Using Naturalistic Driving Data to Assess Vehicle-to-Vehicle Crashes Involving Fleet Drivers
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 limitations of police-reported crash data, which often lacks detailed information on driver behaviors and contributing factors in the seconds preceding a collision. Motivated by the high rate of transportation-related occupational fatalities and the difficulty of accurately identifying causes like distraction or fatigue through self-report or reconstruction, the researchers utilized naturalistic driving data to examine vehicle-to-vehicle crashes involving fleet drivers. The primary objective was to identify roadway conditions, critical pre-crash events, driver behaviors, and response times, specifically comparing rear-end and angle crashes. The researchers analyzed 247 moderate-to-severe vehicle-to-vehicle crashes from a database of 777 events captured by Lytx DriveCam in-vehicle event recorders (IVERs). These devices record 12 seconds of video, audio, and accelerometer data triggered by hard braking, fast cornering, or impacts exceeding 1g. After excluding minor impacts, rear-endings of the equipped vehicle, and non-fleet drivers, the remaining crashes were coded using a methodology developed from existing government and academic standards. Two independent analysts double-coded the six seconds leading up to each crash, focusing on 24 data elements including environmental conditions, driver errors, and specific behaviors. Statistical analyses, including chi-square and t-tests, were used to compare crash types and behaviors. The results indicated that 84% of crashes involved some driver contribution, with recognition errors (e.g., inadequate surveillance) present in 71% of cases and decision errors (e.g., following too closely) in 40%. Rear-end crashes were significantly associated with recognition errors and longer eyes-off-road times (average 3.2 seconds) compared to angle crashes (0.7 seconds). In rear-end crashes, 97% involved a vehicle ahead decelerating or stopping. Driver distraction was prevalent, with cell phone use occurring in 8.3% of crashes; notably, operating or looking at a phone occurred only when drivers were alone. Drivers alone were 3.5 times more likely to use cell phones and nearly twice as likely to engage in personal grooming or talking to themselves compared to those with passengers. Reaction times in rear-end crashes increased from 1.9 seconds with no observed behavior to 2.5 seconds with any behavior, and up to 3.3 seconds with cell phone use. Drowsy driving was rarely detected, appearing in only 1.8% of crashes, likely due to the low frame rate of the cameras. The study concludes that naturalistic driving data provides a more accurate and detailed view of crash causation than police reports, particularly regarding driver distraction and inattention. The findings highlight distinct behavioral profiles for rear-end versus angle crashes, suggesting that rear-end collisions are heavily driven by inattention and inadequate surveillance, while angle crashes involve more decision errors like failure to yield. These insights can inform the development of automotive safety systems and technologies tailored to mitigate specific crash types. The large sample size and objective measurement of behaviors offer significant value for traffic safety research, overcoming the recall bias and limited visibility inherent in previous studies.
Key finding
Among 247 coded moderate-to-severe fleet vehicle-to-vehicle crashes, drivers contributed in 84% of cases and recognition errors (including inadequate surveillance and potentially distracting behaviors) occurred in 71%; rear-end crashes showed about four times longer average eyes-off-road time (3.2 vs. 0.7 s) than angle crashes.
Methodology
naturalistic
Sample size: N=247 moderate-to-severe vehicle-to-vehicle fleet crashes (from 777 reviewed Lytx DriveCam videos)
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_aaa_foundation on 2026-05-23 (5 acquisition events logged).
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | aaa_foundation | — | — | 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 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 19 | 2026-06-11 |
| verify | success | — | — | — | 2 | 2026-06-10 |
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.
- naturalistic crash near crash
- pre crash contributing factors
- incidence prevalence
- crash typology
- driver post crash behavior
- crash reconstruction hf
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