Enhancing Traffic Safety with the Implementation of Crowdsensing Solutions in the Mobile Era
Abstract
Traffic injuries are one of the most severe public health problems. Fueled by the growing availability of traffic-related data sources, data-driven safety studies have been extensively utilized to model traffic risks and enhance driving safety. Among these data sources, mobile crowd sourced (MCS) data shows significant potential to advance current safety studies substantially; however, the implementation of MCS-based solutions is still underexplored. This dissertation explores the potential of MCS-based solutions for enhancing traffic safety. It contains four distinctive research works to reevaluate and capture traffic risks using MCS data. In the first study, I utilized crowdsourced Waze data to re-assess freeway traffic risks. Traditionally, police crash reports (PCR) have been used as the primary source of crash data in safety studies, which cannot capture the unreported risks (near-crashes and traffic incidents). This study provides a new procedure to capture unreported traffic risks by combining PCRs and Waze data. The results demonstrated that Waze could capture a broad range of unreported traffic risks and be potentially used as a surrogate safety measure in the absence of crash data. The second and third studies introduce MCS-based solutions for monitoring road surface conditions. Road surface roughness assessment is essential in road maintenance, which is also closely related to traffic safety. However, continuously monitoring road surface roughness with a high-efficient solution remains a challenging research question.
In these two studies, we proposed new solutions to achieve large-scale monitoring of road surface conditions and the detection of road anomalies using MCS data. The results demonstrated, by mining the MCS data, road surface conditions can be effectively assessed. The last study introduces an innovative approach to characterize hazardous driving scenes, in which drivers are prone to making driving mistakes. This study marks the first attempt to explore the correlation between driving error occurrence and geospatial features. In this study, mobile sensed driving errors were integrated with driving-related geospatial features to form “scenic tuples” to characterize the occurrence of each error. Through mining a long-term collection of scenic tuples, we can extract the individualized hazardous scenes, which has the potential to aid in reducing driving risks.
Citation
Li, Xiao (2019). Enhancing Traffic Safety with the Implementation of Crowdsensing Solutions in the Mobile Era. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /189097.