Accurate, Affordable and Widely Applicable Freeway Travel Time Prediction: Fusing Vehicle Counts with Data Provided by New Monitoring Technologies

Abstract

AbstractToday, technology allows highly accurate direct travel time measurements. These can be attained by identifying vehicles at several locations on the freeway or by directly tracking vehicles’ trajectories. The penetration rate of these technologies is higher than ever before and continuously growing so that the traditional problem of data significance (i.e. not having enough measurements during a short updating period) is being attenuated. This fact has encouraged traffic administrations and private companies to deploy real-time information systems based on these data. However, even in an ideal scenario, direct measurements of travel times are representative of near past traffic conditions for vehicles entering the target stretch, while the objective of real-time information systems is to transmit information about traffic conditions in the near future. This chapter aims to fuse the information provided by input–output diagrams obtained from loop detectors with direct measurements of travel times obtained from automatic vehicle identification (AVI) or tracking technologies. This fusion allows exploiting the accuracy of the direct measurements to correct the count drift in loop detectors. Then, corrected input–output curves can be used to obtain reliable short-term predictions of travel time from vehicles’ accumulation. The proposed data fusion method has been applied to a test site in the AP7 freeway near Barcelona using real and simulated data. Results show that the method is able to provide predicted travel times that anticipate changes in traffic conditions much faster than the simple dissemination of measured travel times, implying lower average and maximum errors of the real-time information systems. The benefits of using the method grow with the severity of congestion and in low surveillance environments, which represent the scenarios where the travel time information is more precious and more difficult to obtain.