Naturalistic Open-Source Pedestrian-Driver Yielding Dataset Collected in Minnesota

本研究介绍了一个名为“Naturalistic Open-Source Pedestrian-Driver Yielding Dataset Collected in Minnesota”的大型开放源数据集，由明尼苏达大学土木、环境和地质工程系创建。该数据集包含超过3000次驾驶员与行人互动的视频数据，涵盖了超过50个不同的情境变量。数据集的创建过程涉及在明尼苏达州的18个无信号交叉口部署定制传感器包，记录自然环境下的驾驶员和行人行为。该数据集主要用于研究驾驶员在无信号交叉口的让行行为，旨在通过分析这些互动来提高交通安全，特别是行人的安全。

This study presents a large-scale open-source dataset named *Naturalistic Open-Source Pedestrian-Driver Yielding Dataset Collected in Minnesota*, developed by the Department of Civil, Environmental, and Geological Engineering at the University of Minnesota. This dataset contains over 3,000 video recordings of pedestrian-driver interactions, covering more than 50 distinct contextual variables. The development of this dataset involved deploying customized sensor packages at 18 unsignalized intersections across Minnesota to capture the naturalistic behaviors of both drivers and pedestrians. This dataset is primarily used to study drivers' yielding behaviors at unsignalized intersections, with the goal of improving traffic safety—especially pedestrian safety—by analyzing these real-world interaction scenarios.