Towards Nucleation of GoA3+ Approval Process

Abstract

The approval of Automatic Train Operation (ATO) from GoA3 on (GoA3+) requires a strong developers’ network to ensure the homogeneous landscape of expert opinions for regulators and courts. Certain technologies needed for GoA3+, especially Computer Vision (CV) powered by Deep Learning (DL), are fast developing and therefore do not exhibit a sufficient degree of professional experience for technical norms, although there is no scarcity at methodical candidates for such an approval process. What appears to be missing is a set of the relevant approval requirements as well as their implications for CV and DL, in order to serve as a common nucleation core for the development of a GoA3+ approval process. This paper aims at providing such a core. THIS CONTRIBUTION REPRESENTS SOLELY AUTHORS’ PROFESSIONAL OPINION, NOT THE ONE OF THEIR EMPLOYER.