Predicting Changes in Driving Safety Performance on an Individualized Level Under Naturalistic Driving Conditions

Cavuoto, Lora; Megahed, Fadel · 2019 · ROSA P / Transportation Informatics University Transportation Center

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical public safety issue of driver-related crashes, particularly among commercial truck drivers, where fatigue and distraction are leading causes. Despite advancements in vehicle technology, driver-related factors contribute to 75–90% of fatal crashes. The research aims to model changes in driving safety performance (DSP) on an individualized level under naturalistic conditions by leveraging data from wearable sensors. The goal is to predict hazardous behaviors and unsafe acts before they occur, enabling proactive interventions rather than reactive detection. The researchers utilized data from the Co-Pilot SE™ wearable headset, which captures head movements and acceleration. The dataset comprised over 10 million events from approximately 200 truck drivers between August 2016 and October 2017. After rigorous cleaning to remove infeasible data points, the analysis focused on 9.8 million events across 198 drivers. Key input variables included trip duration, time of day, median speed, proportion of downward glances (indicating distraction), and the balance of left-to-right mirror checks. Outcome variables were defined as the time between mirror checks and the occurrence of hard brake events. The study employed logistic regression to determine if input variables predicted an "unsafe" mirror check rate, defined as fewer than six checks per minute (greater than 10 seconds between checks), contrary to Federal Motor Carrier Safety Administration guidelines. The findings revealed that 61.7% of events occurred during the day, with drivers looking left 58.2% of the time, right 35.5%, and down 6.3%. Approximately 45% of extended gaps between mirror checks occurred when drivers were on the phone. Logistic regression analysis achieved a 66% prediction performance for classifying safe versus unsafe mirror check rates. The results identified two significant predictors of unsafe performance: the ratio of left-to-right mirror checks and trip duration. An imbalance in mirror checks (higher left-to-right ratio) was associated with longer separation times between checks. Additionally, trips with unsafe mirror check rates had a significantly longer average duration (225 minutes) compared to safe trips (185 minutes). Median speed and the percentage of downward glances were not statistically significant predictors in this model. The significance of this work lies in its demonstration that individualized, data-driven modeling can identify specific behavioral patterns linked to degraded safety performance. By establishing that trip duration and mirror check imbalance are strong predictors of unsafe behavior, the study provides a foundation for real-time monitoring systems. These insights allow for the development of predictive models that can alert drivers or fleet managers to potential fatigue or distraction risks before critical events, such as hard braking or crashes, occur. This approach shifts the focus from post-incident analysis to proactive safety management in commercial trucking.

Key finding

Trip duration and left-right mirror check balance significantly predicted unsafe mirror check rates, while median speed and looking down were not significant predictors.

Methodology

naturalistic

Sample size: 198

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 bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

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

• telematics crash prediction
• naturalistic crash near crash
• exposure measurement
• drowsiness detection algorithms
• temporal
• vigilance

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: behavioral performance data, crash risk outcomes, observational prevalence