Scholarlink

<p>Driver fatigue remains a major contributor to heavy-vehicle crashes worldwide, particularly in long-haul freight operations characterized by extended duty cycles and irregular sleep schedules. This study proposes and validates a multimodal fatigue assessment framework integrating physiological, psychological, and vehicular telemetry data under real-world naturalistic driving conditions along National Highway 44 (NH-44), India. A total of 400 time-synchronized driving segments (201 fresh, 199 fatigued) were collected from 30 professional truck drivers. Multimodal acquisition included wearable physiological sensors, standardized psychological scales, and high-resolution vehicle telemetry, ensuring precise temporal alignment across modalities. Physiological indicators (heart rate, heart rate variability, electrodermal activity, skin temperature, and blood pressure), subjective measures (Karolinska Sleepiness Scale and mood scores), and driving behaviour parameters (speed variability, braking intensity, steering deviation) were integrated using an Analytic Hierarchy Process (AHP)-based weighting approach to construct a Composite Fatigue Index (CFI). Receiver Operating Characteristic analysis yielded an AUC of 0.938, indicating strong discriminative capability. At an optimized threshold of 0.46, the framework achieved 80% accuracy, 86.7% sensitivity, and stable five-fold cross-validation accuracy of 0.833. Although Random Forest and Support Vector Machine models demonstrated slightly higher raw accuracy, the proposed CFI offers superior interpretability, structured weighting transparency, and practical deployability in safety–critical contexts. Geostatistical validation further demonstrated spatial clustering of fatigue hotspots along corridor segments, confirming the geospatial coherence of the classification framework. Overall, multimodal integration significantly enhances early fatigue detection and provides a scalable, interpretable solution for real-time driver monitoring and freight safety management.</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

Multimodal Framework for Truck Driver Vigilance and Driving Behavior Classification Using Physiological, Psychological, and Telemetry Data in Naturalistic Driving Conditions

  • 期刊论文

  • 发表日期：

    2026年3月30日

查看全文
  • Safa Koul,
  • L. Janani,
  • Mohammad Shafi Mir,
  • Saqib Bashir,
  • Sheikh Tanveer Ahmad

摘要

Driver fatigue remains a major contributor to heavy-vehicle crashes worldwide, particularly in long-haul freight operations characterized by extended duty cycles and irregular sleep schedules. This study proposes and validates a multimodal fatigue assessment framework integrating physiological, psychological, and vehicular telemetry data under real-world naturalistic driving conditions along National Highway 44 (NH-44), India. A total of 400 time-synchronized driving segments (201 fresh, 199 fatigued) were collected from 30 professional truck drivers. Multimodal acquisition included wearable physiological sensors, standardized psychological scales, and high-resolution vehicle telemetry, ensuring precise temporal alignment across modalities. Physiological indicators (heart rate, heart rate variability, electrodermal activity, skin temperature, and blood pressure), subjective measures (Karolinska Sleepiness Scale and mood scores), and driving behaviour parameters (speed variability, braking intensity, steering deviation) were integrated using an Analytic Hierarchy Process (AHP)-based weighting approach to construct a Composite Fatigue Index (CFI). Receiver Operating Characteristic analysis yielded an AUC of 0.938, indicating strong discriminative capability. At an optimized threshold of 0.46, the framework achieved 80% accuracy, 86.7% sensitivity, and stable five-fold cross-validation accuracy of 0.833. Although Random Forest and Support Vector Machine models demonstrated slightly higher raw accuracy, the proposed CFI offers superior interpretability, structured weighting transparency, and practical deployability in safety–critical contexts. Geostatistical validation further demonstrated spatial clustering of fatigue hotspots along corridor segments, confirming the geospatial coherence of the classification framework. Overall, multimodal integration significantly enhances early fatigue detection and provides a scalable, interpretable solution for real-time driver monitoring and freight safety management.