Driver distraction effects and explainable workload recognition in extra-long tunnels using visual, ECG, and vehicle-control features
Related Experiment Video
Updated: Nov 5, 2026
Tactile Vibrating Toolkit and Driving Simulation Platform for Driving-Related Research
Published on: December 18, 2020
Yingping Tang1, Zhigang Du1, Xu Ding1
1School of Transportation and Logistics Engineering, Wuhan University of Technology, Wuhan, China.
Objective:
To address driving distraction caused by reduced driver alertness in the middle sections of extra‑long highway tunnels, which poses a threat to driving safety, this study systematically examines the effects of manual, visual, and cognitive distraction. It investigates their influences on visual indicators, ECG signals, vehicle control performance, and subjective workload. Furthermore, this work constructs a high‑precision interpretable driving workload recognition model.
Methods:
Baseline and three distraction conditions were established in driving simulation experiments, with synchronized collection of multidimensional visual, ECG, vehicle control, and subjective data. Repeated‑measures ANOVA, kernel density estimation, and gaze distribution analysis were employed to examine distraction characteristics, and a LightGBM‑SHAP fusion model was constructed for workload recognition and feature contribution quantification.
Results:
Manual and visual distraction both caused gaze deviation and reduced vehicle control stability, while cognitive distraction reduced drivers' visual scanning of the surrounding environment. Manual distraction exerted a certain effect on ECG indicators, and its mean subjective workload was the highest. The LightGBM‑SHAP model outperformed traditional models in recognition accuracy. SHAP analysis identified LF/HF and SDLP as core features for high‑workload classification, and the model demonstrated both high recognition performance and interpretability.
Conclusion:
In this tunnel scenario, under experimentally induced distraction conditions, manual and visual distraction had certain effects on driving stability and drivers' physiological indicators, while cognitive interference also altered driver workload and attention characteristics. The proposed model provides a high‑precision, interpretable method for driving state assessment and is applicable to distraction recognition and warning in tunnel scenarios.
