What, Indeed, is an Achievable Provable Guarantee for Learning-Enabled Safety Critical Systems

TL;DR

This paper discusses the challenges of providing rigorous safety guarantees for learning-enabled systems in safety-critical domains and proposes a two-step verification method to achieve provable statistical guarantees.

Contribution

It introduces a novel two-step verification approach aimed at enabling provable safety guarantees for learning-enabled systems.

Findings

Existing methods cannot achieve provable guarantees

Proposed two-step verification method

Potential for rigorous safety assurance in critical systems

Abstract

Machine learning has made remarkable advancements, but confidently utilising learning-enabled components in safety-critical domains still poses challenges. Among the challenges, it is known that a rigorous, yet practical, way of achieving safety guarantees is one of the most prominent. In this paper, we first discuss the engineering and research challenges associated with the design and verification of such systems. Then, based on the observation that existing works cannot actually achieve provable guarantees, we promote a two-step verification method for the ultimate achievement of provable statistical guarantees.