Guidelines and Regulatory Framework for Machine Learning in Aviation

Abstract

Automation and eventually autonomy are regarded as the enabler for upcoming Urban Air Mobility (UAM) / Advanced Air Mobility segment. Only they could enable unprecedented opportunities for scaling drones and air taxis to a large number of vehicles, making the services available for everyone. Artificial Intelligence (AI) in general, Machine Learning (ML) in particular promise a huge leap towards achieving high levels of automation and further autonomy. Nevertheless, the safety concerns and challenges regarding compliance to the existing software standards are more pressing then ever before. Existing regulatory framework for hardware and software items fail to provide adequate acceptable means of compliance for AI-based systems. Hence, there are currently a number of ongoing efforts to update and augment the current standards. This paper will give an overview of the existing and upcoming regulatory framework for certifying AI-based systems. It will elaborate the EASA documents, artificial intelligence roadmap, Concepts of Design Assurance for Neural Networks (CoDANN), CoDANN II, as well as the concept paper on first usable guidance for level I machine learning applications. Furthermore, suitable guidance from EuroCAE, RTCA, ASTM and AVSI will be discussed.