Title: A cross-vendor and cross-state analysis of the GPS-probe data latency

Abstract: Crowdsourced GPS probe data has become a major source of real-time traffic information applications. In addition to traditional traveler advisory systems such as dynamic message signs (DMS) and 511 systems, probe data is being used for automatic incident detection, Integrated Corridor Management (ICM), end of queue warning systems, and mobility-related smartphone applications. Several private sector vendors offer minute by minute network-wide travel time and speed probe data. The quality of such data in terms of deviation of the reported travel time and speeds from ground-truth has been extensively studied in recent years, and as a result concerns over the accuracy of probe data has mostly faded away. However, the latency of probe data, defined as the lag between the time that disturbance in traffic speed is reported in the outsourced data feed, and the time that the traffic is perturbed, has become a subject of interest. The extent of latency of probe data for real-time applications is critical, so it is important to have a good understanding of the amount of latency and its influencing factors. This paper uses high-quality independent Bluetooth/Wi-Fi re-identification data collected on multiple freeway segments in three different states, to measure the latency of the vehicle probe data provided by three major vendors. The statistical distribution of the latency and its sensitivity to speed slowdown and recovery periods are discussed.