Hybrid Causal Model Based Diagnosis. Application to Automotive Embedded Functions

International audienceThe behavior of embedded systems is commonly characterized by hybrid phenomena in which each operational mode is triggered by commands sent by electronic control units (ECU), involving hardware and software components. While hardware components are inherently continuous, the ECUs introduce discrete switching between the behavioral modes of the apparatus. For continuous systems, the theory of logical diagnosis casts the diagnosis problem within a consistency-based reasoning scheme that requires normal behavior models only. On the other hand, hybrid model based diagnosis methods rely on the availability of fault models and implement abductive reasoning similarly to what is done in discrete event model based diagnosis approaches. In this paper we propose a hybrid model consistency based method in which ideas are borrowed from discrete event system diagnosis to build a partial diagnoser, and from continuous systems to check consistency and track causal dependencies underlying discrepancies between expected and observed behaviours