Drowsiness Monitoring By Steering And Lane Data Based Features Under Real Driving Conditions

Friedrichs, Fabian · 2010 · OpenAlex-citations

DOI: 10.5281/zenodo.42038

archive: archived pipeline: cataloged verified

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

Summary

This paper addresses the challenge of detecting driver drowsiness using vehicle sensor data under real-world driving conditions, a significant safety issue given that drowsiness contributes to approximately 30% of severe traffic accidents. While previous research has largely relied on driving simulator data, this study utilizes a large-scale database of real-world drives provided by Mercedes-Benz. The primary motivation is to develop robust drowsiness monitoring features that account for external influences such as road conditions, traffic, and varying driving styles, which are often absent in controlled simulator environments. The study aims to introduce new features based on Controller Area Network (CAN) data, propose efficient feature extraction methods, and evaluate their performance in distinguishing between awake, questionable, and drowsy states. The methodology relies on a dataset comprising over 1.2 million kilometers of measurements from 54 drivers across 10 vehicles, filtered to 52,768 kilometers of valid drives. Drowsiness levels were assessed using the Karolinska Sleepiness Scale (KSS), a subjective self-rating method, as objective measures like EEG and eye-tracking were not consistently available. The researchers extracted 48 features categorized into lane-based, steering wheel angle-based, and CAN-bus signals. These features included metrics such as lane deviation, steering velocity quartiles, zig-zag events, and time-on-task. To handle individual driver variations, the study employed baselining techniques to normalize features against initial driving behavior. Additionally, the authors introduced Exponentially Weighted Moving Average (EWMA) and Variance (EWVAR) filters for more efficient and adaptive feature processing compared to traditional moving averages. The results indicate that while individual features showed correlations with drowsiness, the classification performance in real-world conditions was lower than that achieved in driving simulators. Feature selection was performed using Sequential Floating Forward Selection (SFFS), which identified the most effective subset of features for classification. The study compared various classifiers, including k-nearest neighbors, linear discriminant analysis, and artificial neural networks. The analysis revealed that features such as circadian weighting, time-on-task, and degree of interaction were frequently selected. However, the classes of awake, questionable, and drowsy drivers exhibited severe overlap in feature space, posing significant challenges for accurate classification. The best-performing models still struggled to match the high accuracy rates reported in simulator-based studies, largely due to the noise and variability inherent in real-world driving data. The significance of this work lies in its comprehensive evaluation of drowsiness detection using real-world data, highlighting the limitations of current feature sets and classification methods when applied outside controlled environments. The study demonstrates that while CAN-based and steering features are promising, they require sophisticated adaptation to driver-specific behaviors and environmental conditions to be effective. The findings suggest that future systems must account for the complexity of real-world driving, including external disturbances and individual driving styles, to improve the reliability of drowsiness monitoring. This research provides a foundation for developing more robust driver assistance systems that can effectively reduce accidents caused by fatigue in everyday driving scenarios.

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
• lane positioning
• microsleep
• steering pattern
• 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: validation psychometrics, tool software