Lessons from Automated Vehicle Collision Data in the United States

The deployment and successful integration of automated vehicles into the traffic stream relies on our understanding of their implications on safety and operation. As such, this study aims to contribute to the existing literature by investigating automated vehicle collision data from the United States, which include vehicles with advanced driver assistance systems (ADAS) and automated driving systems (ADS). The study analyzes temporal trends in the observed collisions and compares the most influential factors on collision frequency and severity. In addition, future collisions are predicted using time-series modeling. The results demonstrate a future decrease in ADS collisions over time that is more significant compared to ADAS collisions. In terms of collision severity, factors such as the presence of a test driver in ADS-equipped vehicles did not seem to improve safety. In terms of collision location, ADS collisions on intersections show a decreasing trend while street or midblock collisions slightly increased. ADS collisions that occurred at high speeds (≥ 50mph) were caused by the other road users in most cases. Injury rates out of the total ADS collisions were 18.9, 11.1 and 11.6% for turning, lane changing, and proceeding collisions, respectively. Adverse weather conditions had no notable effect on ADS collision severity, while they tend to increase ADAS collisions with injuries by 7.4% compared to clear weather.