Approximate predictive control barrier function for discrete-time systems

Nov 18, 2024ยท
Alexandre Didier
Alexandre Didier
,
Melanie N. Zeilinger
ยท 0 min read
Abstract
We propose integrating an explicit approximation of a predictive control barrier function (PCBF) in a safety filter framework. The approximated PCBF is implicitly defined through an optimal control problem and allows guaranteeing invariance of an implicitly defined safe set as well as stability of this safe set within a larger domain of attraction. By extending existing theoretical analysis of the PCBF, we establish inherent robustness of the original algorithm and translate the guarantees to input-to-state stability of the proposed algorithm with respect to possible approximation errors, recovering the same guarantees in the absence of approximation errors. The proposed algorithm allows certifying inputs with respect to state constraint satisfaction through a single function evaluation and filtering unsafe inputs through a control barrier function based safety filter, which is independent of the time horizon of the original predictive optimisation problem, resulting in significant online computational benefits. We demonstrate the stability properties of the proposed algorithm on a linear system example as well as its use a fast safety filter for miniature race cars in simulation.
Type
Publication
arXiv preprint arXiv:2411.11610