Delay estimation models for signalized intersections using differential evolution algorithm

Abstract

Delay is an important parameter in the optimization of traffic signals and the determination of the level of service (LOS) of a signalized intersection since it directly reflects the lost travel time and fuel consumption. The accurate estimation of delay is, therefore, an important issue. The purpose of this study is to develop new delay models with less input parameters by using one of the artificial intelligent techniques. In this research, three types of differential evolution delay estimation models (DEDEM), i.e. linear, exponential and quadratic, were developed using differential evolution (DE) approach. In developing of the delay models, the green ratio (g/C effective green to cycle length) and the degree of saturation (x=v/c; volume to capacity) were considered. The first one changed from 0.35 to 0.60, the second one varied between 0.7 and 1.4. The model outputs were compared analytically to the HCM and Australian (Akcelik) delay models. The study results illustrated that R2, Mean Square Error (MSE) and Mean Absolute Error (MAE) values of DEDEMquadratic, which are 0.97, 207.98, 12.12 respectively, were better than analytical delay models and other types models. As a result, the quadratic form of DEDEM model can be used as an alternative estimation model for delay, and DE approach can be utilized as a model-fitting algorithm as well.