Heterogeneity of Driving Behaviors in Different Car-Following Conditions

Abstract

Many application fields in transportation engineering can benefit from an accurate modelling of car-following behavior. In particular, in recent years, an increased importance is assigned to embed behavioral abilities in ADAS (Advanced Driving Assistance Systems) and in driving automation solutions. However, accurate development of car-following models needs for accounting of the drivers’ heterogeneity, which can be easily observed in car-following data. This paper contributes to analyze different sources of heterogeneity with particular focus on three factors: the dispersion over-time of the behavior of a single driver; the heterogeneous behaviors of different drivers; and the possible bias introduced by some over-simplification of the modelling framework, with particular reference to the type of leading vehicle. Our analyses are based on the observation of car-following trajectories collected in a large experiment involving one hundred drivers. Observed behaviors have been interpreted by means of several car-following models proposed in past. The comparison of the values of the parameters identified for the models (versus observed data) is adopted for the analyses. Moreover, directly observed variables (car-following speed and spacing) are adopted to complement and confirm the analyses. Results show that the greater among the sources of dispersion is the across-driver heterogeneity and that by taking into account such an inherent drivers’ dispersion of car-following behaviors it is possible to better identify also the effect of the modelling oversimplifications induced by not considering the type of leading vehicle.