An integrated framework for multiple traffic anomalies detection on highways using vehicle trajectories

Fast and accurate identification of traffic anomalies on highways is of utmost importance. This study presents an integrated framework for multiple traffic anomaly detection on highways using vehicle trajectories. The framework addresses both macroscopic congestion patterns and microscopic driving behaviors, offering a comprehensive solution that simultaneously detects multiple anomalies within a unified framework. The developed framework comprises three main components: data acquisition and preprocessing, vehicle trajectory recognition, and traffic anomaly detection. The former two components are responsible for acquiring real-time vehicle trajectories on highways. With such trajectory information and the continuously monitored short-term traffic state, the latter component seeks to simultaneously detect all the traffic anomalies via a tailored sub-algorithm for each of them. For macroscopic anomaly detection, an algorithm for detecting stop-and-go waves by constructing localized shockwaves is proposed to capture the propagation of traffic congestion waves even in limited field-of-view scenarios. For microscopic anomaly detection, a dynamic background traffic state updating mechanism is introduced, allowing the framework to adaptively integrate historical traffic data and environmental factors. Additionally, a double-layer stacking framework based on unsupervised methods is designed to integrate diverse feature types and addressing perspective distortions. The developed framework is tested in experiments on both simulation and real-world data on highways. The results confirm its effectiveness in the simultaneous detection of multiple traffic anomalies within an integrated framework.