Study Of OD Estimation Based On Turning Traffic Flow

The Origin-destination (OD) matrices describe traffic flow between the origin and destination nodes in road network and reflect travelers’traffic demands of traffic network. They are the most important basic data for urban traffic planning, traffic control, and traffic flow forecasting. Conventional methods of OD matrices acquirement are highly expensive and time-consuming. It is an effective way to estimate OD matrix by using link traffic flow and other relevant information.Conventional methods of OD matrix estimation are primarily based on link traffic flow. In a real traffic scenario, due to the numbers of links is much smaller than that of OD pairs, which greatly limits the estimation accuracy of these methods. In this work, turning traffic flow information is introduced to improve estimation accuracy of OD matrix. According to the differences of road network and the known conditions, OD estimation problems are divided into four categories, and the corresponding problem-solving schemes are given in detail.The main work of this dissertation is as follows:1. The principle of OD matrix estimation, which is based on traffic flow, is studied. The principle and method of OD matrix estimation based on turning traffic flow are addressed. The traffic assignment models with respect to different road network are analyzed, and then the appropriate model is chose for the two traffic conditions in road network.2. For non-congested road networks, Multipath probability assignment model is studied and acquirement method of traffic assignment matrix is given. Entropy maximization (EM) model is applied to estimate OD matrix without prior information. Furthermore, the related optimization algorithms are discussed. OD matrix estimation problem with prior OD matrix is solved by using generalized least squares(GLS) model.3. For congested road networks, the user equilibrium (UE) model is studied. The OD matrix was estimated by bi-level programming model. In the case of the lack of historical OD matrix, user equilibrium (UE) model is combined with EM model for problem-solving. Otherwise, GLS model and UE model are integrated to estimate OD matrix.4. The detailed experimental schemes are designed for four types of OD matrix estimation problems, and the results are discussed and analyzed accordingly.