A calibration framework of car following models for safety analysis based on vehicle tracking data from smartphone probes

This study introduces a new methodology for acquiring vehicle tracking data with which to calibrate and validate microsimulation traffic models for safety analysis. Most common approaches are based on video image processing algorithms or the use of roadside Bluetooth detectors. In this paper a procedure is presented that makes use of on-board assisted-GPS equipped smartphone probes supplemented by other location services including Wi-Fi positioning system and cell-site multilateration. The calibration procedure was applied to the VISSIM® software using a genetic algorithm to systematically modify the parameters of car following behaviour model in order to fit vehicle tracking data obtained from simulations to the measured ones. Vehicle tracking data are analysed in terms of rear-end interactions among vehicles in traffic stream; these interactions are expressed by the deceleration rate to avoid a crash, a surrogate safety measure accounting for the speed differential between follower and leader vehicles and their closing time.