Traffic Flow Prediction And Control In City Freeway

This thesis consists of two parts: traffic prediction and traffic control.In the traffic prediction part, firstly it analyses the linear and non-linear features of traffic-flow and the advantage of traditional traffic prediction models. Then according to the theory of data-fusion, several relevant time series are constructed. A new prediction approach is proposed by fusing the result from prediction based on the linear part of the traffic time series by ARIMA model and exponential smoothing model and the result from prediction based on the non-linear part of the traffic time series by BP neural network model. Case studies show that the data fusion approach can improve the prediction accuracy compared with using ARIMA or neural network model alone.In the traffic control part, firstly it analyses the features of freeway on-ramp control model, variable speed-limit control model and freeway on-ramp income model. Secondly, according to the theory of coordination control, it's promoted that a new multi-objective optimal control model including average traffic flow density, average traffic flow speed and on-ramp income based on above three traffic control models. The new model is proved more effective by simulation example. Finally, some simple methods for calculating some factor approximately of the new model are presented.