Are we ready for highly automated vehicles? Evaluating automated driving systems’ safety and reliability using inverse probability weighted regression adjustment

Automated Driving Systems (ADS) offer promising benefits in enhancing road safety, mobility, and operational efficiency. However, their integration into real-world traffic systems introduces critical safety challenges, particularly in remote-controlled and fully autonomous configurations. Although the observed proportion of injuries in ADS-involved crashes (17.59%) is lower than in conventional traffic crashes (40%), the persistence of injury risk, given an ADS crash, necessitates further investigation. This study analyzes 705 ADS crash reports (July 2021–July 2024) from the National Highway Traffic Safety Administration’s dataset, applying the doubly robust Inverse Probability Weighted Regression Adjustment (IPWRA) method to address potential selection bias. Findings reveal that crashes involving remote drivers/operators are associated with an 18.3% higher likelihood of injury, while fully autonomous configurations correspond to a 23.3% increase, relative to scenarios involving both in-vehicle and remote operators. These findings underscore ongoing challenges in managing safety-critical events under reduced human oversight. Intersections are associated with significantly elevated injury risk, 22.3% in remote and 18.7% in no-driver scenarios, reflecting their operational complexity. Environmental conditions, including adverse weather and dark lighting, contribute to 10.2% to 19.2% higher injury risk. Crashes involving Vulnerable Road Users (VRUs) yield the most pronounced effects, with a 22.2% increase observed in no-driver configurations. Subject vehicle higher pre-crash speeds and front/rear-end collisions are also strongly associated with elevated injury risk. These findings highlight the need for improvements in ADS sensing, decision-making, and control systems. Policymakers and manufacturers must prioritize safety protocols tailored to complex road conditions to ensure safe ADS integration into traffic systems.