Algorithm for detecting and correcting errors in archived traffic data

An algorithm is presented for correcting errors in archived loop data from freeway traffic-management systems that are the result of poorly calibrated sensors. These errors pose a significant difficulty when archived data are used in off-line analysis because the calibration errors are difficult to detect by using traditional methods. In the presented work, consistency of vehicle counts is used to judge the validity of the data: if vehicles counts are balanced, the data are valid; if vehicle counts are not balanced, the data are not valid. The method also can determine a correction factor. This correction factor is used to create a time series that can be combined with the original data to adjust the volume to create a consistent data set. To illustrate the methodology, an example case is presented that details the process of identifying a pair of reference stations that are properly calibrated. After the reference stations are identified, a poorly calibrated station is identified, and the data from this station are corrected. The result of the correction process is discussed.