Agent-Based Stochastic Modeling of Driver Decision at Onset of Yellow Light at Signalized Intersections

Abstract

This paper introduces two logistic statistical models for the driver stop–run decision at the onset of yellow at signalized intersections to capture the stochastic nature of the driver stop–run decision. One model is a classical frequentist model, whereas the other uses a Bayesian statistics approach. The Bayesian model parameters were calibrated by using the Markov Chain Monte Carlo slice procedure implemented within the MATLAB software. Both models were developed with 3,328 stop–run records, which were collected in a field experiment on the Virginia Smart Road, a limited-access highway between Blacksburg and Interstate 81 in Montgomery County, Virginia. The variables included in each model were driver gender, age, time to intersection, yellow time, approaching speed, and speed limit. Both models were shown to be consistent. For the Bayesian model application, two procedures were illustrated: cascaded regression and Cholesky decomposition. Both procedures produced replications consistent with the Bayesian model realizations, while these procedures captured the parameter correlations without the need to store the set of parameter realizations. The Bayesian model produced valid and transferable behavior by replicating multiple experimental results. The proposed Bayesian approach is ideal for modeling multiagent systems in which each agent has its own unique set of parameters.