Research paper
Predictive safety filters · Checking learned control inputs
An MPC-based filter checks proposed learning-controller inputs and modifies them when needed using a data-driven model and uncertainty bounds. The paper studies constrained pendulum and Bullet quadrotor simulations.
Reviewed 2026-11-05 · Source and bibliography checked
- Original title
- A predictive safety filter for learning-based control of constrained nonlinear dynamical systems
- Publisher
- Elsevier BV
- Authors
- Kim Peter Wabersich, Melanie N. Zeilinger
- DOI
- 10.1016/j.automatica.2021.109597
- Publication date
- 2021-07
- Source language
- en
- Edition
- Automatica; 2021; DOI 10.1016/j.automatica.2021.109597
- Access
- Publisher metadata and open author manuscript
- Publication status
- Published
Related resources and engineering notes
Source overview
An MPC-based filter checks proposed learning-controller inputs and modifies them when needed using a data-driven model and uncertainty bounds. The paper studies constrained pendulum and Bullet quadrotor simulations.
Use and review
Abstract and quadrotor evaluation section checked. Simulation safety depends on model, uncertainty and terminal-policy assumptions; this does not replace validation of a physical robot safety function.
Sources & editions
A predictive safety filter for learning-based control of constrained nonlinear dynamical systems ↗
Elsevier BV · Automatica; 2021; DOI 10.1016/j.automatica.2021.109597
Automatica bibliography; arXiv:1812.05506v4 abstract and section 5.2, https://arxiv.org/abs/1812.05506v4. Journal date 2021-07 is not the archive posting date.