Neutrons Sensitivity of Autonomous Vehicles Stack: from sensors data processing to actuators modification

In this proposal we intend to move a step forward in the reliability evaluation of self-driving vehicles by considering the full stack required to enable automation. Our idea is to characterize the sensitivity of basic procedures of autonomous vehicles through dedicated codes extrapolated from simulators and complex frameworks. Then, we will compare the error rate of the different codes to identify the ones that are more likely to be corrupted. Additionally, we will observe the faults propagation in the autonomous vehicle framework to identify the operations that are more critical, i.e., whose corruption is more likely to cause an accident or unexpected behavior.