A discrete event approach for vehicle failure diagnostics

The paper presents a model-based method for the design of diagnostics for a large-scale system. The method is demonstrated by application to the diagnostic design for the longitudinal control system of a fully automated vehicle capable of platooned operation. It is assumed that the system is modelled by continuous and discrete event models (DEM's). We comment on the abstraction of continuous models into discrete event models and show how DEM's may be constructed in a modular manner for a given set of sensors, observers and controllers defined in the continuous domain. The method is modular in the sense that the model of the integrated system is built by composing unit models using mathematically well-defined operators. We provide a formal way of defining a diagnostic requirement and thereafter proving whether a given diagnostic scheme satisfies the requirement. We also provide a means of generating a FSM specification of the residue processing software. The paper provides two examples to intuitively explain the working of the methods. It also provides the verification results and models for the diagnostic design of the longitudinal control system of a fully automated vehicle.