EUVIMEDEUVIMED entdecken
Crossref

Arguing Residual Risk Acceptance for E2E AI Automated Driving Systems Using Black-Box Risk-Based Testing

Rasmus Adler, Ioannis Sorokos, Jan Reich, Thomas Muehlenstaedt, Sandro Nueesch, Raphael Fonte Boa Trindade

Proceedings of the AAAI Symposium Series · 06.11.2026 · 10 (1) · 208-216

DOI: 10.1609/aaaiss.v10i1.43408 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 06.11.2026. Verfasst von Rasmus Adler und 5 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Residual risk remains after risk-elimination and mitigation measures and must be tolerable for any technical system. For complex AI systems operating in dynamic, uncertain environments, quantifying this risk and demonstrating its tolerability are challenging, especially for Automated Driving Systems (ADS) using end-to-end (E2E) AI models. Their vast scenario space and the difficulty of grouping scenarios by equivalent model processing hinder risk quantification. We propose a black-box, risk-based testing approach and an argumentation framework for assessing residual-risk tolerability from test evidence. The framework operates at macro, scenario, and scenario-instance levels, combines complementary risk-acceptance criteria, and distinguishes technology-neutral evidence of risk reduction from technology-specific safety-integrity evidence. Derived from European regulatory frameworks, including the EU AI Act, it formalizes key evidence relations and illustrates them through a hypothetical driving scenario.

Zitieren

APA

Adler, R., Sorokos, I., Reich, J., Muehlenstaedt, T., Nueesch, S., & Trindade, R. F. B. (2026). Arguing Residual Risk Acceptance for E2E AI Automated Driving Systems Using Black-Box Risk-Based Testing. Proceedings of the AAAI Symposium Series, 10(1), 208-216. https://doi.org/10.1609/aaaiss.v10i1.43408

Vancouver

Adler R, Sorokos I, Reich J, Muehlenstaedt T, Nueesch S, Trindade RFB. Arguing Residual Risk Acceptance for E2E AI Automated Driving Systems Using Black-Box Risk-Based Testing. Proceedings of the AAAI Symposium Series. 2026;10(1):208-216. doi:10.1609/aaaiss.v10i1.43408

BibTeX
@article{Adler2026,
  author = {Adler, Rasmus and Sorokos, Ioannis and Reich, Jan and Muehlenstaedt, Thomas and Nueesch, Sandro and Trindade, Raphael Fonte Boa},
  title = {Arguing Residual Risk Acceptance for E2E AI Automated Driving Systems Using Black-Box Risk-Based Testing},
  journal = {Proceedings of the AAAI Symposium Series},
  year = {2026},
  volume = {10},
  number = {1},
  pages = {208--216},
  doi = {10.1609/aaaiss.v10i1.43408}
}
Kostenlos starten →