Evaluating Real-Time Origin-Destination Flow Estimation Using Remote Sensing-Based Surveillance Data

Traffic management methods often employ network flow models that use data collected by a surveillance system to estimate the current state of the network and predict its evolution over time. Origin-Destination (OD) flows comprise one of the main elements of the network state. This paper focuses on the role of various types of surveillance data in the real-time estimation of dynamic OD flows. Traditionally, traffic surveillance is carried out using inductive loop detectors. While providing an inexpensive means of monitoring traffic, such detectors do not provide any data on traffic conditions beyond the detection point. Recent advances in communication, computing, and electronics are enabling the development of remote sensing-based surveillance systems which involve the observation of the roadway network from a fairly high vintage point. Such sensors can provide extended spatial coverage of the network, thus, capturing vehicle trajectories, queue lengths, and link travel times. Such data have the potential to improve the accuracy of the network state estimates along with traffic predictions and, consequently, result in more effective traffic management. This study quantifies the value of using intersection turning fraction and link travel time measurements, which are available from remote sensing-based surveillance systems, in estimating OD flows in real-time. A network consisting of three adjacent intersections is used to conduct an empirical analysis.