Investigating Relationship Between Driving Patterns and Traffic Safety Using Smartphones Based Mobile Sensor Data

Sahin, Olcay; Paleti, Rajesh; Cetin, Mecit · 2016 · ROSA P / Mid-Atlantic Transportation Sustainability 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 investigates the relationship between microscopic driving patterns and traffic safety, addressing a gap in existing literature that primarily relies on aggregate traffic measures and roadway geometry. The authors argue that traditional Safety Performance Functions (SPFs) may yield biased estimates because they fail to account for high-resolution vehicle dynamics, such as speed and acceleration, which are critical indicators of unsafe driving behaviors. With the advent of smartphone sensors capable of recording fine-grained temporal data, the research aims to quantify how these microscopic metrics correlate with crash frequencies along highway segments. The methodology involved collecting mobile sensor data from smartphones equipped with GPS, accelerometers, and gyroscopes, linked to On-Board Diagnostic (OBD) devices via Bluetooth. Data was gathered using the floating car technique on major interstates in the Hampton Roads region of Virginia during the weekday evening peak period (4:00–6:00 pm) over one year. This probe data was merged with Virginia Department of Transportation (VDOT) crash records, roadway inventory data, and traffic exposure information from 222 Wavetronix sensors. The analysis focused on 513 unique roadway segments. The researchers developed and compared several statistical models, including Poisson, Negative Binomial (NB), NB with Heterogeneous Dispersion (NB HD), Zero-Inflated models, and Generalized Ordered Response (GOR) models. They also incorporated random parameter heterogeneity and spatial dependency effects to account for unobserved factors and the influence of neighboring segments. The findings demonstrate that models incorporating microscopic traffic measures significantly outperform traditional models. The best-performing model was the GOR variant of the NB model with heterogeneous dispersion, random heterogeneity, and spatial effects. Key results indicate that higher traffic exposure, lower vehicle speeds, and greater variation in acceleration profiles are associated with increased crash frequencies. Specifically, a 10% increase in traffic exposure raised crash frequency by 2.1%, while a 10% increase in speed decreased it by 5.5%, suggesting that low-speed congestion is riskier than free-flow conditions. Segments with extreme accelerations but no extreme decelerations experienced 40% fewer crashes, highlighting the danger of stop-and-go movements. Additionally, spatial dependency was significant; acceleration patterns in neighboring segments influenced crash risk, with the spatial effect (9% increase in crash frequency per 10% increase in neighbor acceleration variance) exceeding the direct effect (3.4%). Roadway geometry variables were not significant in the final model, indicating that previous models may have overestimated their impact to compensate for unmodeled dispersion. The study concludes that integrating high-resolution mobile sensor data into crash frequency modeling improves both statistical fit and policy relevance. The authors recommend that transportation agencies update SPFs to include microscopic traffic measures, enabling more accurate safety predictions and better evaluation of countermeasures targeting driving behavior, such as variable speed limits. Furthermore, they advocate for adopting advanced count modeling techniques like NB HD and GOR frameworks, which offer superior flexibility and accuracy compared to standard Poisson and NB models without significant computational complexity.

Key finding

Models incorporating microscopic speed and acceleration data from smartphones significantly outperformed traditional models, revealing that acceleration patterns in adjacent segments influence local crash risk more than local driving behavior.

Methodology

dataset

Sample size: 513

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.

• incidence prevalence
• telematics crash prediction
• exposure measurement
• naturalistic crash near crash
• induced exposure
• sex gender

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: crash risk outcomes, observational prevalence
• Methodological Resource: dataset resource