Analyzing Industrial Architectural Models: Where Verification meets Statistical Model Checking

Automotive systems rely on complex and safety-critical software that implements many of the functions of modern vehicles. The system’s software architecture has to be decided well in advance of production, so it is very desirable to assess the quality of the software architecture in order to obtain quick indications of errors at early design phases. In this paper, we present a constellation of techniques that are tailored for architectural models in EAST-ADL. The methods are complementary in terms of covering EAST-ADL model analysis against a rich set of requirements, and in terms of the varying degree of confidence in the provided guarantees. We overview the current model-driven development in a chosen automotive context, and we propose various analysis techniques that target EAST-ADL architectural models, in an attempt to tackle some of the exposed design needs. Our contributions spans over simulation of EAST-ADL functions in Simulink, model- checking EAST-ADL models with timed automata semantics with an academic prototype called ViTAL, and statistical model-checking in UPPAAL, applied on a network of timed automata obtained by an automatic transformation from the EAST- ADL model, which we propose in this paper. An industrial Brake-by-Wire prototype is the use-case on which we show our results.