Wheel Tracking and Vehicle Identification Using Road-Embedded Sensors

The growing volume of freight traffic and decreasing infrastructure availability due to maintenance emphasize the need for smart road systems such as digital twins to support data-driven road asset management. In this context, we present a framework for tracking individual vehicle wheels and assigning them to their corresponding vehicles using only road-embedded sensors. The approach ultimately enables high-resolution mapping of traffic-induced loads, as required for digital twins. The method relies on timestamps and positions of distributed sensor detections and proceeds in three steps: clustering sensor signals, reconstructing wheel trajectories via binary optimization, and rule-based vehicle identification based on geometric and kinematic consistency. The method targets key characteristics of real-world traffic data, including unknown vehicle geometries and varying speeds. In the evaluated dataset from a research-grade Weigh-in-Motion (WIM) installation on a German highway, the method achieved 100% precision and recall in wheel tracking and near-perfect results in vehicle identification. The method remains robust under dynamic vehicle behavior and up to 5% artificial data loss, demonstrating its potential for infrastructure monitoring and load-informed digital twins of the road.