Neural network-based car following model with considering the drivers’ spatial anticipation capability

Car following models are considered as one of the most important driving behavior models, which have widely been applied in various fields of transportation engineering such as microscopic traffic simulation and safety. Large number of factors affect drivers’ car following behavior. In practice, due to the lack of appropriate data and latent nature of some factors, only few numbers of them have been considered in model development. Drivers’ ability to spatial anticipation of traffic flow is one of the important factors of car following behavior, which has been neglected in construction of car following models, particularly in neural network-based car following models. In order to consider this capability, a car following model based on neural network has been developed. Speed of the subject vehicle, speed of first and second leader and speed of follower as well as first and second lead gap and first lag gap were considered as input variables while the model’s output is acceleration of the subject vehicle. Calibration results indicate the better performance of the proposed model in comparison to the base model (without spatial anticipation). Additionally, in some situations both models have similar performance. Further .investigations show that this is due to the large correlation between first and second lead gap