A new methodology for evaluating the accident risk of Level 3 autonomous vehicles

抄録

Before the era of smart mobility there will be a transitional period in which both autonomous vehicles (AVs) and human-driving vehicles (HDVs) share the same road networks in mixed traffic. The main objective of this research is to evaluate and rank the safety performance of the different AV models. The evaluation is based on studying 190 accident records of level 3 AVs operated in the United States of America. In the methodology, a risk index is developed for classifying the different models of AVs by accident exposure. Then, a clustering and an ordinal logistic model have been developed to investigate the main factor affecting the risk of different AV groups. In the results, the AV models have been ranked based on the developed risk index. Moreover, the results showed that the accident risk of the AVs increases with speed, especially at intersections.