Stephen Boyles

Reliability assessment for traffic data

Given the vast amounts of data automatically collected by traffic detectors, identifying erroneous data is an important and challenging issue. In this paper, we develop a fuzzy logic approach for quantifying the reliability of data obtained from traffic detectors. Previous researchers have proposed multiple criteria for determining erroneous data; broadly speaking, these approaches either consider fundamental consistency (is the data physically plausible?), network consistency (is the data consistent with observations at nearby detectors?), and historical consistency (is the data plausible given past observations at this location?). This paper proposes a classifier incorporating all of these criteria, applying fuzzy logic to integrate these three separate assessments. An example application is given, utilizing data collected in the Dallas, TX, region.