Bicycle Naturalistic Data Collection

Elhenawy, Mohammed; Jahangirl, Arash; Rakha, Hesham A. · 2016 · ROSA P / Connected Vehicle/Infrastructure 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 safety issue of bicycle-motor vehicle crashes at intersections, which account for over 30% of cyclist fatalities in the United States. Motivated by a 47% increase in bicycle commuting rates between 2000 and 2011 and the high risk associated with cyclist violations, the research aims to identify factors influencing cyclist behavior and assess the feasibility of developing violation prediction models. The project seeks to integrate these models into connected vehicle infrastructure to mitigate crashes by predicting violations before they occur. The researchers conducted a naturalistic cycling experiment using data from 20 participants who rode instrumented bicycles in Blacksburg, Christiansburg, and Radford, Virginia. The bicycles were equipped with the MiniDAS system, featuring cameras, GPS, accelerometers, gyroscopes, and speed sensors. Participants rode their normal routes without special instructions to ensure realistic behavior. Data reduction identified 251 crossings at signal-controlled intersections and 2,024 crossings at stop-controlled intersections. The study utilized mixed-effects generalized regression models to identify significant factors affecting violation probabilities. Additionally, the authors tested four machine learning algorithms—multivariate logistic regression, random forest, K-nearest neighbors, and artificial neural networks—to develop binary classification models for predicting violations based on kinetic data measured as cyclists approached intersections. The analysis revealed distinct factors influencing violations at different intersection types. At signalized intersections, right turns and the presence of side or opposing traffic were statistically significant factors; specifically, right turns increased the likelihood of red light violations, while the presence of traffic decreased it. At stop-controlled intersections, significant factors included right turns, left turns, warm weather, and the presence of other road users. Demographic factors such as age and gender were not significant in either context. The violation prediction models demonstrated high accuracy, with error rates ranging from 0% to 10% depending on the prediction distance. Notably, an error rate of 6% was achieved when the cyclist was approximately two seconds from the intersection, a timeframe sufficient for motor vehicle drivers to respond to warnings. The findings confirm the feasibility of using naturalistic cycling data to predict intersection violations with high precision. The identified significant factors provide actionable insights for traffic safety interventions and infrastructure design. Furthermore, the low prediction error rates at a two-second time-to-intersection threshold suggest that these models can be effectively integrated into connected vehicle systems. This integration could enable real-time warnings to drivers or automated infrastructure adjustments, such as signal timing changes, to prevent bicycle-motor vehicle conflicts and enhance overall transportation safety.

Key finding

Violation prediction models achieved a 6% error rate at a 2-second time-to-intersection, providing sufficient warning time for drivers.

Methodology

naturalistic

Sample size: 20

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.

• cyclist safety

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, crash risk outcomes
• Methodological Resource: dataset resource