Model Based Automotive System Design: A Power Window Controller Case Study
Modern day vehicles come equipped with a large number of sensors, actuators and ECU's with sophisticated control algorithms, which requires engineering activities from various disciplines. An automotive system is developed in various stages with multiple stakeholders involved at each stage. Each stakeholder provides a distinct view point on system representation, which makes it challenging to bridge the gaps in developing a holistic understanding of the system functionality. The safety critical nature of automotive systems induces timing and dependability concerns that must be addressed at all stages. Furthermore, the relatively long development life-cycle of automotive systems makes it imperative to have a clear strategy for long term evolution. To deal with these challenges, model based techniques are applied in the industry for automotive systems development. System engineers use a suitable architecture description language (ADL) to represent the system architecture at several levels of abstraction. A number of system architecture description and software architecture standards have been developed in the automotive industry to streamline the development process. However, most of these standards are elaborate and need a fair amount of understanding before they can be applied.
In this work, we explore the application of existing system architecture description and software architecture standards. Our main contribution is a Power Window Controller system demonstrator that covers the methodology described by EAST-ADL and AUTOSAR. Through this case study, we intend to highlight the key aspects and gaps in the application of EAST-ADL and AUTOSAR. Starting from features and requirements, we have analyzed the impact of architectural decisions at each stage of automotive system development. We also performed design verification, timing analysis and dependability analysis to ensure correctness of the system. Lastly, considerations regarding variability have been discussed to support evolution.