Some statistical techniques for analysing Bluetooth tracking data in traffic modelling : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Statistics at Massey University, Palmerston North, New Zealand

Thumbnail Image
Open Access Location
Journal Title
Journal ISSN
Volume Title
Massey University
The Author
The economy and the environment are both affected by traffic congestion. People spend time stuck in traffic, which limits their free time. Every city's road infrastructure is under increased pressure, particularly in large cities, due to population growth and vehicle ownership patterns. Therefore, traffic control and management are crucial to reducing traffic congestion problems and effectively using existing road infrastructure. Bluetooth is a commonly used wireless technology for short-distance data exchange. This technology allows all mobile phones, GPS systems, and in-vehicle applications such as navigation systems to connect with the personal devices of drivers and passengers. A Media Access Control (MAC) address is a unique electronic identifier used by each Bluetooth device. The concept is that, while a Bluetooth-equipped device travels along a road, its MAC address, detection time, and location can be detected anonymously at different locations. Bluetooth technology can be integrated into Intelligent Transportation Systems (ITS) to enable better and more effective traffic monitoring and management, hence reducing traffic congestion. This thesis aims to develop some statistical methods for analysing Bluetooth tracking data in traffic modelling. One of the challenges of using Bluetooth data, particularly for travel time estimation, is multiple Bluetooth detections, which occur when a Bluetooth sensor records a Bluetooth device several times while it passes through the detection zone. We employ cluster analysis to look at the possibility of extracting meaningful traffic information from multiple detections, and the observed gap distribution, which is the time difference between records when multiple detections occur. We also develop a novel regression method to investigate the relationship between data from Bluetooth and Automatic Traffic Counts (ATCs) through weighted regression analysis, in order to explore potential causes of bias in the representativeness of Bluetooth detections. Finally, we seek the practical objective of recovering ATC from Bluetooth data as a statistical calibration problem, following the development of a new time-varying coefficients Poisson regression model.
Bluetooth technology, Mathematical statistics, Traffic flow, Mathematical models