Privacy-aware urban traffic modeling using mobile sensing data

Abstract

The results reveal that in addition to ensuring an acceptable level of privacy, the released datasets from the privacy-aware system can also be applied to urban traffic applications with satisfactory performance. Albeit application-specific, such a "Privacy-by-Design" approach can help broaden the traditional urban traffic modeling field and hopefully shed some light on other related science and engineering fields using mobile sensors.

Advances in global positioning systems (GPS) and wireless communications have prompted the rapid deployment of mobile traffic sensors (e.g., GPS-enabled devices, Bluetooth, smartphones, Connected Vehicles, etc.) that are able to move along with the flow they are monitoring. This research focuses on mobile traffic sensors such as GPS, which can provide detailed tracking capabilities, including fine-grained location traces, speed and other relevant information of individuals or vehicles. These capabilities, and the data they provide, may promise great advances in science and engineering. However, the use of mobile sensing data also poses great challenges. Two key challenges are addressed in this dissertation: the selection of which mobile data elements to be collected and used, and how to apply the collected mobile sensing data for urban traffic applications.

To this end, two important issues are considered. The first is how to satisfy the need for information extraction, i.e., data for transportation and especially urban traffic modeling purposes. Mobile data are fundamentally different from data collected via traditional means: they are more detailed spatially, but usually only provide a sample of the entire flow. As a result, choosing the form of mobile data to be collected and used will have profound implications on the development of new modeling techniques. The second concern is how to address the privacy issues evoked by collecting mobile data, such as location traces of individual drivers. Such concerns can slow down or impede the wide applications of new technologies. These two sometimes conflicting needs, data for modeling and privacy protection, pertain to many mobile-sensing-based traffic applications and need to be addressed in a holistic manner.

In this doctoral research, the researcher proposes the concept of co-designing traffic modeling methods and privacy protection mechanisms for urban traffic applications. The virtual trip lines (VTLs) zone-based privacy-aware system is developed to balance the level of privacy and the modeling needs in urban traffic environment. The proposed system is a combination of access control and privacy-preserving techniques. It trims mobile sensing data and only keeps the data that are essential to traffic applications. In such a privacy-aware system, novel traffic models are developed to use the collected/processed mobile sensing data for a variety of urban traffic applications, including automatic vehicle classification, vehicle trajectory reconstruction and vehicular energy/emission estimation. A data fusion and information integration approach is further developed to combine heterogeneous traffic data sources for fine-grained urban traffic applications.