Qualification methodology for ISO 26262 certification of automotive SoCsystems

Méthodologie de qualification pour la certification ISO 26262 des systèmes-sur-puce pour l’automobile This thesis proposes to set up a flow and a methodology of ISO26262certification for system-type integrated circuits on a digital chipdedicated to driving. These circuits are generally composed of sev-eral Intellectual Properties, IPs, dedicated to different functionssuch as communication or processing of information from sensors (camera, li-dar ...), real-time system, vision and imaging, system management (operatingsystem), security. The ISO26262 methodology requires the extraction of a num-ber of metrics related to the resilience of the system to single and multiple faultsas well as the effectiveness of countermeasures (detection, reporting and correc-tion of errors) and failure modes. The extraction of failure metrics from faulttrees is a method known and documented in the literature. Nevertheless, itsapplication has often been limited to macroscopic electromechanical systemssuch as a car, actuator or sensor chains. On the other hand, these methods arerarely applied in the field of automotive SoCs where the extraction of metricsis still largely manual (usually using a spreadsheet) and dependent on an ex-pert, and where the verification of the effectiveness of countermeasures is bestdone by targeted fault injection on a few sub-parts of the complete system orirradiation under a particle beam. This thesis proposes to develop a reliabilitymetrics extraction methodology based on fault injection per block as well ascomposition methods to obtain the metrics at the level of the complete system.The first part of the thesis will be devoted to the study of the bibliography onthe construction of fault trees, the ISO26262 standard and the declination ofthe different reliability metrics in the case of a digital SoCs type system. Theextraction of metrics at the block level will be based on 2 different methods, oneanalytical based on probabilities, the other experimental based on fault injec-tion. The aim is not to develop new probability codes or fault injection toolsbut to develop a methodology to use them in the context of a SoC to obtain thedesired data. The second part of the thesis will concern the composition of thedata obtained at the functional block level in order to obtain the ISO26262 met-rics at the system level (SoC). It will be a matter of developing a compositionmethod adapted in particular to the characteristics of SoCs (communicatingsystem, performing calculations that must react in real time, ...) and to the faultmodels that characterize them or imposed by the ISO26262 standard. The thirdpart of the thesis concerns the application of the developments described in theprevious paragraph to an SoC-type system and the verification of the resultsobtainined