Experimental determination of factors causing crashes involving automated vehicles

Emergence of technologies to replace human action is occurring in many sectors, with autonomous vehicles being a leading example. Autonomous vehicles do not require human interaction and instead employ various devices to perform essential operations. This paper assesses factors which cause autonomous vehicles to suffer crashes, using field data collected by the Californian Department of Motor Vehicles. Data on these highly automated vehicles (AVs) were clustered based on degree and direction of impact, and analyzed by coding in Excel and RStudio programming. A novel feature of the work is that all clustering, analysis, application of association rules, and determination of degrees of severity of crashes were done by RStudio programming and that the direction of autonomous vehicles impacts was identified based on field data. Our analysis reveals that weather conditions, maneuvering, road conditions, and lighting are major factors in autonomous vehicles crashes. Rear-end crash and minor scratches to autonomous vehicles are the most frequent forms of damage, based on the available data. This study underscores the critical need for enhanced sensor technologies and improved algorithms to better handle adverse weather conditions, complex maneuvers, and varying road and lighting conditions. By identifying the most frequent types of damage, such as rear-end crashes and minor scratches, this research provides valuable insights for manufacturers and policymakers aiming to improve the safety and reliability of autonomous vehicles. The findings can inform future design improvements and regulatory measures, ultimately contributing to the reduction of crash rates and the advancement of autonomous vehicle technology.