Classification of Drivers' Workload Using Physiological Signals in Conditional Automation
DOI: 10.3389/fpsyg.2021.596038
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 safety challenges associated with conditional automation (SAE Level 3) in vehicles, where drivers may engage in non-driving-related tasks (NDRTs) but must be ready to take over control during hazardous situations. High mental workload (MWL) induced by secondary tasks can impair takeover performance, increasing accident risk. The research aims to determine if physiological signals can accurately classify drivers’ MWL in real-time to support dynamic safety systems. While previous methods relied on subjective ratings or task performance metrics, these are often impractical for continuous, real-world monitoring. The authors propose using machine learning to analyze autonomic nervous system indicators, specifically electrodermal activity (EDA), electrocardiogram (ECG), and respiration, which offer non-intrusive, continuous data streams suitable for embedded sensors. The experimental design involved 90 participants driving in a fixed-base simulator for 25 minutes under conditional automation. The study employed a between-subjects design: half the participants performed a verbal cognitive task to induce high MWL, while the other half only monitored the environment (low MWL). Three physiological signals were recorded: EDA (skin conductance), ECG (heart rate and variability), and respiration. The researchers compared three classifiers, various sensor fusion strategies, and different data segmentation levels (time windows) to evaluate classification accuracy. The goal was to identify the optimal configuration for detecting high workload states using physiological data alone, without relying on driving performance metrics which are less relevant when the vehicle is automated. The results demonstrated that physiological signals could effectively distinguish between low and high mental workload conditions. The best-performing model achieved a classification accuracy of 95%. The study found that sensor fusion—combining data from multiple physiological sources—improved performance in certain configurations. Crucially, the analysis of data segmentation revealed that increasing the time window size improved model performance for windows smaller than 4 minutes, but performance decreased for windows larger than 4 minutes. This suggests an optimal balance between having enough data for robust feature extraction and maintaining timely detection. The most effective classification was achieved using 4-minute recordings of respiration and skin conductance data. The significance of this work lies in its demonstration that high mental workload in conditionally automated driving can be detected with high accuracy using non-intrusive physiological sensors. The finding that 4-minute windows provide optimal performance offers practical guidance for designing real-time monitoring systems in future vehicles. By accurately identifying when a driver’s cognitive load exceeds safe thresholds, automated systems could adapt their behavior, such as modifying takeover request alerts or adjusting automation levels, thereby enhancing safety and user experience. This study supports the feasibility of integrating physiological monitoring into the human-machine interface of next-generation automated vehicles to mitigate risks associated with driver distraction and delayed takeover responses.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-08-09 |
| archive | success | canonical_url | — | — | 1 | 2026-08-09 |
| extract | success | cached | — | — | 3 | 2026-08-10 |
| clean | success | clean | — | — | 1 | 2026-08-09 |
| chunk | success | chunk | — | — | 1 | 2026-08-09 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-08-09 |
| promote | success | — | — | — | 1 | 2026-08-09 |
| summarize | success | llm | qwen3.6-27b-nvidia | summ-v5 | 2 | 2026-08-10 |
| tag | success | vector_similarity | — | — | 11 | 2026-08-11 |
| verify | partial | — | — | — | 2 | 2026-08-10 |
Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- workload measurement
- drowsiness detection algorithms
- stress driving
- mental demand
- distraction detection algorithms
- neuro workload indices
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
- Theoretical Contribution: theory or model