Predicting driver's takeover time based on individual characteristics, external environment, and situation awareness

Area of Science:

• Human-computer interaction
• Automated driving systems
• Driver behavior analysis

Background:

• Safe transitions in conditional automated driving rely on timely driver takeover.
• Predicting driver takeover time is essential for developing robust safety systems.
• Existing models often lack comprehensive integration of individual, environmental, and cognitive factors.

Purpose of the Study:

• To construct a predictive model for driver takeover time.
• To identify key individual, environmental, and situation awareness variables influencing takeover time.
• To compare the performance of different predictive modeling approaches.

Main Methods:

• High-fidelity driving simulation experiments with 18 designed takeover events.
• Extraction of 15 basic and 3 dynamic factors from individual characteristics, external environment, and situation awareness.
• Application of XGBoost and Shapely for model construction and variable contribution analysis.

Main Results:

• The BM+SA model achieved a high goodness of fit (Adjusted_R² = 0.7746).
• XGBoost outperformed other machine learning models (SVM, Random Forest, CatBoost, LightBoost).
• Situation awareness, individual characteristics, and environmental factors showed decreasing importance; scan/gaze duration increased takeover time, while pupil area and self-reported SA decreased it.

Conclusions:

• The developed BM+SA model effectively predicts driver takeover time.
• Situation awareness is a critical factor influencing driver takeover performance.
• Findings support the development of real-time driver monitoring systems and optimized human-machine interaction in automated vehicles.