Modelling road hazards and the effect on AV safety of hazardous failures

Autonomous vehicles (AV) are about to appear on our roads within the next few years. However, to achieve the final breakthrough, not only functional progress is required, but also the fundamental safety questions must be solved. Among those, a question demanding special attention is the need to assess the overall safety of an AV and quantify that it is safe enough to take part in normal traffic despite its inherent imperfections. This paper addresses this problem and offers a probabilistic model, which allows one to study how imperfections of an AV perception system and of mechanisms responsible for AV safety (e.g., Safety Monitors), can impact AV safety in the presence of road hazards. We also demonstrate how the model can be used to validate if the AV is safe enough, to understand the criticality of (perception) errors, and to identify areas/parameters that have more influence on safety than others.