Citation

Cook AR, Cleveland DE. Transp. Res. Rec. 1974; 495: 1-11.

Copyright

(Copyright © 1974, Transportation Research Board, National Research Council, National Academy of Sciences USA, Publisher SAGE Publishing)

DOI

unavailable

PMID

unavailable

Abstract

This paper investigates the feasibility of using freeway traffic-flow data compiled by electronic surveillance and control systems for the detection of accidents and other lane blockage incidents that temporarily disrupt traffic flow. Research was conducted on the John C. Lodge Freeway in Detroit where traffic data consisted of 1-minute compilations of volume and occupancy recorded by ultrasonic presence detectors. Nineteen detection algorithms, including one being used in Los Angeles, were evaluated with a sample of 50 representative afternoon peak-period incidents. The technique of exponential smoothing of occupancy or volume data to detect incident-generated shock waves was found to be the most effective. This algorithm detected 42 percent of the incidents with virtually no false alarms and every incident with an 8 percent false-alarm rate. Most of the incidents were detected within 1 minute of the onset of congestion at a detector station. Algorithm effectiveness was not affected by detector spacings ranging from 1,460 to 4,815 ft (445 to 1468m), volumes from 1,200 to 2,000 vph per lane, occupancies from 9 percent to 45 percent, precipitation, or the particular lane blocked. The algorithms could not distinguish accidents from less serious incidents, but because they directly related each incident to its impact on traffic operations their incorporation in control systems could improve system response to incidents.