Data underlying the research on driver takeover responses in conditionally automated driving

This dataset was collected from a fixed-base driving simulator experiment designed to examine driver responses to takeover requests in Level 3 conditionally automated driving. Each of the 57 participants (33 male, 24 female; mean age = 38.51 ± 17.23 years) completed nine takeover scenarios. These scenarios were generated by combining three levels of traffic density (0, 10, 20 vehicle/km) with three levels of cognitive workload induced by non-driving-related tasks (0-back, 1-back, 2-back). The order of nine scenarios was balanced using a Latin Square design, thereby reducing the potential order and learning effects and enhancing the dataset’s reliability for experimental replication and the development of generalizable models.<br>The dataset includes multimodal data capturing (1) driver characteristics (takeover style, risk-taking attitude, etc.), (2) scenario information (traffic density and non-driving-related task), (3) vehicle operational data (velocity, acceleration, steering wheel angle, etc.), (4) subjective scenario experience (situational awareness, spare capacity, etc.), and (5) physiological signals (eye movements and heart rate). Detailed descriptions of the variables and formats are provided in the data_dictionary.csv.

本数据集采集自一项固定基座驾驶模拟器实验，旨在探究有条件自动驾驶（Level 3）场景下驾驶员对接管请求的响应表现。共计57名参与者参与实验（男性33名，女性24名；平均年龄为38.51±17.23岁），所有参与者均完成了9组接管场景任务。上述场景由三类交通密度等级（0、10、20辆/公里）与三类由非驾驶相关任务诱发的认知负荷等级（0-back、1-back、2-back）组合生成。实验采用拉丁方设计（Latin Square design）对9组场景的呈现顺序进行平衡，以此降低潜在的顺序效应与学习效应，提升了数据集在实验复现与可泛化模型开发中的可靠性。
本数据集包含多模态采集数据，具体涵盖以下五类：（1）驾驶员特征（接管风格、冒险态度等）；（2）场景信息（交通密度与非驾驶相关任务类型）；（3）车辆运行数据（车速、加速度、方向盘转角等）；（4）主观场景体验（情境感知、剩余认知容量等）；（5）生理信号（眼动与心率）。变量与数据格式的详细说明已在data_dictionary.csv文件中给出。