SAFETYLIT WEEKLY UPDATE

We compile citations and summaries of about 400 new articles every week.
RSS Feed

HELP: Tutorials | FAQ
CONTACT US: Contact info

Search Results

Journal Article

Citation

Lee H, Coifman BA. Transp. Res. C Emerg. Technol. 2012; 24: 141-156.

Copyright

(Copyright © 2012, Elsevier Publishing)

DOI

10.1016/j.trc.2012.02.005

PMID

unavailable

Abstract

Loop detectors are the most commonly used vehicle detectors for freeway management. There has been considerable research to screen the quality of loop detector data using macroscopic measurements (e.g., five min aggregate data) and more recently using microscopic data (i.e., individual vehicle actuations). However, some significant detector errors have not received much attention due to the difficulty of identifying their occurrence. This paper examines one such error, splashover: the erroneous detection in one lane of a vehicle from an adjacent lane. If not caught, the splashover events will cause the flow and occupancy measurements to be too high. We examine the nature of splashover and develop an algorithm to find detectors exhibiting chronic splashover problems.

At the crux of this algorithm, an erroneous pulse arising from splashover in one lane should usually be bounded by the valid pulse from the vehicle in its lane of travel. The algorithm calculates the rate of suspected splashover: the percentage of pulses in the subject lane that are bounded by a pulse in an adjacent lane. However, a difficulty arises because any given splashover event in the data stream is usually indistinguishable from the non-splashover event of two vehicles passing the detector station at the same time yielding valid concurrent actuations. To control for non-splashover events we calculate a dynamic threshold rate of false positives as a function of the observed traffic conditions. Recognizing the fact that chronic splashover is non-transient, for the reasons explained in the paper, the algorithm is intended to be applied during free flow periods since these conditions are opportune.

Concurrent loop detector data and video ground truth are used for the evaluation from a total of 21 directional data sets, covering 94 adjacent lane pairs from 68 loop detectors at 15 loop detector stations, both with and without actual splashover. The algorithm exhibited good performance. Finally, with only minor modifications, the algorithm should also be applicable to other detector technologies, e.g., side-fire microwave radar sensors.

NEW SEARCH


All SafetyLit records are available for automatic download to Zotero & Mendeley
Print