Nonparametric Evaluation of Adaptive Traffic Signal Control System Based on Reliability Measures

Abstract

Adaptive traffic control systems often require additional sensors and more advanced computing technologies than time-of-day control, resulting in a more costly system installation. A more comprehensive evaluation than traditionally undertaken is needed to justify these increased costs. Conventional measures to evaluate traffic signal control have been average measures, such as average speed, travel time, delay, etc. Depending on the selected measures, the evaluation results may favor different systems. Thus, to address the need for a more comprehensive evaluation, this study explores measures that aim to quantify travel time reliability. From the transportation system perspective, reliability is commonly defined according to the level of travel time variation or the probability that travelers will arrive at their destination within a given time. To quantify this, reliability measures are often closely associated with the width of travel time distribution, or 95th percentile travel time, implying that estimating a reliability measure with a reasonable accuracy requires more data than estimating an average measure. The evaluation of traffic signal control on arterials has typically relied on test vehicle data in which the sample size is limited in most applications. Given the above background, the main objective is to provide a method to evaluate the effect of an adaptive traffic signal control system using reliability measures based on test vehicle data. The method is presented using the data collected in the SCATS pilot study in Cobb County GA, U.S. To achieve the objective, existing travel time reliability measures were reviewed, and appropriate measures are selected for the evaluation. To overcome the sample size limitation, a bootstrap sampling technique to estimate the confidence interval of a reliability measure was adopted to compare the control performance before and after the new system was adopted. The comparison results showed that in selected time period-route combinations, SCATS performed better than the before system, but the overall performance is comparable each other. Also it was found that traffic control system performance evaluation may lead to different results depending on the measure selection.