Stratified Traffic Congestion Measurement Models And Algorithms For Megacities

Due to the mega-scale nature, huge population size, and complicated network structures, megacities have demonstrated unique but different traffic congestion phenomena and patterns from small and/or medium-sized cities. In the field of traditional traffic congestion measurement, numerous models have been developed for cities. However, few have been developed specifically for megacities. After a comprehensive synthesis of state-of-the-art in congestion measurements and an analysis of traffic flow characteristics, the research in this dissertation developed stratified traffic congestion measurement models for megacities by applying statistical and signal processing techniquesIn order to develop the measurement models, the dissertation first defined four dimensions for measuring congestions for megacities, including congestion intensity, congestion duration, congestion breadth, as well as congested segments and intersections. A stratified traffic congestion measurement framework for megacities, including macroscopic, mesoscopic, as well as microscopic levels, was developed for determining the basic functions of congestion measurement models and their inter-relationships. The models were developed as follows:First, macroscopic models for measuring the congestion intensity were developed. By introducing the cumulative logistic regression theory into the measurement of congestion intensity for road segments, a travel-speed based measurement model for the congestion intensity was presented. Then, a weighted vehicle-hours-traveled (VHT) based traffic congestion indices (TCI) model was developed by using stratified sampling and comprehensive evaluation method. Second, mesoscopic models for measuring the congestion duration and breadth were developed. After analyzing the relationship between the independent and dependent variables in the measurements of temporal and spatial congestions, the non-linear regression technique was used to develop the travel-speed based models for the measurement of severe congestion duration (SCD) and congestion duration (CD) for a specific road class, in which the exponential and cubic functions were selected for SCD and CD models respectively. Further, multivariate linear regression techniques were adopted to develop models for the measurement of severe congestion breadth (SCB) and congestion breadth (CB), in which a bivariate linear regression of travel-speed and stop-time was used in the SCB model, and a simple linear regression of travel-speed was used in the CB model. Finally, microscopic models for the identification of key congested segments and intersections were developed. Identification algorithms were developed in light of an analysis of the typical traffic congestion characteristics of the key congested segments and intersections. In the modeling process, the wavelet transformation was used to extract the approximate components of travel speeds, which were used in the identification of periodical congestion characteristics. The method to identify the primary congestion was proposed based on an analysis of characteristics of the spatial-temporal distribution of congestions. The characteristics of both periodical and primary congestions were used to identify the key congested segments and intersections.The validity and accuracy of proposed models were tested by using the statistical test methods and by comparing the model estimation with the field observed data. The test demonstrated that the proposed models and algorithms can be readily used to evaluate traffic congestions for megacities.