Machine Learning-Based Traffic Breakdown Prediction Utilizing Connected and Automated Vehicles
(Room Palms Ballroom - Booth 100)
16 Jul 19
5:30 PM
-
7:00 PM
Traffic breakdown prediction is a major component of traffic management system. Early prediction of traffic breakdown will allow traffic management agencies to activate operational plans to prevent, delay, or mitigate the adverse impacts of breakdown. Previous studies have focused on utilizing macroscopic traffic measures to predict breakdown. Connected vehicle data will provide much more detailed data at the individual vehicle data that have the potential to allow better prediction of breakdown. This study investigates the use of microscopic features along with macroscopic features to estimate the probability of breakdown in mixed traffic of human-driven vehicles and connected and automated vehicle (CAV). These microscopic features are: standard deviation of individual’s vehicle speed, number of oscillation (stop-go) and time exposed time to collision (TET). The impact of the introduction of these microscopic parameters on predicting breakdown is assessed by comparing the results with those obtained using only macroscopic measures.