Optimization of Traffic Signal Offsets with High Resolution Event Data

Abstract

Recently, high-resolution data has emerged as an economical means of obtaining detailed information about signal operations. This has enabled the development of metrics to examine progression, such as cyclic flow profiles similar to those used in the TRANSYT model. Those profiles could further be used to predict changes to traffic flows under trial offset adjustments, enabling offset optimization. An algorithm to do so, called “Link Pivot,” has been used on numerous signalized arterials in Indiana and elsewhere over the past 8 years. This paper reviews the success of the predicted changes through statistical analysis. A linear regression of 569 predicted and actual percent on green values yields an R2 of 0.90. Next, predicted and actual arrival profiles were compared for 357 approaches for which detailed data was available. This analysis finds that the prediction method was successful in about 95% of the approaches considered in this evaluation. The paper dissects the remaining cases and discusses how they could be accommodated.