Quantifying Loop Detector Sensitivity and Correcting Detection Problems on Freeways

摘要

Loop detectors are the most commonly used vehicle detector for freeway management. A loop detector consists of a physical loop of wire embedded in the pavement connected to a sensor located in a nearby cabinet. The sensor detects the presence or absence of vehicles over the loop and typically allows a user to manually select the sensitivity level of operation to accommodate for a wide range of responsiveness from the physical loop. In conventional practice, however, it is difficult to know the physical loop's responsiveness, which makes selecting the appropriate sensitivity level difficult. If the sensitivity and responsiveness are poorly matched it will degrade the detector's data and the performance of applications that use the data, including: traffic management, control, and traveler information. To resolve this often overlooked problem, this paper presents an algorithm to assess how well a loop detector's sensitivity is set by calculating the daily median on-time from the data reported by the loop detector. The algorithm can be incorporated into conventional controller software or run off-line. The result can be used both to correct the detector on-times for an inappropriate sensitivity setting in software (e.g., through a multiplicative correction factor) and to trigger an alarm to dispatch a technician to adjust the hardware sensitivity. Plotting the daily median on-time over months or years can show how the detector performance evolves. The approach is then transposed to dual-loop detectors to identify and correct for inaccurate spacing between the paired detectors. Finally, the methodology is evaluated by comparing the loop detector speeds against the concurrent velocities from a GPS-equipped probe vehicle. Although the focus of this paper is on loop detectors, with only minor modification the algorithm should also be applicable to other detector technologies that emulate loop detector operation, e.g., side-fire microwave radar.