A Model of Driver Behavior at High Speed Signalized Intersections

Abstract

Driver behavior at high speed signalized intersections is modeled in this paper as a binary decision (stop or go). A driver is assumed to stop if the time to arrive at the stop bar, at the onset of amber, is smaller than some critical time. Assuming that time perception and the critical time are normally distributed among drivers, the model's parameters are estimated using a probit calibration routine. The approach is shown to significantly reduce the sample size needed for estimating dilemma zone boundaries and therefore it represents a significant cost saving. As shown in the paper, the dilemma zone boundaries can be directly estimated (analytically) from the model without using stopping probability curves and the implied prior aggregation by speed. The efficient use of the data is a key to the reduced sample sizes needed for estimating the model.