Monitoring driver drowsiness in partially automated vehicles: Added value from combining postural and physiological indicators

Perrotte, Gaëtan; Bougard, Clément; Portron, Arthur; Vercher, Jean-Louis · 2024 · Crossref

DOI: 10.1016/j.trf.2023.12.010

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of monitoring driver drowsiness in partially automated vehicles (Level 2 automation), where traditional Driver Monitoring Systems (DMS) relying on facial cues and vehicle trajectory data become less effective. As drivers transition from active operators to supervisors, they may engage in non-driving tasks or look away from the road, rendering camera-based facial analysis unreliable. The authors investigate whether combining postural indicators (seat pressure and movement) with physiological indicators (cardiac and respiratory data) provides added value for detecting the full spectrum of drowsiness, from alertness to sleep. The experimental design involved 22 participants driving for approximately 100 minutes in a static simulator under Level 2 automation. The session included a 90-minute autonomous phase characterized by long periods of no traffic, interspersed with a traffic jam phase intended to reactivate drivers. Data were continuously recorded using a Drowsimeter for ocular metrics (PERCLOS70), a BIOPAC system for physiological signals (ECG and respiration), and textile pressure sensor mats for postural data. Drowsiness was classified into five states—Alert, Slightly Drowsy, Drowsy, Extremely Drowsy, and Asleep—based on PERCLOS70 thresholds. Statistical analyses, including ANOVA and post-hoc tests, compared these indicators against the alert baseline, with data rescaled to minimize inter-individual variability. The results revealed distinct physiological and postural signatures for different drowsiness states. Slight drowsiness was characterized by a higher heart rate, slower breathing, and an increased number of movements on the seat, suggesting a struggle to maintain alertness. In contrast, being asleep was marked by a lower heart rate, reduced heart rate variability, and a slouched posture with increased contact pressure on the posterior seat and headrest. Ocular metrics confirmed expected trends, with blink duration increasing significantly in deeper drowsiness states. Postural features, particularly center of pressure movements and headrest usage, showed significant effects across all drowsiness levels, demonstrating sensitivity to the transition from wakefulness to sleep. The study concludes that combining postural and physiological indicators offers a robust alternative for monitoring driver states in automated driving contexts. Unlike facial recognition, seat-based sensors are non-intrusive and effective even when the driver is not facing the camera. The findings highlight that early drowsiness involves active resistance (increased movement and heart rate), while sleep onset involves physiological relaxation and postural collapse. This distinction is critical for developing DMS capable of detecting intermediate drowsiness states, thereby enhancing safety in partially autonomous vehicles by ensuring drivers remain capable of taking control when required.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified_with_issues.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms
• drowsiness
• microsleep
• truck driver fatigue

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Methodological Resource: tool software