Skip to Main Content
Car-Following model is a basic model in traffic microscopic simulation, using to analyze and describe the way one vehicle (driver) follows its leader in a single lane of traffic. In the past the collection of car-following field data was limited almost exclusively to test tracks or driving simulators, information of drivers on "open roadway" were not included, so car-following models were not formally calibrated or validated. A car-following decision support model is developed in this paper using an error back-propagation neural network (ANN) which has three level neural units and uses four variables, DS, RS, Vn+1, and DV as its input. The outcomes of the model are the accelerations or decelerations of the following vehicle which represent the reaction of the following driver. The data samples for model training and test are collected using "Five-Wheel System".