Investigating the Feasibility of Vehicle Telemetry Data as a Means of Predicting Driver Workload

Taylor, Phillip; Griffiths, Nathan; Bhalerao, Abhir; Xu, Zhou; Gelencser, Adam; Popham, Thomas · 2017 · Crossref

DOI: 10.4018/ijmhci.2017070104

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 investigates the feasibility of using vehicle telemetry data to predict driver workload, addressing the safety risks associated with cognitive distraction. Traditional workload monitoring relies on intrusive physiological sensors or subjective self-reports, which are impractical for continuous, real-world application. The authors propose that non-intrusive vehicle telemetry, accessible via the Controller Area Network (CAN-bus), can serve as a viable proxy for driver workload. The research aims to determine if predictive models built from telemetry data can accurately estimate workload levels defined by physiological markers. The researchers collected the Warwick-JLR Driver Monitoring Dataset (DMD) using thirteen participants driving a Range Rover Sport on a controlled test track. To induce varying levels of cognitive workload, participants performed N-back tasks (0-back, 1-back, and 2-back) while driving at approximately 70 mph. Data collection included physiological measures—Heart Rate (HR), Heart Rate Variability (HRV), Skin Conductance (SC), and Electrodermal Response (EDR)—alongside over 1,000 vehicle telemetry signals recorded at 20 Hz. The experimental protocol involved habituation, baseline driving, distraction periods with N-back tasks, and recovery phases. Statistical analysis, including two-way t-tests and ANOVA, was conducted to compare normal driving against distracted conditions. The results demonstrated significant correlations between driver workload and both physiological and telemetry data. Physiologically, HR, SC, and EDR frequency increased significantly during distraction periods, particularly during the more demanding 1-back and 2-back tasks, while HRV showed no significant change. Telemetry analysis revealed that signals directly related to vehicle control, such as throttle position and steering wheel angle momentum, exhibited significant statistical differences between normal and distracted driving. Indirect signals, such as engine speed and gear selection, also showed relationships with workload, whereas unrelated signals like air conditioning controls did not. The study further outlines a data mining methodology to build predictive classification models using these telemetry features to estimate workload status based on physiological ground truth. The findings confirm that vehicle telemetry data contains sufficient information to distinguish between normal and high-workload driving states. By identifying specific telemetry signals that correlate with physiological stress markers, the study supports the feasibility of developing non-intrusive, real-time driver monitoring systems. This approach offers a practical alternative to intrusive sensors, enabling vehicles to adapt functionality or alert drivers when cognitive workload exceeds safe levels, thereby enhancing driving safety without compromising user comfort or privacy.

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.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

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

• workload measurement
• exposure measurement
• mental demand
• telematics crash prediction
• stress driving
• 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