Traffic delay estimation using artificial neural network (ANN) at unsignalized intersections

Abstract

This study was carried out to the modeling of control delays at unsignalized intersection using Artificial Neural Network (ANN). Although, there are several methods for estimation of delay, they lead to different results. A comparative analysis for estimation of the control delay using Malaysian Highway Capacity Manual (MHCM) showed that the theoretical model was not consistent with actual delays observed from sites. Such a finding implies that MHCM's model was not directly capable to the analysis of control delay at unsignalized intersections in Malaysia. Data pertaining to the analysis of control delay was collected from three intersections of various configurations using video camera recording technique. An ANN with two hidden layers and several sizes of neurons in the hidden layers were developed. Two mathematical models for estimation of control delay from minor road with a reasonable accuracy were developed using the outputs from the ANN's model. A statistical analysis revealed that there is good agreement between formulas acquired from the ANN's model and those from the field studies. The results of this research showed that the neural network is able to predict control delay incurred on minor road vehicles at unsignalised intersection more accurately. The analysis revealed that heavy vehicles had the lowest effect on the proposed formulas, where by increasing from 10% to 50%, the values of control delay could increase from 1% to 3%, while the movement flow and conflicting flow had the highest impact, where within the same ranges; control delay could increase until 39%.