What's on your mind? A Mental and Perceptual Load Estimation Framework towards Adaptive In-vehicle Interaction while Driving

Gomaa, Amr; Alles, Alexandra; Meiser, Elena; Rupp, Lydia Helene; Molz, Marco; Reyes, Guillermo · 2022 · arXiv

DOI: 10.48550/arxiv.2208.05564

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 estimating driver mental workload (MWL) and perceptual load (PL) to enable adaptive in-vehicle interfaces that enhance safety and user experience. While MWL has been extensively studied, PL—defined as the extent to which a task consumes available perceptual capacity—remains underexplored in driving contexts, particularly regarding its distinct psychophysiological signatures. The authors aim to differentiate these two constructs and develop a machine learning framework for real-time estimation using non-intrusive sensors compatible with modern vehicles. The study employed a within-subject dual-task design with 45 participants in a driving simulator. Participants performed a primary lane-changing task while simultaneously engaging in one of two secondary tasks: an auditory n-back task to induce varying levels of MWL (1-back, 2-back, 3-back) or a visual search task to induce varying levels of PL (2x2, 3x3, 4x4 matrix sizes). Psychophysiological data were collected using a chest-mounted ECG sensor for heart rate (HR) and heart rate variability (HRV), and a head-mounted eye-tracker for pupil diameter, processed using the Low/High Index of Pupillary Activity (LHIPA). Driving performance was measured via lateral deviation from an ideal path. The authors utilized an automated machine learning pipeline (auto-sklearn) to classify load levels, employing nested cross-validation to ensure generalization. Statistical analysis revealed that MWL significantly influenced HR and HRV, with easy conditions differing significantly from medium and hard conditions combined, though no significant difference existed between medium and hard levels. In contrast, PL showed little to no significant effect on these psychophysiological measures, despite successful manipulation of task difficulty. Reliability analysis excluded LHIPA and driving performance metrics from further modeling due to low internal consistency. Consequently, machine learning models using eye or driving data performed at chance levels. However, models using heart rate data alone achieved up to 89% accuracy in binary classification (low vs. medium/high MWL). Multi-class MWL classification reached approximately 49% accuracy, while PL classification remained at chance levels across all models. The findings indicate that while non-intrusive heart rate sensors can effectively distinguish between low and elevated mental workload, they are insufficient for detecting perceptual load or differentiating between medium and high workload levels. The lack of psychophysiological sensitivity to PL suggests that current non-intrusive measures may not capture the specific cognitive demands of visual distraction. The authors provide an open-source framework and dataset, highlighting the potential for adaptive interfaces based on MWL while noting the need for better metrics to capture perceptual load in driving scenarios.

Key finding

Mental workload during driving can be effectively measured through psychophysiological indicators and secondary task performance, with cognitive load showing measurable impacts on lane keeping and speed control even when drivers report no subjective awareness of impairment.

Methodology

lab_experiment

Sample size: 28

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 discover_arxiv on 2026-05-04 (5 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-07; verification: verified.

Topics

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

• workload measurement
• mental demand
• stress driving
• situational awareness
• drowsiness detection algorithms
• distraction detection algorithms

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