Stochastic Multiclass Traffic Assignment with Consideration of Risk-Taking Behaviors

Abstract

A study of the problem of predicting traffic flows under traffic equilibrium in a stochastic transportation network is presented. Travelers’ risk-taking behaviors are explicitly modeled with respect to probabilistic travel times. Traveling risks are quantified from the travel time distributions directly and are embedded in the route choice conditions. The classification of risk-neutral, risk-averse, and risk-prone travelers is based on their preferred traveling risks. The formulation of the model clarifies that travelers with different risk preferences have the same objective–to save travel time cost–though they may make different route choices. The proposed solution algorithm is applicable for networks with normal distribution link travel times theoretically. Further simulation analysis shows that it can also be applied to approximate the equilibrium network flows for other frequently used travel time distribution families: gamma, Weibull, and log-normal. The proposed model was applied to a test network and a medium-sized transportation network. The results demonstrate that the model captures travelers’ risk-taking behaviors more realistically and flexibly compared with existing stochastic traffic equilibrium models.