Real-Time Data-Based Decision Support System for Arterial Traffic Management
Research Team
Mohammed Hadi, Ph.D., Florida International University
Lily Elefteriadou, Ph.D., University of Florida
Project Description
Traffic congestion along arterial streets is increasingly becoming a critical issue that needs to be addressed by transportation agencies. Compared to the relatively mature management of freeways, arterial traffic operations and management are lagging behind. To address such a gap, Intelligent Transportation System (ITS) devices, such as traffic detectors, Bluetooth/Wi-Fi readers, and so on, are installed or planned to be installed along a number of arterial streets. The data generated from these devices provide an enriched source for monitoring arterial traffic and estimating performance measures. However, these measures are usually estimated offline to check arterial street performance and to verify the effectiveness of traffic operations. The real-time traffic operations of urban streets still rely more on visual examination of the videos generated from closed circuit television (CCTV) cameras at traffic management centers. To mitigate congestions during incidents or special events, some transportation agencies manually adjust signal timing based on operator’s observations of queues. A robust and automated decision support system is needed to help transportation agencies to better manage arterial traffic in real time.
The goal of this project is to develop a real-time data-based decision support system for arterial management. The developed system will not only automatically estimate system performance and identify the traffic state based on data from multiple sources, but also predict the short-term traffic conditions using advanced machine learning techniques. Recommendations will be further provided by the system based on predicted traffic condition as well as agencies’ past operational experience to assist agencies in determining optimal arterial traffic management and control strategies.
Product
Clustering – Developed a hybrid unsupervised (clustering) for off-line identification of the traffic state and supervised machine learning approach (decision tree) for real-time classification of the traffic state. The approach was tested utilizing data from Sunrise Blvd. in Broward County.