Economical valuation of architectural decisions within automotive

Today most innovations made within the automotive domain are driven by electronics. The automotive customers demand new functionality with every new product release and the time-to-market is constantly shortened. Automotive embedded systems are often resource constrained and trade-offs between the system behavior and the resources required is of great importance. The cost sensitive automotive industry has to optimize the use of the system's limited resources, but in the meantime also be flexible. The system needs to support a large number of vehicle configurations over many years of production. The design decisions are usually based on many factors that pull in different directions such as maintenance, portability, usability etc. The growing complexity of the product and the many uncertain factors create a need for support in the design process. To better understand this problem decision methods used within an R&D department of an international vehicle manufacturer has been investigated through interviews and surveys. The survey reveals that a majority of the respondents use unstructured methods for resolving decision issues. When respondents were asked about their preferences there was an expressed need for more structured methods. In this research several existing methods have been surveyed and the methods most relevant to this issue are further described in this thesis. The main contribution of this thesis is an evaluation method using Real Options. The method provides the opportunity to analyze the cost of designing for flexibility to cope with a future growth of the product, based on the estimated value of the future functionality. To improve the usability an evaluation process is defined to aid engineers. This process provides a way of valuing system designs and enables the engineer to think about the future in a systematic manor. To analyze the resource usage within an embedded system a method is proposed on how to evaluate the resource efficiency of functions implemented within an automotive embedded system. The challenge of this work has been to develop methods that are found helpful to the industry and are easy enough to use so that designers are willing to try them again.