
year = {2025},
volume = {70},
number = {11},
pages = {7414--7429},
doi = {10.1109/tac.2025.3575565}
}
EUVIMED