Spatiotemporal trajectory characteristic analysis for traffic state transition prediction near expressway merge bottleneck

摘要

The theoretical analysis of traffic flow with empirical vehicle trajectory data contained within this study allows for the explanation, reconstruction, and prediction of spatiotemporal transition characteristics of traffic conditions. Using an unmanned aerial vehicle (UAV) during a morning rush hour on a working day, observations of congestion evolution near an on-ramp bottleneck of an expressway was captured. The empirical high-fidelity trajectory data of 621 vehicles were extracted. The major findings include:(1) Macroscopic perspective: Three traffic states (free flow, metastable traffic flow, and jam) were observed with different spatiotemporal physical structures and their critical characteristics in a continuous first-order phase transition. The spontaneous spatiotemporal traffic breakdown and capacity drop were also captured, the critical points of which were also recorded. The features of widening synchronized flow pattern, general pattern, and the nature of nucleation described in the three-phase traffic theory were identified, which provides a new observational cognition for congestion formation and development.(2) Microscopic perspective: The impact of lane changes on immediate vehicles were modeled and quantified with the fluctuation magnitude of speed and the theoretical lag distance, which provides a method to identify the source of perturbation and determine the microscopic critical threshold of vehicle between macroscopic phase transitions. The research shows a lane changer could cause a forced deceleration of an immediate vehicle on a target lane after inserting itself with critical spacing headway below approximately 15-20 m. The average deceleration duration of vehicles in different traffic states can be captured as crucial driving features from downstream to help predict traffic state transition in a real bottleneck.