In order to simplify the development process, most automated driving technologies employ limitations in their operational scope. Comparing automated driving performance in these specific contexts to manual driving performance in general then provide an unfair comparison. Instead, the reference should be made as specific and as precise as possible.

The objective of the present study was to identify basic and relevant numbers required to evaluate the safety performance of automated driving by identifying conflict situations for targeted traffic environments and establishing crash rates for these conflict situations. Results in terms of crash rate differences confirmed that relevant references must be specified according to the automated driving technology limitations, i.e. traffic environment, vehicle type and conflict situation. The need to distinguish between crash outcomes (fatal, severe or minor injuries, or property damage only crashes) is obvious. The number of kilometres necessary to establish future significant differences in these crash rates indicate that retrospective traffic safety evaluations will be far from realistic for many years to come.

This report provides an overview on basic numbers required to evaluate traffic safety performance of automated driving cars, a discussion on the methodology to achieve these numbers, and insights into expectations on future traffic safety evaluations of automated driving traffic safety performance.