RGB cameras failures and their effects in autonomous driving applications

RGB cameras are one of the most relevant sensors for autonomous driving applications. It is undeniable that failures vehicle may compromise task, possibly leading to unsafe behaviors when images subsequently processed by system altered. To support definition safe and robust architectures intelligent systems, in this paper we define failure modes a camera, together with an analysis effects known mitigations. Further, build software library generation corresponding failed feed them six object detectors mono stereo self-driving agent simulator. The resulting misbehaviors respect operating clean allow better understanding related safety risks image-based