Exploring Naturalistic Driving Data for Distracted Driving Measures

Ishak, Sherif S.; Osman, Osama A.; Codjoe, Julius; Jenkins, Syndney; Karbalaieali, Sogand; Theriot, Matthew; Bakhit, Peter · 2017 · ROSA P / Louisiana Transportation Research 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 challenge of identifying reliable surrogate measures for distracted driving using the Second Strategic Highway Research Program (SHRP 2) Naturalistic Driving Study (NDS) dataset. Motivated by the need to quantify crash risks associated with secondary tasks and the limitations of previous studies due to small sample sizes, the researchers aimed to explore the extensive SHRP 2 NDS data to detect distracted driving behaviors. The specific secondary tasks analyzed included talking or listening on a hand-held phone, texting or dialing on a hand-held phone, and interacting with an adjacent passenger. The study also sought to outline a framework for a crash index model to quantify the risk associated with these behaviors. The methodology involved extracting time-series data for five performance measures: GPS speed, lateral acceleration, longitudinal acceleration, throttle position, and yaw rate. The researchers first employed multiple logistic regression (MLR) to determine the odds of a driver engaging in secondary tasks based on these measures, analyzing results across different age groups (16–29 and 70–89) and genders. Due to poor model fit indicated by undesirable Hosmer and Lemeshow Test p-values in the MLR analysis, the team subsequently applied neural network modeling, leveraging its capability for nonlinear pattern recognition. Data preparation included cleaning, aggregation, and stratification to ensure independence among groups. The MLR results were largely inconclusive regarding statistical fit, though lateral acceleration emerged as a useful indicator for detecting phone-related distractions. Demographic analysis revealed that longitudinal acceleration performed better in predicting distractions for drivers aged 70–89, while lateral acceleration was more effective for younger drivers (16–29). For both genders, lateral acceleration proved more effective in predicting texting/dialing and talking/listening. However, the neural network analysis yielded more robust results, demonstrating that the five selected performance measures could serve as effective surrogate measures for distracted driving. The neural network models successfully detected drivers’ engagement in secondary tasks with high accuracy, overcoming the limitations of the linear regression models. The significance of this research lies in validating the utility of SHRP 2 NDS data for distracted driving studies and establishing specific vehicle dynamics metrics as reliable surrogates for distraction. The study concludes that artificial intelligence tools, specifically neural networks, are superior to traditional statistical methods for detecting nonlinear patterns in driving behavior associated with distraction. Additionally, the report provides a proposed framework for calculating a crash risk index, offering a pathway for future research to quantify the safety implications of distracted driving. These findings support transportation officials in developing more accurate safety assessments and enforcement strategies based on naturalistic driving data.

Key finding

Neural network models successfully utilized lateral and longitudinal acceleration, along with other vehicle dynamics data, to accurately detect driver engagement in secondary tasks like texting and phone conversation.

Methodology

naturalistic

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.

• distraction detection algorithms
• temporal
• naturalistic crash near crash
• visual
• exposure measurement
• manual

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: observational prevalence, behavioral performance data
• Methodological Resource: dataset resource