Methodology for determining maximum injury potential for automated driving system evaluation

Scenario-based testing is being proposed as a method for evaluating the performance of an Automated Driving System (ADS). A scenario, in the driving context, is described as “a temporal sequence of scene elements, with actions and events of the participating elements occurring within this sequence” (Riedmaier et al. Citation 2020). To evaluate an ADS, the ADS may be placed in a number of scenarios contained in a collection, or database. This database could include scenarios that represent nominal driving situations, reactions to critical events, or test the failsafe operation of an ADS.

The outcome of the scenarios, e.g., crash, no crash, or injury risk mitigation, is often heavily dependent on the evasive maneuvers taken by the actors in the scenario. To be able to evaluate the performance of an ADS using scenarios, those scenarios should represent a potential collision of sufficiently high severity so that the ability of the ADS to avoid an outcome involving possible serious injury can be assessed. For example, if a scenario database is composed of only scenarios with vehicles traveling at low speed in the same direction, then an assessment of an ADS’s ability to avoid serious injury collisions is not meaningful because there would be little possibility for a serious injury to occur regardless of the ADS’s performance. The objective of this study is to propose a method to estimate the probability of an injury outcome in a scenario, e.g., Maximum Abbreviated Injury Scale of 3 or greater, given the worst case of little or no evasive maneuvers before the collision. This injury potential can be used as a means for comparing scenario-based testing datasets to each other or to real-world crash data to determine the appropriateness of the scenario database.