Modeling adaptive cruise control vehicles from experimental data: model comparison

This article uses experimentally collected car following data from seven different commercially available adaptive cruise control (ACC) vehicles to calibrate microscopic models for each system's car following behavior using three different common car following models. Calibration is conducted by selecting the model parameters that minimize the error between the simulated vehicle trajectories and the experimental data. The goal of this study is two-fold: (i) assess which car- following models typically used to describe human driving behavior are best for describing ACC car-following dynamics, and (ii) provide best-fit calibrated car following models for seven different commercially available ACC vehicles, which can be used to understand the traffic flow impact of ACC systems via simulation analysis. We find that the intelligent driver model and the optimal velocity model with a relative velocity term perform best, and with similar performance to one another, while the Gazis-Herman-Rothery model as calibrated does not capture all the ACC car following dynamics.