Lane-level routable digital map reconstruction for motorway networks using low-precision GPS data

• We identify nodes, construct centerlines and determine lanes from smartphone GPS. Smooth, segment-based centerline unbiased by GPS density distribution over lanes. Accurate road width based on an adapted two-stage local outlier factor criterion. Compatible highly accurate estimation of using adaptation GMM. Providing a platform for the extraction lane-level traffic data. The construction routable digital maps trajectory data has attracted lot attention, especially in recent years, with ease cheapness collecting required Such maps, if they are constructed at lane-level, have many applications analysis, study driving behavior floating car In this paper, we present three-step automatic method QuickBundles node detection network, dissimilarity matrix Fréchet distance construction, Gaussian Mixture Method lane estimation. results smooth, as well compatible accuracy, connectivity, compatibility, validity, robustness proposed been tested various ways. paper show that method, while being low cost, can be used extracting longitudinal lateral behavior, drivers' lane-changing maneuvers. Due to unique features remains constant along entire length segments. addition, validation metrics shows obtain reliable estimates speed volume flow different Being derived actual trajectories, inferred markings may deviate physical ones driver population systematically deviates laterally, instance curves; depending application use case resulting map, desirable or undesirable feature.