Investigations On State Forecasting Methods Of Urban Traffic Networks

In recent years, China was spending efforts on improving the transport infrastructure. However, with fast-developing economic, it could not meet the increasing traffic demand only by strengthening the transport infrastructure construction or improving the traffic network capacity, and that will lead to traffic jams or even crippling traffic disasters. ITS(Intelligent Transport Systems) is a new-style management system which makes full use of new and high technology to coordinate relationship of the drivers, the vehicles and the roads. It is composed of Advanced Transportation Information Service System(ATIS), the Advanced Traffic Management System(ATMS), etc., which are used to realize traffic guidance and traffic control. For both traffic flow guidance system and traffic control system, the state forecasting is an important foundation. Under such a background, this dissertation investigates the state forecasting methods of urban traffic networks(including urban expressway networks and urban surface way networks), and also the feasibility and effectiveness to combine with Model Predictive Control to apply on traffic control systems. We hope through these efforts, these methods could be used into actual transport guidance and control applications. The main work of this dissertation can be summarized as the following aspects:1) Investigation of forecasting method of urban expressway single road, based on data-driven forecasting method. This investigation makes use of the vehicle velocity historical data of Shanghai South North Viaduct, which is provided by Shanghai Traffic Information Center, and the Hybrid Particle Swarm Optimization Neural Network with error compensation mechanism, to forecast vehicle velocity in future minutes.2) Investigation of forecasting method of urban expressway network, based on the Cell Transmission Model(CTM). This investigation develops an optimized algorithm of the original CTM, through the off-ramp diversion coefficient online computing instead of setting it as a fixed value. This optimization could improve the forecasting accuracy of the original CTM.3) Investigation of forecasting method of urban surface way network, based on the CTM theory. Original CTM improves from modelling a single road to a network which contains at most three-legged junctions, but it will face some problems describing a network contains four-legged junctions, thus it could not apply on modelling of urban surface way networks. This dissertation proposed an extension algorithm of the original CTM, which is called Cell-link Modelling(CLM). The CLM could apply on the urban surface way networks based on CTM theory.4) From the control aspect, analyze the feasibility and effectiveness of CLM applying on traffic control system using Model Predictive Control. Compare the CLM to the classical S model, in order to illustrate the CLM’s superiority in computation complexity aspect.