The HighD Dataset: Is This Dataset Suitable for Calibration of Vehicular Traffic Models?

摘要

A large-scale naturalistic vehicle trajectory dataset from German highways called highD is used to investigate the car-following behavior of individual drivers. These data include trajectories of 1,10,000 vehicles with the total length of 16.5 h. Solving a nonlinear optimization problem, the Intelligent Driver Model is calibrated by minimizing the deviations between observed and simulated gaps, when following the prescribed leading vehicle. The averaged calibration error is 7.6%, which is a little bit lower compared to previous findings (NGSIMI-80). It can be explaind by the shorter highD trajectories, predominantly free flow traffic and good precision metrics of this dataset. The ratio between inter-driver and intra-driver variability is inversigated by performing global and platoon calibration. Inter-driver variation accounts for a larger part of the calibration errors than intra-driver variation does.