조건부 자율주행에서 제어권 전환 수행도와 제어권 전환 요청 시기에 따른 운전자의 시각적 탐색 전략에 대한 연구

Abstract

Objective: The purpose of this study is to analyze visual search strategies depended on takeover performance and takeover request lead time (TORlt) in Level 3 autonomous driving and derive eye movements for safe takeover. Background: Driver distraction while driving is a major accident factor. If their eye movements are dispersed or fixed at one point, this hinders information acquisition and proper decision-making, which leads to fatal accidents. Especially in Level 3 autonomous driving, as the driver is not obliged to monitor the driving scene, to ensure safe and successful takeover, drivers need to collect meaningful information during TORlt. Hence the study of the driver's eye movements and attention allocation in TORlt is essential. Method: Eye movement data were collected using a glasses-type eye tracker in an autonomous driving simulator experiment, and classified into high- and lowperformance groups. The static aspects (number of fixation, fixation duration) as well as dynamic aspects (visual entropy) of eye movement were statistically analyzed through two-way ANOVA. Results: Results showed significant differences in both aspects of takeover performance and TORlt. Significant differences were observed in the destination lane and the side mirror located in the destination lane by the takeover performance. Moreover, there were significant differences in the forward area, non-driving related task (NDRT) and a sum of fixation duration by TORlt. Regarding the dynamic aspects, the visual entropy values became significantly higher as TORlt increased and as the takeover was performed better. Conclusion: The study quantitatively revealed meaningful areas of interest and eye movements for successful takeovers. Application: The results can be applied in designing in-vehicle systems and in the further analysis of dynamic eye movements in takeover situations.