Document presents research on Safety Performance Indicators (SPIs) for continuous monitoring and evaluation of Automated Driving Systems safety. Leading indicators anticipate outcomes through metrics including time-based, distance-based, probability-based and deceleration-based measures; lagging indicators confirm outcomes after occurrence such as crashes and injuries. Analysis of 21 datasets identified SHRP2 as primary source with 1402 crashes and 5262 near-crashes. Performance evaluation using Distance Headway and Time-to-Collision achieved Area Under the Curve close to 0.9 distinguishing crashes and near-crashes from normal driving. Main bottleneck identified is data quality and scarcity of reliable safety event labels.
German In-Depth Accident Study: an ongoing project between BASt, the University of Hannover, and the Automotive Research Association (FAT). See http://www.gidas.org.