Arterial Traffic Flow Estimation Based on Vehicle-to-Cloud Vehicle Trajectory Data Considering Multi-Intersection Interaction and Coordination

Abstract

Conventional detection methods for intersection traffic flow heavily rely on fixed-location inductive loop, video image processing, infared, and microwave radar detectors. The emerging connected vehicles (CV) technologies can potentially reduce such dependencies on conventional vehicle detectors with the vehicle-to-cloud (V2C) CV data. This paper proposes an analytical method for traffic flow estimation in urban arterial corridors based on CV trajectories collected through V2C communication. Different from the existing single-intersection models, the proposed model considers traffic states and the traffic signal coordination among adjacent intersections, therefore, can capture the delay and queuing dynamics in arterial corridors. The queue spillback phenomenon is explicitly considered by applying the shockwave theory. The proposed model is evaluated based on real-world vehicle trajectory data from the DiDi platform collected on an arterial network in Chengdu, China with a penetration rate of less than 10% of the overall traffic. The flow estimation results are compared with traffic counts collected from video detectors. The model parameters are calibrated with more than 300,000 GPS points during a typical workday and tested on a different workday. The evaluation results show a mean absolute percentage error within the range of 4–7% among all intersections, outperforming the results generated by the existing single-intersection model. The results indicate the promising potential of using the proposed methods to evaluate intersection performance without heavy investment in on-site detectors.