Development Of Models And Algorithms For Estimating Time-Varying OD Demands Based On Dual Sources Of Data

Origin-Destination (OD) demands are the most basic data for traffic and transportation planning. With the rapid increase in motorization and traffic congestion, advanced traffic and transportation planning, management and control strategies are becoming important measures to ease the traffic congestion, while the derivation of time-varying OD demands is the essential data basis in the evaluation and optimization of these measures. As such, extensive effort has been made on studies associated with the estimation of time-varying OD demands by various scholars. However, due to the complexity of the topology of the urban road network, the varied behaviors of route choices, and the lack of adequate real-world data, current time-varying OD demand models have been mainly developed based on the stationary source data. There is still a lack of effective application-oriented time-varying OD estimation models incorporating multiple source data.In this context, it develops time-varying OD demands models and algorithms of urban road network based on dual sources data. The dissertation first synthesizes state-of-the-art from the perspectives of main factors associated with the OD estimation, which include OD estimation models and its related problems. Then, it analyzes the characteristics of stationary source data and moving source data and the advantages of OD estimation combining two sources of data, and further provides the characteristics of time-varying OD demands model of urban road network. Based on the above analysis, a framework of application-oriented time-varying OD demands model of urban road network combining two sources of data is presented, in which time-varying splitting rate is introduced. In this framework, three key issues are provided including the data quality control models for RTMS data and Floating Car Data (FCD), nonlinear programming time-varying OD model based on combined two sources of data, and the sampling size and algorithm analysis of time-varying splitting rates. With respect with the data quality control models, the RTMS data quality control model based on a linear interpolation Floyd route mending model based on dynamic travel time, and OD estimation method of FCD are proposed. With respect to the nonlinear mathematical programming time-varying OD estimation model, the OD estimation model is developed that combines the consideration of three factors:link flows, OD demands and travel time. In this model, the weight value determination algorithm based on ideal point method and heuristic algorithm based on modified GENOCOPIII are proposed. With respect to the sample size analysis and the algorithm for the time-varying splitting rates, a sampling model based on the interval estimation of means from normal populations and the algorithm for determining the time-varying splitting rates under different data conditions are developed.The proposed models are applied to estimating time-varying OD demands of the Western 3rd Ring-Road corridor network in Beijing. A quantitative analysis of estimated OD demands, traffic flow, and travel time at specific sites and for network is conducted using the indices such as correlation coefficients and mean relative errors. The results show that the proposed time-varying OD demands model is capable of effectively estimating the dynamic OD demands of urban road network. Finally, the raw GPS data are further explored through data mining techniques to obtain more data for OD estimation, and recommendations are provided in terms of application of the proposed model as well as the online OD estimation.