Urban Traffic Flow Forecasting Model of Double RBF Neural Network Based on PSO

Abstract

The real time adaptive control of urban traffic, as a complex large system, usually needs to know the traffic of every intersection in advance. So traffic flow forecasting is a key problem in the real time adaptive control of urban traffic. This paper's research object is two typical adjacent intersections of city road. A double RBF NN model with classifying coefficient is presented. The space of high dimensional input samples is divided into two lower dimensional subspaces by the model. Then the nonlinear degree of the space samples is reduced greatly. Particle swarm optimization (PSO) algorithm is used to determine the parameters of two RBF NN respectively. The method not only simplifies the structure of RBF NN, but also enhances training speed and mapping accurate. The simulation results show the effectiveness of the model