EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics

Felix Berkenkamp iD, Andreas Krause, Angela P Schoellig

Machine Learning · 2023-10 · 112 (10) · 3713-3747

DOI: 10.1007/s10994-021-06019-1 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Machine Learning, online erschienen am 24.06.2021. Verfasst von Felix Berkenkamp und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Science, ETH Zurich, Zurich, Switzerland. Die Arbeit wurde bisher 199-mal zitiert (Crossref) und nennt 51 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung, Natural Sciences and Engineering Research Council of Canada, Connaught New Researcher Award.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Selecting the right tuning parameters for algorithms is a pravelent problem in machine learning that can significantly affect the performance of algorithms. Data-efficient optimization algorithms, such as Bayesian optimization, have been used to automate this process. During experiments on real-world systems such as robotic platforms these methods can evaluate unsafe parameters that lead to safety-critical system failures and can destroy the system. Recently, a safe Bayesian optimization algorithm, called SafeOpt , has been developed, which guarantees that the performance of the system never falls below a critical value; that is, safety is defined based on the performance function. However, coupling performance and safety is often not desirable in practice, since they are often opposing objectives. In this paper, we present a generalized algorithm that allows for multiple safety constraints separate from the objective. Given an initial set of safe parameters, the algorithm maximizes performance but only evaluates parameters that satisfy safety for all constraints with high probability. To this end, it carefully explores the parameter space by exploiting regularity assumptions in terms of a Gaussian process prior. Moreover, we show how context variables can be used to safely transfer knowledge to new situations and tasks. We provide a theoretical analysis and demonstrate that the proposed algorithm enables fast, automatic, and safe optimization of tuning parameters in experiments on a quadrotor vehicle.

Schlagwörter

Stichwörter der Autor:innen

Bayesian optimizationReinforcement learningRoboticsSafe explorationSafety constraints

Autor:innen und Institutionen

  • Department of Computer Science, ETH Zurich, Zurich, Switzerland.
  • Institute for Aerospace Studies, University of Toronto, Toronto, Canada.

ORCID-Profile: Felix Berkenkamp

Förderung

  • Schweizerischer Nationalfonds zur Förderung der Wissenschaftlichen Forschung
  • Natural Sciences and Engineering Research Council of Canada
  • Connaught New Researcher Award
  • ETH Zurich

Zitieren

APA

Berkenkamp, F., Krause, A., & Schoellig, A. P. (2023). Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics. Machine Learning, 112(10), 3713-3747. https://doi.org/10.1007/s10994-021-06019-1

Vancouver

Berkenkamp F, Krause A, Schoellig AP. Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics. Machine Learning. 2023;112(10):3713-3747. doi:10.1007/s10994-021-06019-1

BibTeX
@article{Berkenkamp2023,
  author = {Berkenkamp, Felix and Krause, Andreas and Schoellig, Angela P},
  title = {Bayesian optimization with safety constraints: safe and automatic parameter tuning in robotics},
  journal = {Machine Learning},
  year = {2023},
  volume = {112},
  number = {10},
  pages = {3713--3747},
  doi = {10.1007/s10994-021-06019-1}
}
Kostenlos starten →