Model-Based Analysis of Critical Resources in Automotive System Architectures

Doctoral thesis, 2013

Modern cars contain more and more innovative user functions, some of which can only be implemented in software and electronics. Electronics and software allows the vehicle system to be flexible, and at the same time creates additional complexity. This complexity needs to be managed by the E/E (Electrical/Electronic) system architecture, which serves as a platform for user functions implemented in electronics and software. The E/E system architecture needs to be developed to find a balance between product cost and flexibility (which often means additional cost for electronics). The architectures need to be flexible enough for different types of cars produced over a long time-period, but they should not contain product cost for hardware that is never used. Efficient development of E/E system architectures calls for early validation methods. The objective of this thesis is to identify critical resources in the E/E system architecture and evaluate the use of these resources throughout the life-cycle of the architecture. The methods shall be able to use available information with as little changes as possible, to minimize the manual effort involved. This is important in an industrial context, where all added work is questioned. The research in this thesis is based on an empirical approach. Analysis methods for evaluating critical resources in E/E system architectures have been developed using requirement specifications, documents, models, and software implementations from the automotive industry. The methods have been validated in case studies conducted in the automotive industry. Interviews of domain experts have been used to collect additional data providing more complete context for when and how to apply the analysis methods. This thesis shows how the ROM type memory size can be analyzed within 15% accuracy using only the information available in early development phases. To reduce the manual effort required to analyze the memory size, a tool called CompSize was developed. In addition, the thesis shows how other critical resources in the system architecture such as processor capacity and communication bandwidth can be analyzed in early phases using the AQOSA framework. It is also shown how architecture development tools like AQOSA complement the domain knowledge of an architect. Furthermore, the thesis proposes a development process for automotive system architectures, with the purpose of increasing the understanding of the context for the analysis methods. The analysis methods introduced in this thesis are suitable as support for architectural trade-off analysis that will allow architects to obtain early feedback on design decisions. This is important in order to avoid late design changes that are often difficult and costly to make, and that are associated with high risk. Especially, the methods are important during deployment of software onto hardware to identify so-called bottlenecks in the system architecture. Bottlenecks are when critical resources in the system architecture are overloaded; resources like processing capacity and memory space in an ECU, and the bandwidth in a communication bus.