The Effects of Data Inaccuracy on the Performance of Traffic Signal Timing Plans

Problem

A variety of traffic signal timing systems are currently in use in the transportation network. The accuracy of the timing systems in use depends heavily on the accuracy in predicting vehicle arrivals. With less accurate systems, traffic may pile up behind lights that don't change often enough, or just a few vehicles may have to wait a long time for a light to change even when there is no cross traffic.

This study made no valuation of the various signal timing systems. Rather, it sought to quantify when and how the performance of a signal timing plan will degrade when vehicle arrival information that is inaccurate to varying degrees is used to calibrate the plan. The study also sought to explore the level of improvement that can be achieved if very accurate vehicle arrival information can be obtained for the calibration of signal timing plans.

Method

An isolated intersection (i.e., not an arterial or network) was used as a model. The intersection was assumed to have no left or right turns. The performance of signal timing plans was then evaluated for a calibration that uses imperfect vehicle arrival information at this intersection.

In reality, vehicle arrival rates vary with time of day, whereas arrival distributions vary spatially from intersection to intersection. To account for these variations, this study included sensitivity analysis based on 18 vehicle arrival rate and distribution scenarios. The paper further discusses the methodology used to complete the study, including selection of a delay formula, use of variance-to-mean ratios, and detailed description of the 18 different vehicle arrival rate scenarios used as signal timing system calibrations.

Findings

Overall, the change in signal timing delay is insignificant when less accurate data are used to calibrate a signal timing plan. The effect of over- or under-estimation of traffic flow on total vehicle delays at the timed traffic signal increases with flow level. When traffic flow is low, less than a 1% increase in delay is observed. When flow is moderately heavy, less than a 4% increase in delay is observed. Significant increase in delay (greater than 10%) is observed only when the flow level is underestimated in a situation of high flow.

Underestimation of flow can also result in an increase in queue length and the number of stops. The increase in total number of stops is minor for critical intersections (less than 5%) but large for non-critical intersections (greater than 10% when the measured flow deviates from the actual flow by 10%).

In general, the performance of a signal timing plan is more sensitive to the accuracy of in input data when applied to an intersection with balanced flow than to an intersection with unbalanced flow.