Predicting fatigue-related aberrant driving behaviours in taxi drivers: a multi-modal approach using physiological, sleep, and demographic data
File(s)
Accepted version
Author(s)
Type
Journal Article
Abstract
Taxi drivers face elevated risks of fatigue-related aberrant driving behaviours (ADBs), exacerbated by sleep-related factors, yet predictive models incorporating these factors remain underexplored. This study employed a multi-modal approach to assess the predictive value of heart rate variability (HRV) metrics for fatigue-related ADBs and to assess the added value of integrating demographic and sleep data. A four-day naturalistic driving study involving 38 taxi drivers combined survey-based data with physiological, trajectory, and sleep monitoring. Participants with severe obstructive sleep apnea (OSA) exhibited significantly more hard accelerations, abrupt braking, and sharp steering compared to those without severe OSA. Three predictive models were developed: a long short-term memory (LSTM) model using HRV metrics; hybrid model A that combined HRV metrics, self-reported data, a hospital-based apnea-hypopnea index (AHI), and prior-night sleep disorder indices; and hybrid model B that extended hybrid model A by incorporating sleep disorder indices from the two preceding nights. The hybrid models outperformed the unidimensional LSTM model, with hybrid model B achieving the best performance (accuracy: 94.26%, specificity: 96.08%, sensitivity: 82.07%). The oxygen desaturation index (ODI-3%) emerged as the strongest predictor of ADB risk. These findings underscored the value of cumulative sleep data for fatigue-related ADB prediction and fatigue management.
Date Issued
2026-01-01
Date Acceptance
2026-05-04
Citation
Behaviour and Information Technology
ISSN
0144-929X
Publisher
Taylor and Francis Group
Journal / Book Title
Behaviour and Information Technology
Copyright Statement
Copyright © 2026 Informa UK Limited, trading as Taylor & Francis Group. This is the author’s accepted manuscript made available under a CC-BY licence in accordance with Imperial’s Research Publications Open Access policy (www.imperial.ac.uk/oa-policy)
License URL
Publication Status
Published online
Date Publish Online
2026-05-19