Research On On-Ramp Metering Of Urban Expressway Based On Support Vector Regression

With the rapid expansion of city scale and the excessive concentration of population, traffic congestion in metropolitan has become the main factor of restricting urban development. The construction of urban expressway effectively alleviates the traffic congestion. However, with the rapid increasing of motor vehicles, the method of only building roads can’t completely solve the problem of traffic congestion. As one of the most practical and effective methods, on-ramp metering of urban expressway is carried out in this thesis. Based on the study of traffic flow model, support vector machine algorithm and genetic algorithm, on-ramp metering model based on support vector regression machine is established.First of all, the basic concepts and principles of on-ramp metering of urban expressway are systematically expounded, and the dynamic density model, dynamic flow model and dynamic velocity model are studied. On this basis, the ALINEA model and the on-ramp metering model based on BP algorithm are in-depth studied, and the advantages and disadvantages of these two on-ramp metering model are summarized.Secondly, the basic concepts, key ideas and concrete solving process of support vector regression machine are researched and elaborated, and the selection and optimization problem of SVR model’s parameter is studied. After thorough research of genetic algorithm, the selection and optimization algorithm of SVR model’s parameter is constructed.Thirdly, on the basis of researching of short-term traffic density prediction, a traffic density forecast model based on SVR algorithm is constructed. Combined with SVR traffic density forecast model, three on-ramp metering models based on support vector regression machine are established according to the difference of the input variable.Finally, on the basis of researching on traffic flow data acquisition method, traffic flow data of on-ramp on the east side of Anhui bridge were collected. Traffic density forecast model and on-ramp metering model based on SVR algorithm are realized and analyzed by MATLAB, and the simulation results prove the feasibility and effectiveness of those models.