Multiclass Traffic Assignment Model Considering Heterogeneous Stochastic Headways of Autonomous Vehicles and Human-Driven Vehicles

This study proposed a stochastic link capacity in mixed flows, considering heterogeneous stochastic headways for autonomous vehicles (AVs) and human-driven vehicles (HVs), respectively. Using this link traffic capacity, a multi-class traffic assignment problem considering mixed flows is proposed. Though previous study assumed an approximation of random variable computation to derive mixed link capacity, by assuming lognormally distributed stochastic headway, we showed the mixed link capacity model that ensure the link capacity analytically following lognormal distribution without any approximate computation of stochastic variables. Furthermore, the analytical relationship between the mean and standard deviation of the stochastic link capacity and the market penetration rate of AVs in a link is derived. We also showed the conditions for the mean and variance of the link travel time to be monotonically increasing with the link flow. The proposed traffic assignment model assumes that AVs and HVs follow UE and SUE, respectively. Numerical calculations were performed to validate the proposed model using a test network.