Estimation of stochastic link capacity and link performance function including uncertainty of driver’s behaviour

Abstract

Recently, user equilibrium models with uncertainty were proposed to describe stochastic travel time in road networks. The accurate estimation of a stochastic link capacity is important for such models. This study develops a method for estimating a stochastic link capacity by considering the uncertainty of the driver’s behaviour. We postulate that the stochastic link capacity follows a lognormal distribution. The characteristics of the lognormal distribution suit real traffic observation, e.g. nonnegativity, asymmetricity, and long-tailed. Furthermore, the estimated link capacity can derive the analytical likelihood function for the link performance function. Then, the link performance function can be estimated analytically by explicitly considering the source of uncertainty. The proposed methods can contribute to the estimation of stochastic link capacity and the link performance function for the user equilibrium model with stochastic travel time. A numerical calculation demonstrates the proposed method using real traffic observations in Sapporo, Japan.