Research On Traffic Condition Prediction Of Urban Expressway

Traffic control and guidance system is a core subsystem in ITS, and the key issue to implement this is how to make a real-time and precise traffic condition prediction. Unfortunately, the traffic flow is an intrinsically nonlinear system, and it has a strong correlation with the historical data and upstream traffic flow. Actually, it is very difficult to describe the traffic flow using precise mathematic model, so this thesis hopes to use other technology to research. Because of their good capability for prediction, Kalman filter theory and BP neural network theory are used to forecast the traffic volume, speed and occupancy, based on the analysis of the traffic characteristic of urban expressway. In the thesis, the author uses real traffic data to prove their prediction capability, and also compares their performance. Through experiments, it can conclude: 1) Both of them have good performance on traffic parameter prediction; 2) Kalman filter model is very simple to apply on-line; 3) BP neural network model has a better prediction performance than Kalman filter model, but it is complex to realize and can't be used on-line. In the thesis, these models make good use of historical data and upstream traffic flow to predict traffic volume, speed and occupancy. Both of these two algorithms are verified with real traffic data. The results show that these methods are feasible and effective, and they could be used in real-world applications.