Bridging the Gap Between ISO 26262 and Machine Learning: A Survey of Techniques for Developing Confidence in Machine Learning Based Systems. 2020-01-0738
Machine Learning (ML) based technologies are increasingly being used to fulfill safety-critical functions in autonomous and advanced driver assistance systems (ADAS). This change has been spurred by recent developments in ML and Artificial Intelligence techniques as well as rapid growth of computing power. It is clear that ML-enabled systems can deliver value as part of a production ADAS program. However, demonstrating that ML-based systems are capable of achieving the necessary level of safety integrity remains a challenge. Current research and development work focused on establishing the reliable and safe operation of ML-based systems is disjoint and typically presents individual techniques that might be used to gain confidence in these systems. As a result, there is minimal guidance for adapting an established ISO 26262 compliant automotive engineering program to enable the development of ML-based systems.
This paper presents a literature survey of recent ML literature to identify techniques and methods that can contribute to meeting ISO 2626 requirements. The surveyed literature is mapped onto the ISO 26262 V-model and the applicability of individual techniques and methods are discussed for each major phase of development. The overall themes observed in literature suggest that current research and development is focused into three main areas: 1) architecture for ML-based systems, 2) verification and validation using formal mathematical methods, and 3) verification and validation using testing and simulation techniques.
Jose Serna, Simon Diemert, Laure Millet, Rami Debouk, Ramesh S, Jeffrey Joyce
Critical Systems Labs Inc., General Motors LLC