Identifying Lane-Mapping Errors at Freeway Detector Stations

Abstract

With several loop detectors, numerous junctions, and site-specific configuration files, many opportunities exist for errors at a freeway detector station in mapping from a loop detector input to its physical location and lane in the freeway. If these configuration errors are not caught, they propagate to the station's data and any control decisions based on those data. Great care is necessary to prevent such errors, yet they persist in the field even with the added expense of trying to prevent them. This paper develops and tests two algorithms to identify such errors without using a priori knowledge about the detector station configuration. By correlating events between two loop detectors, the first algorithm tests all active detectors to match pairs from dual-loop detectors and to identify any single-loop detectors at the station. The second algorithm uses a hypothesis-test-based K-means method to group lanes from common directions based on time series velocity. Both algorithms exhibited good performance; they were tested over many days and detector stations. Although the analysis focuses on loop detectors, the methodology should also apply to emerging technologies that mimic loop detectors, such as wayside-mounted microwave radar.