Study On Car-following Model Based On Artificial Neural Network

Car-following model is an important part of the traffic flow theory. It is important for the study of traffic safety, transportation management, capability of road transportation and service level. The ultimately object of this thesis is to construct a car-following model which can well and truly reflect the traffic flow character of urban expressway in Beijing. The main study in this thesis is listed as followed:The car-following models that have existed are analyzed and classified. Then the technology route of this study is confirmed: On the base of practice, depending on high precision instrument to collect car following data, exercising scientific theory methodology, combining with computer simulation.Plenty of car-following behavioral data is collected with the data collection method using the instrumented GPS vehicles. After data processing, abundant traffic flow data about car-following can provide data base for the construction of car-following model.With statistical technique, the driver reaction time is gained on base of the collected data. It affords parameter for the car-following model.Input variables and output variable is decided according to study of car-following theory. Then car-following models based on BP and RBF neural network are designed and trained. The data used for train network is get from experiment data.Car-following model based on ANN and traditional dynamics are simulated on computer. According to the simulation result, the models are compared and appraised.Through the study, the conclusion can be drown that the models build in this paper can afford a feasible and efficient model for the traffic flow study and micro-simulation of urban expressway.