Citation

Chen H, Zhao X, Li Z, Li H, Gong J, Wang Q. Transp. Res. F Traffic Psychol. Behav. 2023; 95: 281-296.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.trf.2023.04.012

PMID

unavailable

Abstract

The driver's takeover behavior is critical to safety during automated driving in fog zones. This paper aims to explore the influence mechanism of takeover behavior, this paper analyzes the influence of various factors on takeover behavior from the perspective of event duration, and the influence mechanism of takeover behavior can be promoted. A takeover behavior test platform of L3 automated driving technology was developed based on a driving simulator, and a high-fidelity driving simulation experiment was carried out with the fog zone selected as the takeover scenario. The takeover request time (TOR) was set as 5 s and 10 s, and the no-driving-related task (NDRT) was set as work tasks and entertainment tasks. 42 drivers were invited to participate in the experiment where takeover behavior data were obtained. Takeover response time and takeover correct time are representative indicators of takeover behavior, which are extracted from takeover response and takeover performance. The survival analysis (the Kaplan-Meier method and Cox proportional-hazards method) was used to explore the effects of factors (gender, age, driving age, TOR, NDRT) on survival time (takeover response time, takeover correct time). The KM model results show that driver attribute factors (gender, age, and driving age) had significant difference on takeover response time. Females were more cautious about the takeover, and their response time is shorter than that of males. The takeover response time increased with age and driving age, with statistical difference. Only TOR had a statistical difference in the takeover correct time, while other factors had no statistical significance. The takeover correct time of TOR-10 s (median = 8.50 s) is higher than the TOR-5 s (median = 3.65 s). However, the mean value of takeover correct time corresponding to 5 s is higher than 10 s. The Cox model results are consistent with the KM model, and the model results show that external variables (TOR, NDRT) have significant influence on the takeover correct time. The work task requires more immersion, and the longer time it takes for drivers to recover the situational awareness of driving, which leads to an increase in the takeover correct time. Based on the survival analysis method, this research studies the influence of factors on takeover behavior from the perspective of the duration of takeover events, which can lay a foundation for in-depth study on the influence mechanism about takeover behavior. The study can also provide references for the improvement of drivers' takeover behavior, and the optimization design of the takeover warning system.

Language: en

Keywords

Autonomous Vehicles; Driving Simulator; Influence Factors; Survival Analysis; Takeover Behavior