TRAFFIC DETECTOR ERRORS AND DIAGNOSTICS

The results from research into the use of vehicle detectors, with an emphasis on the diagnosis and correction of detector errors, are described. Of primary interest is the development of a diagnostics scheme in which the average vehicle on time is examined as a test statistic. By comparing this value against the average on times for a station of detectors, the validity of detector operation can be checked. This scheme has been tested at the San Francisco-Oakland Bay Bridge and has been found to yield good results. The false-alarm rate is low compared with that for the occupancy diagnostic method, and sensitivity to true detector failures is improved. This test has also been carried out on inductive loop data from Los Angeles and Chicago. Other experimental work has shown that for magnetometers, the measurement of occupancy is greatly influenced by the manner in which the detector is tuned. Thus methods for improving the consistency of detector tuning and minimizing errors are suggested. It has also been found that pulse breakups are a common operational problem, especially in congested conditions and with heavy vehicles. This can lead to errors in measured occupancy and counts of several percent. Tests have shown that breakups are inherent in the design of the hardware, but that compensation can occur with software. An algorithm for this has been designed and implemented that reduces these errors and improves estimation of vehicle lengths. Missed vehicles, spurious pulses, and lane changes have been found to constitute a small fraction of abnormal detector signals.