Latent class model for car following behavior

Abstract

Car-following behavior, which describes the behavior of a vehicle while following the vehicle in front of it, has a significant impact on traffic performance, safety, and air pollution. In addition, car-following is an essential component of micro-simulation models. Over the last decade the use of microscopic simulation models as a tool for investigating traffic systems, ITS applications, and emission impacts, is becoming increasingly popular. The paper presents a flexible framework for modeling car-following behavior that relaxes some limitations and assumptions of the most commonly used car following models. The proposed approach recognizes different regimes in driving such as car-following, free-flow, emergency stopping, and incorporates different decisions in each regime, such as acceleration, deceleration, and do-nothing depending on the situation. A case study using NGSIM vehicle trajectory data is used to illustrate the proposed model structure. Statistical tests suggest that the model performs better than previous models.