Research paper
Estimated safe regions · Switching between learning and recovery
A supervisor switches between a learning controller and a predefined corrective controller using an estimated safe region. Corrective-action outcomes update that estimate; the study evaluates pendulum and quadcopter control examples.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- A General Framework to Increase Safety of Learning Algorithms for Dynamical Systems Based on Region of Attraction Estimation
- Publisher
- Institute of Electrical and Electronics Engineers (IEEE)
- Authors
- Zhehua Zhou, Ozgur S. Oguz, Marion Leibold, Martin Buss
- DOI
- 10.1109/tro.2020.2992981
- Publication date
- 2020-10
- Source language
- en
- Edition
- IEEE Transactions on Robotics; 2020; DOI 10.1109/tro.2020.2992981
- Access
- Publisher metadata and open author manuscript
- Publication status
- Published
Related resources and engineering notes
Source overview
A supervisor switches between a learning controller and a predefined corrective controller using an estimated safe region. Corrective-action outcomes update that estimate; the study evaluates pendulum and quadcopter control examples.
Use and review
Checked author abstract and institutional bibliography. Simplified models and estimated regions do not eliminate failure; corrective-controller reliability and switching conditions require separate validation.
Sources & editions
A General Framework to Increase Safety of Learning Algorithms for Dynamical Systems Based on Region of Attraction Estimation ↗
Institute of Electrical and Electronics Engineers (IEEE) · IEEE Transactions on Robotics; 2020; DOI 10.1109/tro.2020.2992981
IEEE TRO 2020 bibliography; author abstract https://zhehuazhou.github.io/publication/2020general and TUM institutional publication record. Quadcopter evaluation source is TUM-hosted author manuscript.