Identifying Vehicle Turning Movements at Intersections from Trajectory Data

Abstract

Automatically generating performance measures for signalized intersections is important for agencies to assess how individual movements perform at various intersections. This information can be used to prioritize maintenance activities, capital investments, or conduct quantitative before/after assessment. Recently, various studies have utilized connected vehicle trajectory data to generate performance measures. However, most of the trajectory-based performance measures techniques require map matching, road geofencing, and/or geospatial references to identify the movements that individual vehicles follow at a signalized intersection. These labor-intensive movement identification methods diminish scalability since geographic references need to be defined at the specific location where an analysis is intended. This study utilizes connected vehicle trajectory data to automatically identify individual vehicle turning movements at signalized intersections. Entry and exit trajectory headings at an intersection are obtained to generate movement clusters that are then processed using k-means to automatically detect number of clusters and centroids to then assign a particular turning movement. Over 1.1 million trajectories were analyzed with the proposed methodology on a location south of Indianapolis in July 2020. Individual trajectory turning movements were successfully identified at signalized intersections, with matching accuracies of up to 98% with a common road geofencing method.