Advanced Driver Fatigue Research

NHTSA · 2008 · ROSA P / United States. Federal Motor Carrier Safety Administration

archive: archived pipeline: cataloged verified

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

Summary

This report details the Advanced Driver Fatigue Research project, conducted by the Center for Intelligent Systems Research (CISR) at George Washington University and funded by the Federal Motor Carrier Safety Administration (FMCSA). The primary objective was to develop and validate an unobtrusive drowsy driver detection system specifically for commercial motor vehicles. While CISR had previously established that steering-pattern analysis could effectively detect drowsiness in passenger automobiles, it remained unclear whether this method would translate to truck driving due to differences in vehicle dynamics, driver experience, and steering feel. The study aimed to confirm the validity of a steering-only detection approach for truckers and to evaluate its performance against systems that also incorporate eye-tracking data. To achieve this, the researchers constructed a full-size, fixed-base truck driving simulator in partnership with the French National Institute for Transport and Safety Research (INRETS). The simulator featured a realistic cab, force-feedback steering, and a 135-degree field of view, validated for realism by expert truck drivers. The experimental design involved fourteen licensed commercial truck drivers who participated in three sessions: a practice session, a baseline morning session, and a late-night session designed to induce fatigue. During the night sessions, most subjects experienced episodes of drowsiness. Data collected included steering activity, eye closure metrics (PERCLOS), and video recordings. The researchers developed two Artificial Neural Network (ANN) models to classify driving intervals as wake or drowsy: one using only steering angle data and another combining steering data with eye closure inputs. A unique data preprocessing scheme was employed to discretize steering signals, allowing a single ANN model to generalize across different drivers and driving styles. The results demonstrated that the steering-only approach was equally valid for truck drivers as it was for automobile drivers. The ANN model relying solely on steering activity achieved an 85% accuracy rate in predicting drowsy intervals, with a false alarm rate of 14%. The combined steering-and-eye-tracking model performed slightly better, with 88% accuracy and a 9% false alarm rate. Crucially, the steering-only system proved highly effective at crash prevention; it issued a timely warning in 100% of the first two crashes experienced by any subject. On average, the system provided 5.6 warnings in the five minutes preceding a crash, with the initial warning occurring approximately 3 minutes and 56 seconds before the incident. The study confirmed a strong correlation between high-amplitude steering corrections and drowsiness, validating the hypothesis that fatigued drivers exhibit fewer micro-steering corrections and more macro-corrections. The significance of this research lies in the demonstration that an unobtrusive, steering-only detection system is a viable and cost-effective solution for commercial motor vehicles. Unlike eye-tracking systems, which can be degraded by low-light conditions or eyewear, the steering-based system is robust and requires minimal additional sensors. The findings suggest that this technology is ready for prototype development and field operational testing. By providing early warnings, such systems can mitigate the severity of fatigue-related crashes, addressing a leading cause of serious injuries and fatalities in trucking. The report also outlines design goals for warning systems, emphasizing the need for appropriate timing and intensity to ensure driver acceptance and effective behavioral adaptation.

Key finding

The steering-only drowsy detection system issued timely warnings in 100% of the simulated crashes, with the first warning occurring on average 3 minutes and 56 seconds before impact.

Methodology

simulator

Sample size: 14

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. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• drowsiness detection algorithms
• drowsiness
• truck driver fatigue
• sleep deprivation
• time on task
• vigilance

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, validation psychometrics