Mission- and safety-critical systems are increasingly reliant on software. However, their development and qualification has become increasingly challenging: increasing size and more coupling between software and hardware layers has driven development cost to unaffordable levels. Mismatched assumptions at the architecture level, impacting quality attributes such as performance, timing, safety, reliability, are the leading causes and are difficult to test for.
To cope with these issues, architecture must be established to identify problems earlier in the development process. Architecture-centric modeling technology has led to a model-based virtual system integration approach that permits early discovery of problems (integration, performance, etc.). Despite advances in this area, such technologies have found limited use due to the perceived modeling and analysis cost with limited value. Many reasons are invoked, such as the efforts required for making an architecture that matches the software.
In this paper, we present a methodology for automating the production of an architecture skeleton from existing assets (code, functional model, etc.). This first architecture model is then enhanced to match system properties and then processed to verify stakeholders' requirements. By automating the process we reduce the human modeling efforts (and their related time and costs) and leverage the investment in analysis capabilities integrated while reducing the potential additional cost of using it in an existing practice.