New Algorithms for Filtering and Imputation of Real-Time and Archived Dual-Loop Detector Data in I-4 Data Warehouse

Abstract

Loop detectors have been used to gather traffic data for over four decades. Loop data diagnostics have been extensively researched for single loops. Loop data diagnostics for the dual loops laid along 63 km (39 mi) of I-4 in Orlando, Florida, are specifically addressed here. In the I-4 data warehouse, dual-loop detectors provide flow, speed, and occupancy every 30 s. The mathematical relationships among flow, speed, occupancy, and average length of vehicles were used to flag bad data samples provided by a loop detector. A value called the entropy statistic is defined and used to determine the detectors that are stuck. Regression techniques were applied to fill the holes formed by the bad or missing samples. Various pairwise regression models were developed and described, and their performance on the loop data from January and February 2003 was analyzed. The best model was identified as the pairwise quadratic regression model with selective median, which is currently being used to impute missing data in real time. Results are presented of the application of these algorithms to archived loop detector data in the I-4 data warehouse.