Abstract
Traffic-operational projects typically are evaluated by using several measures of effectiveness, including vehicle delay, queue sizes, and vehicle stops. Other emerging measures include vehicle fuel consumption and emissions. Although these emerging measures are difficult to determine in the field, they can be estimated by using second-by-second speed measurements of individual vehicles with Global Positioning System (GPS) technologies. GPS can provide such information with relatively high accuracy (within 1 m/s), but the presence of any erroneous data in the measured data as a result of signal losses may result in unrealistic vehicle fuel consumption and emission estimates. The ability of various data-smoothing techniques to remove such erroneous data without significantly altering the underlying vehicle speed profile is investigated. The techniques reviewed involve differentiating the measured speed profile to generate an acceleration profile that is checked for validity. Several smoothing techniques are then applied to the acceleration profile, including data trimming, simple exponential smoothing, Epanechnikov kernel smoothing, robust kernel smoothing, and robust simple exponential smoothing. The results of the analysis indicate that the application of robust smoothing (kernel of exponential) to vehicle acceleration levels, combined with a technique to minimize the difference between the integral of the raw and smoothed acceleration profiles, removes invalid GPS data without significantly altering the underlying measured speed profile. A sensitivity analysis also demonstrates that the proposed datasmoothing technique can efficiently smooth random variations in speed profiles due to errors in GPS speed measurements.
Get full access to this article
View all access options for this article.