FACTORS AFFECTING THE RELATIVE PERFORMANCE OF TRAFFIC RESPONSIVE AND TIME-OF-DAY TRAFFIC SIGNAL CONTROL

This paper investigates whether vehicle detectors in a computerized traffic control system provide better information on network traffic demands for control purposes than could be obtained from historical average data. Tests were performed on data from Washington, D.C., Toronto, and Fort Wayne. These tests showed that the vehicle detectors and the associated predictor algorithm could give only slightly better results than historical average data. Sensitivity calculations showed also that the network MOE's were nearly the same when the traffic signal timing was optimized for the actual volumes in the network as when it was optimized for substantially different network volumes unless saturation resulted at one or more intersections. The combination of results indicates that it is unlikely that a traffic responsive system will perform better than a well designed time-of-day traffic signal system.