Authors: Haoyi Xiong*, University of Iowa, Xun Zhou, University of Iowa, Amin Vahedian, University of Iowa, Dave Bennett, University of Iowa
Topics: Transportation Geography, Quantitative Methods, Temporal GIS
Keywords: Spatiotemporal Data Mining, Transportation, Traffic Congestion, Network, Event Detection
Session Type: Paper
Start / End Time: 3:05 PM / 4:45 PM
Room: Roosevelt 0, Marriott, Exhibition Level
Presentation File: No File Uploaded
Given the availability of large trajectory data from vehicles, detection of real world traffic congestion become feasible to achieve. Extensive research have been done in this field in past decades, but this problem still remains challenging. This study focusing on detecting sets of connected congested roads caused by unpredictable events (e,g. traffic accidents, road closure) and the corresponding footprints of congestion propagations (i.e. how those congestions spreading out among road network). The information could be used for urban road network structure planning and routing services. Two challenges remain in this problem: first, complex urban traffic congestion patterns due to various traffic needs in different space and time; second, hidden footprints of congestion propagation in large amount of vehicle trajectories. This study will address these two challenges by proposing a data-driven approach. A case study will be conducted in Chengdu, China, and our model performance will be evaluated against past STSS baseline approaches.