Real-time knowledge of system performance enables transportation agencies to manage traffic conditions and keep the public informed of periods of congestion. The ability to accurately measure arterial performance has been limited primarily by the availability of appropriate and reliable data. Unlike freeway networks where speed, flow, and occupancy measures are generally sufficient to quantify the performance of the link, the link-node combination of interrupted-flow arterial facilities requires additional information to accurately describe performance. There are a wide variety of potential data sources for arterial performance measures. These sources include cycle-level traffic signal data, bus automatic vehicle location (AVL) data, and Media Access Control (MAC) address readers. Each of these data sources has different properties, advantages and limitations. No individual element seems adequate as a sole source for arterial performance measure calculations. This research will investigate the fusion of the three above-listed data sources for arterial performance measurement.
In general, performance of an arterial can either be characterized from the perspective of the traveling public or the operating agency. The metrics for these two audiences are generally different and the most useful measure depends on the intended audience, the context, and the communications medium. Several authors have proposed that some measure of travel time is the most appropriate metric for the traveling public, while operational agencies are generally interested in intersection-level performance. This research will investigate both measures for the traveling public as well as operational measures.
Although it is possible to quantify arterial performance metrics in other tools (such as simulation or deterministic models) the purpose of our research is develop methods to report performance in real-time.
Sensors at traffic signals provide performance information. However, such detectors only provide information for travel near signalized intersections and using signal information to predict travel times between intersections is difficult. In contrast, data sources such as bus AVL data and MAC address readers provide good information about travel between intersections. Previous research has suggested that combining a point-based data source (i.e. traffic signal data) with a link-based data sources (i.e. bus AVL data) can improve the accuracy of arterial performance measures; however, this previous research was limited by a lack of detailed data. This project will have available to it a significantly-improved data set consisting of cycle-level signal data, bus AVL data and MAC address reader data from a single corridor. This research will investigate the improvement in performance measurement that can be obtained by fusing these three data sources. Improvements are expected due to the fact that the data sources provide complementary information (link vs. node) as well as to the unique opportunity to improve data cleaning and quality by combining sources.
The outcomes of this research have additional potential impacts. Existing detectors on urban streets are not often optimally placed or configured for accurate performance measurement. The deployment of additional sensors, such as MAC address readers, would potentially help alleviate that problem and provide additional detection where needed. However, such installations require resources to install and maintain and should only be done if the installations have been shown to improve performance measurement. This research will provide preliminary findings that will quantify this issue.
