HN-VCTM: a variable-length cell transmission model for highway traffic flow inference under sparse sensing using toll data

This research introduces a method for inferring highway traffic flow at detector-less sections, utilizing toll data. The approach processes vehicle entry/exit records to generate origin-destination (OD) matrices. These flows are allocated via an enhanced Logit-based dynamic traffic assignment model, incorporating speed distributions by vehicle type, calibrated using toll data. Central to the method is a newly devised Highway Network Variable-length Cell Transmission Model (HN-VCTM). This model employs topology-aware cell division, aligning cell boundaries precisely with toll stations and gantries for maximal data exploitation. This model allows simulation of flow at any unmonitored location. Raw simulation results from the HN-VCTM, validated on a real 7-day dataset covering 106 toll stations in Yunnan, China, achieved MAPEs of 21.08% on weekends and 21.48% on weekdays. The framework demonstrates how existing toll infrastructure enables large-scale traffic state estimation under sparse sensing conditions, thereby reducing dependency on specialized sensors.