Towards Context-Aware Modeling of Situation Awareness in Conditionally Automated Driving
Lilit Avetisyan, X. Jessie Yang, Feng Zhou
TL;DR
This study advances context-aware estimation of driver situation awareness (SA) in SAE Level 3 driving by integrating multimodal physiological signals, eye-tracking metrics, and demographic data collected in a driving-simulator takeover task. A LightGBM regression model, interpreted with SHAP, predicts real-time SA and identifies a top-12 feature subset that yields $RMSE=0.89$, $MAE=0.71$, and $Corr=0.78$, revealing how age, AV knowledge, arousal, and visual attention relate to SA. The work shows that risk perception and automation errors modulate SA and driver attention, supporting the design of adaptive HMIs and personalized interventions to improve safety during takeovers. Despite limitations of a desktop simulation and a single SA measure, the approach demonstrates the feasibility of multimodal SA prediction and offers actionable insights for safer driver–AV interaction in conditional automation.
Abstract
Maintaining adequate situation awareness (SA) is crucial for the safe operation of conditionally automated vehicles (AVs), which requires drivers to regain control during takeover (TOR) events. This study developed a predictive model for real-time assessment of driver SA using multimodal data (e.g., galvanic skin response, heart rate and eye tracking data, and driver characteristics) collected in a simulated driving environment. Sixty-seven participants experienced automated driving scenarios with TORs, with conditions varying in risk perception and the presence of automation errors. A LightGBM (Light Gradient Boosting Machine) model trained on the top 12 predictors identified by SHAP (SHapley Additive exPlanations) achieved promising performance with RMSE=0.89, MAE=0.71, and Corr=0.78. These findings have implications towards context-aware modeling of SA in conditionally automated driving, paving the way for safer and more seamless driver-AV interactions.