Applicability of machine learning architectural patterns in vehicle architecture: A case study

Paper i proceeding, 2021

Machine learning (ML) has grown in importance in the last decade and has become one of the mainstream software technologies used. Together with containerization, data-driven development and microservices, it has also been used in the automotive industry, mainly for autonomous driving functions, but also, for example, for interior driver state monitoring or voice control speech recognition. The goal of this case study is to document experiences of which of the emerging ML architectural patterns are used in modern cars already, which are planned for the future and how they are used. We study the software architecture of a modern car product line based on existing patterns and workshops with software architects. The results show that many ML-specific patterns are used or planned to be used in the near future. Only a handful of patterns are not applicable or not planned to be used. However, we have also found that the established description of the patterns is not suited for the automotive software architectures, which can jeopardize its correct broad usage in the industry. We conclude that the patterns should be described more clearly and that they are more used than we could have anticipated based on the literature.