Composite dependability modeling for in-vehicle networks

This paper discusses an emerging dependability assessment framework for CAN (Controller Area Network)-based in-vehicle applications. The dependability assessment of in-vehicle systems from their designed architecture is a challenging issue for automobile industry. However, since modern in-vehicle systems consist of massive components such as ECUs (Electronic Control Units), CAN equipments and software, it is not always easy to evaluate the whole system dependability from their interactions and interdependency. In this paper, with commonly-used architecture description languages such as UML (Unified Modeling Language) and SysML (Systems Modeling Language), we divide the system structure into several components, and give behavioral description of components, where each component corresponds to a state-dependent model in terms of its probabilistic behavior. We adopt a probabilistic model checking approach to analysis of the component behavior, and describe it by means of the PRISM language. Our composite dependability modeling is a promising technique not only in the in-vehicle application but also in enterprise applications to perform the quantification of system dependability in design. We also give an illustrative example on the dependability modeling of an auto-cruise system in the design phase.