Impact of autonomous vehicles on urban traffic flow at the urban network level: an analysis using real autonomous driving data

Abstract

This study evaluates the impacts of Autonomous Vehicles (AVs) on traffic flow at the urban network level. Real-world AV data collected by Waymo was used to express car-following behavior, which includes cases of AVs following Human-driven vehicles (HVs) and vice versa. The parameters of intelligent driver model, which is one of the car-following models, were calibrated based on the data by using the DIRECT+SQP algorithm. The impacts of AVs on traffic flow were tested based on the scenarios varying by AV's Market Penetration Rate (MPR) and traffic demand in the microscopic simulation of urban networks in Daejeon City, South Korea. The calibration result reflects that AVs display smoother speed variations, longer time headway, and larger minimum gap compared to HVs. Because of such AV's driving behaviors, the road capacity deteriorates by approximately 35% when all vehicles are AVs in the simulation analysis. Interestingly, the results of deterioration in traffic flow as increases in AV's MPR contrast with the conducive effects of AV's introduction asserted in the previous study. This finding reveals that improvements can only be expected if future AVs surpass the driving performance of current human drivers and highlight the necessity for continued advancements in driving technology, connectivity, and safety. This study can be a stepping stone to making traffic operational strategies to prepare for future modal shifts to AVs.