Data-Driven Approach to Reverse Engineer Detector Metadata at Signalized Intersections with Unknown Control Plans

The association of phases to detectors is an essential piece of information needed to manage an inventory of detectors and to measure intersection performance using automated traffic signal performance measures (ATSPM). Often, agencies operating traffic signals do not possess readily available, up-to-date information on detector assignments. This paper provides a data-driven method for automatic identification of detector types and phase assignments at signalized intersections using high-resolution event data. Event data includes the timestamps of detector activations and de-activations as well as phase state changes at 0.1-s resolution. The data preparation starts by sorting data into data structures for analysis. The detector data is associated with individual signal cycles. These data are then aggregated into five-second time intervals. We propose a machine learning method, Occupancy Pattern Association (OPA), to identify the type of detector (stop bar presence or count) and the assignment of phases actuated by the detector. The effectiveness of the proposed algorithms is illustrated by processing a wide range of intersections across three different corridors in different states of Utah, Nebraska and Iowa. The accuracy of detector type identification is 88% or higher among all intersections, and detector configurations at seven intersections are identified with 100% accuracy. The proposed method holds promise for addressing the challenge of developing and maintaining an inventory of detector configurations.