Scholarlink

In conditional automation, supervised driving remains necessary, requiring drivers to take over when automation fails. This study examines takeover performance by considering both takeover readiness and scenario complexity. Takeover readiness was assessed through eye-tracking metrics and motor state, while scenario complexity was evaluated based on the properties of the encountered scenarios. A simulated driving cabin experiment was conducted with 64 participants, resulting in 565 valid takeovers, including 24 labeled as “poor takeovers” due to significant deviations from normal performance. Six critical features influencing takeover quality were identified. Among these, four were related to takeover readiness, and two were associated with the takeover scenarios encountered. Based on the feature importance ranking, physical readiness appears to play a more critical role than cognitive readiness alone in enabling a faster takeover with smoother vehicle dynamics. This research sheds light on real-time takeover quality identification, enabling the development of human-centered assistance for future intelligent vehicles.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Identifying Key Features for Classifying Takeover Performance in Level 3 Autonomous Vehicles with Consideration of Takeover Readiness

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Hsueh-Yi Lai,
  • Tse-Yi Kuo

摘要

In conditional automation, supervised driving remains necessary, requiring drivers to take over when automation fails. This study examines takeover performance by considering both takeover readiness and scenario complexity. Takeover readiness was assessed through eye-tracking metrics and motor state, while scenario complexity was evaluated based on the properties of the encountered scenarios. A simulated driving cabin experiment was conducted with 64 participants, resulting in 565 valid takeovers, including 24 labeled as “poor takeovers” due to significant deviations from normal performance. Six critical features influencing takeover quality were identified. Among these, four were related to takeover readiness, and two were associated with the takeover scenarios encountered. Based on the feature importance ranking, physical readiness appears to play a more critical role than cognitive readiness alone in enabling a faster takeover with smoother vehicle dynamics. This research sheds light on real-time takeover quality identification, enabling the development of human-centered assistance for future intelligent vehicles.