Fall 2021 Series

Student Presenter: Dr. Zhiyong Cui Postdoc in the CEE Department at University of Washington. Dr. Cui is also a University of Washington Data Science Postdoctoral Fellow.

Student Advisor: Professor Yinhai Wang Department of Civil Engineering at University of Washington.

Original Air Date: September 13, 2021

Description

Traffic forecasting is a critical component of modern intelligent transportation systems for urban traffic management and control. Learning and forecasting network-scale traffic states based on spatial-temporal traffic data is particularly challenging due to the time-varying traffic patterns and the complicated spatial dependencies on road networks. The existence of missing values in traffic data makes this task even harder. With the rise of deep learning, this work attempts to answer: how to design proper deep learning models to deal with complicated network-wide traffic data and extract comprehensive features to enhance prediction performance, and how to evaluate and apply existing deep learning-based traffic prediction models to further facilitate future research? To address those key challenges in short-term road traffic forecasting problems, this work develops deep learning models and applications to: 1) extract comprehensive features from complex spatial-temporal data to enhance prediction performance, 2) address the missing value issue in traffic forecasting tasks, and 3) deal with multi-source data, evaluate existing deep learning-based traffic forecasting models, share model results as benchmarks, and apply those models into practice. To learn localized features from the topological structure of the road network, deep learning frameworks incorporating graph convolution operations are proposed to learn the interactions between roadway segments and predict their traffic states. Further, to fill in missing values in the graph-based traffic network, a graph Markov network is proposed, which can infer missing traffic states step by step along with the prediction process. In summary, the proposed graph-based models not only achieve superior forecasting performance but also increase the interpretability of the interaction between road segments during the forecasting process.

Student Presenter: Dr. Ziyuan Pu, Lecturer at Monash University.

Student Advisor: Professor Yinhai Wang Department of Civil Engineering at University of Washington.

Original Air Date: December 7th, 2021

Description

Real-time traffic data is essential for the advancement of emerging data-driven transportation technologies. The existing sensing technologies work well for identifying the mobility patterns of motorized vehicles. However, it is still a big challenge for transportation agencies to obtain reliable data of transit riders and non-motorized travelers in today’s practice with existing traffic sensing technologies. To fulfill the data needs of understanding and modeling the mobility of transit riders and non-motorized traffic device-based wireless sensing methods and technologies have been developed to acquire relevant data. The basic idea of device-based wireless sensing technology is to capture the Media Access Control (MAC) address of Wi-Fi or Bluetooth enabled mobile devices. The MAC address can be used as a global unique identifier to re-identify mobile devices at different sensing locations, and thus travelers can be detected by identifying their mobile devices instead of detecting travelers directly. This provides a novel means for transit riders and non-motorized traffic data collection. Nevertheless, limitations still exist for wireless sensing technologies due to the uncertainties caused by the sensing mechanism, including traffic mode uncertainty, localized spatial uncertainty, and population uncertainty. These bring considerable errors that can generate significant biases in the wireless sensing data. The major objective of this research is to mitigate the impacts of the uncertainties based on the proposed wireless sensing methods and technologies for transit rider and non-motorized traffic data acquisition. Large-scale field tests are conducted to evaluate the efficiency and accuracy of the proposed methodology. This dissertation fills up the gap about effective traffic data acquisition methods for transit rider and non-motorized traffic and thus supporting the transportation systems with reliability, equality, and sustainability.