Modeling Automated Vehicle Interactions with Traffic Signs: A Data-Driven Approach

Abstract: The introduction of Automated Vehicles (AVs) marks a significant shift in the transportation landscape, influencing a wide range of stakeholders, including transportation agencies, policymakers, urban planners, the automotive industry, and society at large. To effectively manage this transition, it is crucial to develop advanced behavior models for AVs that can accurately capture the complex dynamics of vehicle interactions, diverse environmental conditions, and the uncertainties inherent in real-world operations. Although theoretical models of AV behavior exist, their evaluations have largely depended on assumptions that are disconnected from actual conditions, leaving the true impact of AVs on traffic operations unclear.
This paper addresses this gap by developing a comprehensive suite of interaction models between AVs and traffic signs—specifically, stop signs and yield signs—using the Waymo Open Datasets. Through rigorous calibration and validation, these models faithfully represent realistic AV behaviors, including approaching a traffic sign, merging at a yield sign to enter a priority road, making right turns at stop signs, executing one-step left turns at stop signs onto roads without medians, performing two-step left turns onto roads with medians, and navigating all-way stop intersections. A case study is conducted to evaluate how AV interactions with traffic signs affect overall traffic performance, focusing on both mobility and safety dimensions. Additionally, a series of future scenarios are simulated, varying key microscopic model parameters, infrastructure configurations, and AV technology penetration rates.
The outcomes of this paper aim to equip transportation stakeholders with essential insights to facilitate the responsible and efficient integration of AVs into the broader transportation system, supporting informed decision-making and promoting sustainable urban development.