Scholarlink

<p>This study examines drivers’ gaze behavior in conditionally automated driving (CAD) to evaluate situation awareness (SA) using eye-tracking data and explainable artificial intelligence (XAI). SA was assessed in a laboratory experiment involving 39 participants using the Situation Awareness Global Assessment Technique (SAGAT) and the Situational Awareness Rating Technique (SART). Machine learning models and Shapley Additive Explanations (SHAP) analysis were applied to identify significant eye-tracking features for SA prediction. The findings suggest that metrics such as total number of fixations, total dwell time, and number of switches are strong predictors. Also, situational factors influence visual gaze behavior during SA estimation. The findings of this research enable the provision of a real-time, non-intrusive method to enhance driver safety in CAD by improving the understanding of visual attention patterns during control transitions.</p>

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

Modeling Drivers’ Situation Awareness During Takeover Using Machine Learning and Explainable AI

  • 期刊论文

  • 发表日期：

    2026年3月13日

查看全文
  • Young Woo Kim,
  • Geonwoo Kim,
  • Seonyoung Park,
  • Sol Hee Yoon

摘要

This study examines drivers’ gaze behavior in conditionally automated driving (CAD) to evaluate situation awareness (SA) using eye-tracking data and explainable artificial intelligence (XAI). SA was assessed in a laboratory experiment involving 39 participants using the Situation Awareness Global Assessment Technique (SAGAT) and the Situational Awareness Rating Technique (SART). Machine learning models and Shapley Additive Explanations (SHAP) analysis were applied to identify significant eye-tracking features for SA prediction. The findings suggest that metrics such as total number of fixations, total dwell time, and number of switches are strong predictors. Also, situational factors influence visual gaze behavior during SA estimation. The findings of this research enable the provision of a real-time, non-intrusive method to enhance driver safety in CAD by improving the understanding of visual attention patterns during control transitions.