Estimating Markers of Driving Stress through Multimodal Physiological Monitoring

TL;DR

This study develops a multimodal machine learning system to estimate driving stress by analyzing physiological signals, behavioral responses, and environmental factors in a controlled simulation, aiming to enhance road safety and driver well-being.

Contribution

It introduces a novel approach combining physiological, behavioral, and contextual data to accurately estimate driving stress in real-time.

Findings

Model effectively detects stressors in simulated driving scenarios

Physiological signals correlate with observable vehicle control patterns

Temporal analysis reveals dynamics of stress responses

Abstract

Understanding and mitigating driving stress is vital for preventing accidents and advancing both road safety and driver well-being. While vehicles are equipped with increasingly sophisticated safety systems, many limits exist in their ability to account for variable driving behaviors and environmental contexts. In this study we examine how short-term stressor events impact drivers' physiology and their behavioral responses behind the wheel. Leveraging a controlled driving simulation setup, we collected physiological signals from 31 adult participants and designed a multimodal machine learning system to estimate the presence of stressors. Our analysis explores the model sensitivity and temporal dynamics against both known and novel emotional inducers, and examines the relationship between predicted stress and observable patterns of vehicle control. Overall, this study demonstrates the potential of linking physiological signals with contextual and behavioral cues in order to improve real-time estimation of driving stress.