Large-Scale Freeway Network Traffic Monitoring: A Map-Matching Algorithm Based on Low-Logging Frequency GPS Probe Data

Abstract

Low-logging frequency GPS probe data have become a major data source for large-scale freeway network traffic monitoring. A critical step in GPS data processing is map matching. However, traditional map-matching algorithms are developed for in-vehicle navigation with high-logging frequency GPS data, noting that high-logging frequencies can be 1 s, whereas low-logging frequencies can be a few minutes. Such algorithms map a new GPS positioning point instantaneously given its historical points and network topology. Using high-logging frequency data-based map-matching algorithms for low-logging frequency data can cause several problems. First, large mapping errors in previous GPS points can easily propagate to the current points. Second, one-point-a-time processing is not effective and not necessary for traffic monitoring. Multiple GPS points can be processed together to determine routes more effectively. In this article, the authors propose a map-matching framework for low-logging frequency GPS probe data. The proposed framework (a) incorporates curve matching and probabilistic analysis modules of high-logging frequency map-matching algorithms and (b) introduces a new route determination algorithm for multipoint processing on the basis of fuzzy logic and a concurrent version of the N-shortest path algorithm. The authors evaluated the proposed model using field GPS data sets collected in Los Angeles, California. Evaluation methods include not only traditional random mapping case inspection but also a comparison between the GPS-detected speed and the ground truth loop-detector speed to evaluate its effectiveness for traffic monitoring. The evaluation results illustrate the effectiveness and robustness of the proposed framework.