In mixed traffic, the decision-making and/or control of connected automated vehicles (CAVs) largely depends on accurate description and prediction of HVs’ behaviors. To better understand the behavioral heterogeneities of HVs, the paper proposes a method to distinguish car-following behaviors in specific leader-follower contexts. The car-following data of the Next-Generation Simulation (NGSIM) dataset are classified into four leader-follower compositions, namely, truck-car (T-C), car-car (C-C), car-truck (C-T), and truck-truck (T-T). Based on the calibration results of a few well-known car-following models, principal component analysis (PCA) and clustering analysis are applied to the calibrated parameters to discover the behavior patterns and to find the probabilistic distributions of the parameters for the classified car-following (CCF) models. Simulation results show that compared to the unified car-following (UCF) models, the estimation error of calibrated CCF models is reduced by 17.96%-59.32%, which indicates that the proposed method provide a more accurate description of car-following heterogeneities.

CICTP