Big data; urban expressway; traffic safety; traffic efficiency
In an age of data explosion, almost every aspect of social activities is impacted by the abundance of information. The information, characterized by alarming volume, velocity and variety, is often referred to as "Big Data". As one fundamental elements of human life, transportation also confronts the promises and challenges brought about by the Big Data era. Big Data in the transportation arena, enabled by the rapid popularization of Intelligent Transportation Systems (ITS) in the past few decades, are often collected continuously from different sources over vast geographical scale. Huge in size and rich in information, the seemingly disorganized data could considerably enhance experts' understanding of their system. In addition, proactive traffic management for better system performance is made possible due to the real-time nature of the Big Data in transportation. Operation efficiency and traffic safety have long been deemed as priorities among highway system performance measurement. While efficiency could be evaluated in terms of traffic congestion, safety is studied through crash analysis. Extensive works have been conducted to identify the contributing factors and remedies of traffic congestion and crashes. These studies lead to gathering consensus that operation and safety have played as two sides of a coin, ameliorating either would have a positive effect on the other. With the advancement of Big Data, monitoring and improvement of both operation and safety proactively in real-time have become an urgent call. In this study, the urban expressway network operated by Central Florida Expressway Authority's (CFX) traffic safety and efficiency was investigated. The expressway system is equipped with multiple Intelligent Transportation Systems (ITS). CFX utilizes Automatic Vehicle Identification (AVI) system for Electronic Toll Collection (ETC) as well as for the provision of real-time information. Recently, the authority introduced Microwave Vehicle Detection System (MVDS) on their expressways for more precise traffic monitoring. These traffic detection systems collect different types of traffic data continuously on the 109-mile expressway network, making them one of the sources of Big Data. In addition, multiple Dynamic Message Signs are currently in use to communicate between CFX and motorists. Due to their dynamic nature, they serve as an ideal tool for efficiency and safety improvement. Careful examination of the Big Data from the ITS traffic detection systems was carried out. Based on the characteristics of the data, three types of congestion measures based on the AVI and MVDS system were proposed for efficiency evaluation. MVDS-based congestion measures were found to be better at capturing the subtle changes in congestion in real-time compared with the AVI-based congestion measure. Moreover, considering the high deployment density of the MVDS system, the whole expressway network is well covered. Thus congestion could be evaluated at the microscopic level in both spatial and temporal dimensions. According to the proposed congestion measurement, both mainline congested segments and ramps experiencing congestion were identified. For congestion alleviation, the existing DMS that could be utilized for queue warning were located. In case of no existing DMS available upstream to the congestion area, the potential area where future DMS could be considered was suggested. Substantial efforts have also been dedicated to Big Data applications in safety evaluation and improvement. Both aggregate crash frequency modeling and disaggregate real-time crash prediction were constructed to explore the use of ITS detection data for urban expressway safety analyses. The safety analyses placed an emphasis on the congestion's effects on the Expressway traffic safety. In the aggregate analysis the three congestion measures developed in this research were tested in the context of safety modeling and their performances compared. Multi-level Bayesian ridge regression was utilized to deal with the multicollinearity issue in the modeling process. While all of the congestion measures indicated congestion was a contributing factor to crash occurrence in the peak hours, they suggested that off-peak hour crashes might be caused by factors other than congestion. Geometric elements such as the horizontal curves and existence of auxiliary lanes were also identified to significantly affect the crash frequencies on the studied expressways. In the disaggregate analysis, rear-end crashes were specifically studied since their occurrence was believed to be significantly related to the traffic flow conditions. The analysis was conducted in Bayesian logistic regression framework. The framework achieved relatively good classifier performance. Conclusions confirmed the significant effects of peak hour congestion on crash likelihood. Moreover, a further step was taken to incorporate reliability analysis into the safety evaluation. With the developed logistic model as a system function indicating the safety states under specific traffic conditions, this method has the advantage that could quantitatively determine the traffic states appropriate to trigger safety warning to motorists. Results from reliability analysis also demonstrate the peak hours as high risk time for rear-end crashes. Again, DMS would be an essential tool to carry the messages to drivers for potential safety benefits. In existing safety studies, the ITS traffic data were normally used in aggregated format or only the pre-crash traffic data were used for real-time prediction. However, to fully realize their applications, this research also explored their use from a post-crash perspective. The real-time traffic states immediately before and after crash occurrence were extracted to identify whether the crash caused traffic deterioration. Elements regarding spatial, temporal, weather and crash characteristics from individual crash reports were adopted to analyze under what conditions a crash could significantly worsen traffic conditions on urban expressways. Multinomial logit model and two separate binomial models were adopted to identify each element's effects. Expected contribution of this work is to shorten the reaction and clearance time to those crashes that might cause delay on expressways, thus reducing congestion and probability of secondary crashes simultaneously. Finally, potential relevant applications beyond the scope of this research but worth investigation in the future were proposed.
If this is your thesis or dissertation, and want to learn how to access it or for more information about readership statistics, contact us at STARS@ucf.edu
Doctor of Philosophy (Ph.D.)
College of Engineering and Computer Science
Civil, Environmental, and Construction Engineering
Engineering and Computer Science
Length of Campus-only Access
Doctoral Dissertation (Open Access)
Dissertations, Academic -- Engineering and Computer Science; Engineering and Computer Science -- Dissertations, Academic
Shi, Qi, "Urban Expressway Safety and Efficiency Evaluation and Improvement using Big Data" (2014). Electronic Theses and Dissertations. 725.