Neural Network for Gap Acceptance at Stop‐Controlled Intersections

Abstract

The behavior of gap acceptance by vehicles at intersections with stop signs involves the complex interaction of numerous geometric, traffic, and environmental factors. Several methods, including empirical analysis, and theoretical, logit, and probit models have been used to estimate gap acceptance at stop‐controlled intersections. In the past, neural networks have been used to examine problems involving complex interrelationship among many variables and found to perform better than conventional methods. This paper describes the development of a neural network and a binary‐logit model for predicting accepted or rejected gaps at rural, low‐volume two‐way stop‐controlled intersections. The type of control, the turning movements in both the major and minor directions, size of gap, service time, stop type, vehicular speed, queue in the minor direction, and existence of vehicle in the opposite approach were found to influence the driver's decision to accept or reject a gap. The results of the neural network and the binary‐logit model were compared with the observations recorded in the field. The results revealed that the neural network correctly predicted a higher percentage of accepted or rejected gaps than the binary‐logit model.