Calibration of Acceleration-Based and Multi-Anticipative Car-Following Models by NGSIM Trajectory Data

Recently, the Next Generation Simulation (NGSIM) has attracted many researchers' interests because of its representative data and rich contents. Based on the NGSIM data, this paper calibrates the multi-anticipative car-following models, including multi-anticipative IDM, multi-anticipative FVDM, MHVD model, and MVD model. Before calibration, this article pre-processes the accelerations and velocities since there are several defects in the original data. The calibration of car-following models and multi-anticipative car-following models is grounded on the Levenberg-Marquardt algorithm; furthermore, the calibration results of car-following models can provide important reference value for researchers, such as the appropriate estimating ranges of model parameters, while the results of calibrating multi-anticipative car-following models not only introduce the parameter estimation ranges but also testify that, in reality, the multi-anticipative models are more suitable and two preceding vehicles can be a better choice.