Using Vehicular Protection Mechanisms to Enable Fault-Aware Safety Verification of Autonomous Vehicles

Responsibility-Sensitive Safety (RSS) became a widespread solution for online verification of autonomous vehicle (AV) safety. Nevertheless, RSS is built on strong premises concerning the capabilities of AVs. In this paper, we explore the Data-Centric Design of AVs using the SmartData method, which supports the automatic derivation of safety-property monitors from design documents, thus promoting online verification of sensitivity to potential faults in the system components described by vehicular protection mechanisms. The proposed solution explores knowledge from vehicular protection mechanism to promote adaptability to RSS in face of potential faults. The adaptability is enabled by exploring predictors of protection mechanisms to promote a time-to-triggering metric that adjusts the RSS parameters, thereby providing awareness of potential faults. Furthermore, we propose an adaptation to proper responses to a dangerous longitudinal situation in the face of a near-future loss of the brake actuation. We prove that the proposed solution complies with the original RSS proposal by extending its original inductive proof for safety.