The combination of multiple functions having different and complementary capabilities enables the emergence of Autonomous Vehicles. Their deployment is limited by the level of complexity they represent together with the challenges encountered in real environments with strong safety concerns. Thus a major concern prior to massive deployment is on how to ensure the safety of autonomous vehicles despite likely internal (e.g. malfunctions) and external (e.g. aggressive behaviors) disturbances they might undergo. This paper presents the challenges that undergoes the design and development of autonomous vehicles with respect to their functional architecture and adaptive behaviors from a safety perspective. For the purpose of the rationales, we define needs and requirements that lead to the formulation of an architectural framework. Our approach is based on paradigms and technologies from non-automotive domains to address non-functional system properties like safety, reliability and security. The notion of micro-services is also introduced for the self-safety of autonomous vehicles. These are part of the proposed framework that should facilitate the analysis, design, development and validation for the adequate composition and orchestration of services aimed to warrant the required non-functional properties, such as safety. In the present paper, we introduce the structural and behavioral adaptations of the framework to offer a holistic and scalable vision of the safety over the system.

The design of complex systems, as in the case of autonomous vehicles, requires a specialized systems engineering methodology and an adapted modelling framework. In particular, the integration of non-functional requirements, as important as the Safety, requires from this methodological framework the well-adapted semantic expression of constraints as well as their traceability during all phases of analysis, design and implementation. This paper focuses on the study of model-based autonomous system design and investigates the design flows and initiatives grasping with this complex computational model. The specialization of the ARCADIA methodology will be illustrated in a real industrial case.