Not All Miles are Equal: Why Time and Location Matter When Benchmarking Autonomous Safety
Waymo’s peer-reviewed studies in *Traffic Injury Prevention* show crash risk varies sharply by time and location, with late-night urban driving posing the highest danger. The research introduces granular human benchmarks to compare autonomous safety more accurately.
Waymo’s two new studies, published in *Traffic Injury Prevention*, highlight that crash risk is not uniform across time or geography. Traditional safety comparisons often rely on broad national averages, which obscure critical differences between regions, road types, and driving hours. The research demonstrates that fatal crash rates for human drivers in Memphis, for example, are 8.4 times higher than in Boston, while surface streets carry a fatal crash rate 2.3 times higher than freeways across 50 major U.S. urban areas.
The studies introduce a framework to break down human crash data by location and time, enabling more precise benchmarks for evaluating autonomous vehicle safety. Feng Guo, a Virginia Tech professor and lead data scientist at VTTI, emphasized that meaningful safety assessments must account for spatial and temporal disparities. Waymo’s approach pairs crash records with traffic volume data to create localized, time-specific risk profiles, addressing gaps in traditional aggregated analyses.
Risk escalates significantly during late-night hours and weekends, when human crash rates surge 2 to 6 times higher than daytime averages. The data reveals that overnight driving—just 1.5% of total human mileage—is disproportionately hazardous due to factors like fatigue and impaired driving. Jonathan Adkins of the Governors Highway Safety Association noted that autonomous technology could intervene during these high-risk windows, potentially preventing behavior-related crashes.
Waymo’s fleet operates four times more overnight miles than the average human driver, yet its crash rates remain lower across all time windows. Comparing 127 million autonomous miles to human drivers in identical conditions, Waymo avoided 359 injury crashes, with 53% occurring between 8:00 PM and 3:59 AM. The findings underscore the importance of dynamic benchmarks for assessing autonomous safety and aim to foster industry-wide consensus on localized risk evaluation.