Feature dependencies in automotive software systems: Extent, awareness, and refactoring

Abstract

Many automotive companies consider their software development process to be feature-oriented. In the past, features were regarded as isolated system parts developed and tested by developers from different departments. However, in modern vehicles, features are more and more connected and their behavior depends on each other in many situations. In this article, we describe how feature-oriented software development is conducted in automotive companies and which challenges arise from that. We present an empirical analysis of feature dependencies in three real-world automotive systems. The analysis shows that features in modern vehicles are highly interdependent. Furthermore, the study reveals that developers are not aware of these dependencies in most cases. For the three examined cases, we show that less than 12% of the components in the system architecture are responsible for more than 90% of the feature dependencies. Finally, we propose a refactoring approach for implicit communal components, which makes them explicit by moving them to a dedicated platform component layer.