Flexible car-following models incorporating information from adjacent lanes

In recent years, technological advances have significantly improved Driver Assistance Systems and there has been an increasing interest in autonomous vehicles. Aiming at safety, reliability and convenience, autonomous vehicles require detailed car-following models that could model driving behavior in an efficient way. In this research, an existing flexible car-following model is enriched by incorporating additional information about density of two adjacent lanes. This research aims to explore if the additional information on density of adjacent lanes could improve the accuracy of the car-following model. More realistic detailed models could provide a robust solution to autonomous driving. The updated model is applied to reconstructed NGSIM data using a flexible regression technique, loess method. For a more in depth analysis, a meta- model is developed to evaluate the magnitude of the effect of the considered predictor variables on the proposed model. Finally, conclusions are drawn and future prospects are suggested.