Evaluation of Crash Risk during Large Scale Automated Vehicles (AVs) Deployment through Real-time Crash Prediction

抄録

Successful integration of automated vehicle (AV) technology requires safety evaluation during their large scale deployment. This paper focus on a literature review on safety impact of AVs by examining the behavioral models and parameter settings. For the later research component, traffic and crash data from Route 4 of Tokyo Metropolitan Expressway Company Limited was used to develop a RTCPM using Dynamic Bayesian Network (DBN) and a microscopic traffic simulation environment was reproduced using PTV VISSIM tool to create the environment for AV introduction. The findings from literature review identifies several challenging issues including appropriate parameter setting, lower marker shares of AV, and lower automation level on the existing network. The simulation model outcome suggests that crash risk can be reduced by 11.1% and 13.1% under normal driving and cautious driving behavior respectively with mixed scenarios.