Travel Time Estimation And Prediction For Urban Road Networks

The urban traffic guidance system (UTGS) is an important component of the intelligent transportation system (ITS). By use of the travel time estimation and prediction of urban road networks, the UTGS could provide the best driving route for traffic participants. Thus everyone’s travel time could be saved, and the traffic congestion could be relieved thereby.However, unlike other traffic control system, UTGS is not supported by mandatory regulations. That makes the ability to accurately estimate and predict travel time in urban transportation networks essential to the UTGS. Considering the travel time parameter as stochastic variable, we propose an expected travel time estimation and prediction algorithm innovatively. The main contributions of the dissertation in this paper are summarized as follows:Basically, missing and corrupted raw traffic data are unavoidable in real transportation system, affecting the accuracy of the entire follow-up studies. To improve the qualities of the traffic data, pre-processing which includes data filtering and imputation must be used. Focusing on traffic flow data and based on the main error analysis, a three-step screening method is presented; and in according to the spatial and temporal correlation of the traffic flow data, a data recovering method employs time series analysis and multivariable linear regression is proposed.Secondly, based on the analysis of the characteristics of urban road networks, an expected travel time estimation algorithm is proposed. According to the moving process of the vehicle in the urban road, the link-based travel time is divided into three parts: the time spent on the road (without the signal intersection), the delay spend in intersection and the time needed to traverse the intersection. To model the stochastic system, queue theory and the bureau of public roads (BPR) model are used.Finally, in order to prepare for the route choice, a path-based travel time prediction algorithm is proposed. To conquer the problems of the absence of traffic data in the future and the uncertainty of the arrival time, the concepts of several-step prediction and scrolling addition are applied.