Fan Ding
Published: 2017
Total Pages: 252
Get eBook
Traffic details monitoring for a large-scale freeway is always a long-term significant and practical topic in both academic and industrial transportation community. Conventionally, traffic monitoring systems are using roadside equipment data. However, costs of such equipment including both maintenance and installation are expensive. To apply cellular data as a new and indirect data source on traffic states estimation has emerged for nearly two decades. Full cellular activity data refers to the complete records of real-time cellphone communication signals generated by cell towers while maintaining mobile services both on- and off-call. Full cellular activity data is a big data resource, and such data is related to phone calls, texting, web browsing, video and audio streaming, location-based service and other activities. Existing cellular probe-based traffic states estimation methods heavily rely on on-call wireless location technologies signal transition data such as location update (LU), handoff (HO), and timing advance (TA) data. However, in fact, signal transition data is a minuscule subset of the full cellular data and only generated when a phone crosses location area boundaries during an active phone call. In addition, those methods also rely on routine road tests to obtain the relation between cell towers (CT) and freeway segments. Existing safety studies suggest that making phone calls while driving is a safety hazard, and phone calls during driving become forbidden within the improvement of law-making. This research presents a design of traffic monitoring system using the full cellular data for traffic status detection and estimation. Detailed descriptions of each module in such system are given. Features, including the link average CT heat, the link pseudo-speed and the link phone count, are defined and introduced in this research. Two algorithms, a rule based self-adaptive algorithm, and a machine learning based model, are developed to determine the freeway congestion level based on these features. The proposed system is going to be implemented for a major freeway corridor in China. Results are validated by fixed-point radar detector data. As a data-driven technique, the proposed method shows its advantages when there are only limited funds to implement a traffic monitoring system for the large-scale freeway.