Kaito MATSUO

A traffic assignment model considering area-specific route choice criteria

Teppei KATO

 In recent years, autonomous driving technology has rapidly developed, and improvements in traffic efficiency due to the introduction of autonomous vehicles are widely expected. In Japan, the government has set a goal of achieving Level 4 autonomous driving on expressways by 2025. Autonomous driving is therefore expected to be introduced first on high-standard roads, such as expressways, where pedestrians and bicycles are not present. In contrast, on low-standard roads such as ordinary urban streets, additional infrastructure development may be required for the implementation of autonomous driving. Autonomous driving is expected to improve not only link-level traffic efficiency through microscopic vehicle control, such as shorter headways and increased road capacity, but also network-level efficiency through macroscopic traffic management, including route guidance, as its adoption expands.
 Considering this transitional period of autonomous vehicle adoption, it is likely that macroscopic route guidance will be feasible on high-standard roads, while manual driving and selfish route choice by drivers will continue on low-standard roads for the time being. Based on this background, this study aims to predict traffic flow and examine appropriate road policies during the transitional phase in which autonomous driving areas and manual driving areas coexist. To achieve this, a traffic assignment model is developed.
 In the proposed model, stochastic user equilibrium (SUE), which considers perception errors in travel time, is assumed in manual driving areas. Meanwhile, system optimal (SO) traffic assignment, which minimizes the total travel time of the entire network including autonomous driving areas, is assumed for the overall network. Therefore, the proposed model is formulated as a system optimal assignment model with equilibrium constraints in manual driving areas, resulting in a mathematical program with equilibrium constraints (MPEC).
 The model is applied to a test network and numerical experiments are conducted. First, it is numerically confirmed that the obtained solutions satisfy equilibrium conditions even when they are locally optimal. Case studies that vary the size of manual driving areas show that total travel time decreases as selfish route choice diminishes. However, significant improvements in link congestion rates are not observed. Additional experiments assuming capacity expansion due to autonomous vehicle adoption indicate improvements in both total travel time and congestion levels, although capacity increases on some links may worsen congestion on others. The synergistic effects of capacity increases in both autonomous and manual driving areas are also examined.