DrivingResearch.

Driver State Classification: Identifying High Cognitive Load and Drowsiness Through Driver Performance

Ayas, Suzan; He, Dengbo; Donmez, Birsen · 2026 · IEEE Transactions on Intelligent Transportation Systems

DOI: 10.1109/tits.2025.3633499

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 critical challenge of accurately distinguishing between high cognitive load, alertness, and drowsiness in driver monitoring systems (DMS). Existing literature often treats these states in isolation or shows conflicting results regarding their physiological and performance markers, leading to potential misclassifications. Such errors are dangerous, as interventions designed for one state (e.g., alerting a drowsy driver) may exacerbate another (e.g., increasing stress in an already cognitively overloaded driver). The authors aim to identify specific driving performance and physiological metrics that differentiate these three states and evaluate the efficacy of machine learning models in classifying them within a single framework. The researchers conducted a within-subject driving simulator experiment with 27 participants. High cognitive load was induced using auditory-verbal n-back tasks (1-back and 2-back), while drowsiness was induced through extended periods of monotonous driving. Drowsiness levels were rated by trained observers using webcam video analysis based on behavioral indicators. Data collected included driving performance metrics (speed, lane position, steering wheel movement) and physiological measures (heart rate, heart rate variability, galvanic skin response). Mixed linear models analyzed differences between states, while five machine learning algorithms (Random Forest, XGBoost, SVM, KNN, and MLP) were trained to classify drivers into high cognitive load, alert, or drowsy categories. Results indicated distinct patterns for each state. High cognitive load was associated with greater physiological arousal, increased speed variation, reduced average speed, and decreased standard deviation of lane position. Conversely, drowsiness correlated with lower physiological arousal, increased average speed, and higher variability in lane position and steering wheel angle. Machine learning analysis revealed that tree-based ensemble models, specifically Random Forest and XGBoost, performed best. Using simple features like averages and standard deviations, these models achieved up to 76% average accuracy in multi-class classification. The models differentiated high cognitive load with an Area Under the Curve (AUC) of approximately 85% and drowsiness with an AUC of approximately 79%. The findings demonstrate that driving performance and physiological data can successfully distinguish between cognitive overload, alertness, and underload (drowsiness) in a single model. This capability is significant for the development of functional DMS, particularly given regulatory mandates for driver attention warning systems. By accurately separating these states, future systems can deploy appropriate, state-specific interventions—such as reducing interface complexity for overloaded drivers or issuing alarms for drowsy drivers—thereby minimizing classification errors and enhancing road safety.

Key finding

Tree-based ensemble machine learning models achieved up to 76% average accuracy in classifying high cognitive load, alert, and drowsy driver states using driving performance and physiological metrics.

Methodology

simulator

Sample size: 27

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.

StageOutcomeToolModelPromptAttemptsCompleted
discover success — — — 1 2026-05-28
archive success canonical_url — — 1 2026-06-06
extract success cached — — 3 2026-06-10
clean success clean — — 1 2026-06-07
chunk success chunk — — 1 2026-06-07
embed success embed Qwen/Qwen3-Embedding-8B — 1 2026-06-07
enrich success semantic_scholar — — 4 2026-06-15
promote success — — — 1 2026-06-04
summarize success llm qwen3.6-27b-prismaquant summ-v5 2 2026-06-10
tag success vector_similarity — — 15 2026-06-11
verify success — — — 2 2026-06-10

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.

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).