Article Preview
Top1. Introduction
Commercial automobiles have become complex high-technology products in a relatively short time span. Different factors contribute to this complexity. One of them is the increasing number of vehicle functionalities supported by software, electronics and mechatronic technologies; a trend that does not seem to slow down. The involvement of carmakers in the development of these functionalities differs from one vehicle domain to the other (chassis, body, powertrain), ranging from black box integration to white-box developments. Another factor is the way in which car manufacturers have evolved from their historical mechanical and manufacturing background to the intricate organizations that develop the automobile products of today. The advent of the electrical vehicle makes this last two factors even more evident, not only because of the “untraditional” technologies that carmakers need to master, but also because the arrival of new stakeholders, actors and interests around the electrical vehicle mean that the traditional scope of the automobile has changed.
Generally, this evolution has translated into development processes that are not as efficient, flexible and agile as they could or should be (Chale et al. (2012)). The need to master these different complexity-inducing factors and improve the efficiency of product development, plus the arrival of the ISO 26262 standard (which besides from safety-related aspects, also raises issues concerning development processes of automotive systems, currently under-formalized) have motivated the adoption of model-based system engineering. Model-based system engineering advocates the use of models, conforming to a common semantic meta-model, all along the system development process. The meta-model specifies a common unambiguous semantics formalizing system engineering terminology and then providing a common language for system descriptions, i.e. models. Models, produced along the development process, provide system descriptions at different abstraction levels. Abstraction levels help human reasoning and analysis capabilities allowing system specifications to be refined and incrementally validated as long as the comprehension of the system increases. The meta-model approach is also attractive for system development as meta-models and their related models can be easily extended to support an open ended evolution of domain specific concepts. The resulting (information) models are conceptual models in the sense of a conceptual structure: the models impose a machine adequate outer structure upon the otherwise unbound creativity of the engineers, at the same time trying to be as flexible as possible to support a stimulating creativity process. The more concrete the abstraction level becomes (i.e., the more formally accessible system details are described) the lesser becomes the creativity freedom because the information model becomes more constraining until the code—as the final, formal specification—does not offer any more freedom as its respective set of instructions. This means that the information model described in this paper warily reduces the freedom that initially is of utmost necessity for the creativity process of the engineer along the abstraction levels in favour of the productivity of the computer, which in turn helps to find inconsistencies, redundancies, and partly even missing system description aspects by advanced analyses and optimization capabilities to effectively improve development processes.
Thanks to these capabilities, the adoption of model-based design has several benefits including an improvement of quality, through a more rigorous and costless traceability between requirements, design, analysis and testing. While the benefits of model-based design are widely understood, there is no COTS solution today providing a full-fledged model-based environment for automotive systems. The first problem is that many commercial solutions use proprietary meta-models that scarcely fit automotive design needs. Moreover, ideally, the meta-model should be shared in the entire automotive domain, and then proprietary languages should be avoided opting instead for standard languages. UML extensions as SysML, could be an option, but SyML, per se, does not support many concepts of vital importance for the automotive domain, as for instance, concepts for safety analysis, timing analysis and variability. To support these concepts UML needs to be specialized through specific profiles. Even though some efforts have been spent in that direction in literature—e.g. for safety (Cancila et al. (2009)), for timing (OMG MARTE (2011))—we did not reach the stage in which these efforts are unified and integrated in SyML.