Traffic safety measures using multiple streams real time data : final report
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 report presents a methodology for estimating composite traffic safety risk measures that vary temporally and spatially, leveraging multiple streams of real-time data. The research addresses the complexity of traffic crashes, which result from conflicts influenced by roadway conditions, traffic signals, weather, driver behavior, and vehicle health. Motivated by advances in connected vehicle technology and the Internet of Things, the project aims to develop advanced analytics that generate dynamic safety risk profiles for individual drivers. These profiles are intended to support Advanced Driver Assistance Systems (ADAS) by providing proactive, customizable alerts based on real-time risk assessments. The study utilized data from the Strategic Highway Research Program 2 (SHRP 2) Naturalistic Driving Study (NDS), combined with Straight Line Diagram (SLD) data and historical crash records. Due to funding constraints and confidentiality restrictions regarding video data, the researchers employed a filtering mechanism to select relevant subsets of the dataset. The available data included baseline, near-miss, and crash events for random drivers, featuring time-series kinematic and statistical data covering 10–15 seconds prior to and 10 seconds after an event. The methodology involved feature engineering to prepare the data and the development of a predictive model using elastic-net regularized multinomial logistic regression. This model estimates the probability of three driving states: crash, near-miss, and normal driving. The system architecture supports both offline cloud-based processing and an on-board smartphone application that calculates risk scores at defined intervals (e.g., every 10–15 seconds) to create a driver’s risk profile over a trip. The findings demonstrate a novel data-driven approach to real-time traffic safety risk prediction, quantifying risk as the likelihood of adverse outcomes. The researchers introduced five measures of goodness to evaluate model performance, focusing on sensitivity and specificity for minority classes (crashes and near-misses). Using 10-fold cross-validation on a subset of SHRP 2 data, the study confirmed the applicability of the elastic-net model, noting it as the first attempt to apply this specific regularization technique to traffic accident research. The models were designed to prioritize lower Type II errors (false negatives) over Type I errors (false positives), given the higher cost of missing a potential crash. The results indicate that the model can accommodate different resolutions of driving outcomes and that performance could improve with larger sample sizes of crash and near-crash events or through resampling methods like bootstrapping. The significance of this work lies in its potential to enhance ADAS performance for mixed traffic environments containing both autonomous and human-operated vehicles. The developed predictive models can support collision warning systems, driver safety risk profiling, and dynamic hotspot analysis for roadway segments. The study concludes that while the current model provides a robust foundation for real-time risk prediction, future work should focus on expanding data collection, exploring deep learning algorithms, and refining the human-machine interface for practical implementation. The research also highlights the importance of customizing alerts based on individual driver characteristics and surrounding circumstances to maximize safety effectiveness.
Key finding
The study developed an elastic net regularized multinomial logistic regression model that successfully predicts traffic safety risk by classifying driving events into crash, near-miss, and normal categories using SHRP 2 naturalistic driving data.
Methodology
dataset
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).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | rosap | — | — | 2 | 2026-05-23 |
| archive | success | — | — | — | 1 | 2026-05-23 |
| extract | success | cached | — | — | 2 | 2026-06-10 |
| clean | success | — | — | — | 1 | 2026-06-01 |
| chunk | success | — | — | — | 1 | 2026-06-01 |
| embed | success | — | — | — | 1 | 2026-06-02 |
| enrich | success | — | — | — | 1 | 2026-05-23 |
| promote | success | — | — | — | 1 | 2026-05-23 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 3 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 19 | 2026-06-11 |
| verify | partial | — | — | — | 2 | 2026-06-10 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- telematics crash prediction
- induced exposure
- naturalistic crash near crash
- exposure measurement
- adas effectiveness
- anticipation
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
- Methodological Resource: dataset resource
- Theoretical Contribution: computational model