Incident detection for freeways based on a dual-state traffic factor state network

The accurate real-time detection of freeway incidents can help decision-makers intervene as early as possible, minimizing the impact of the incident on traffic. This paper proposes a dual-state traffic factor state network (DS-TFSN) that combines the macro traffic state and the micro vehicle driving state. Based on the correlation between macro traffic factors, micro vehicle factors, and external environmental factors, the concept of accidental factors is added to the DS-TSFN; the impact of these accidental factors on traffic factors and the traffic state is analyzed, allowing the accurate determination of freeway system state and incidents. Further, Petri Net is used to describe the changes in the traffic state under normal and incident-influenced states. The results of the case study show that the accuracy of using comprehensive features for incident detection is 98.9%, which is 11.6% higher than when using macro features and 4.5% higher than when using micro features, respectively.