Automated Vehicle Crash Rate Comparison Using Naturalistic Data
DOI: 10.13140/rg.2.1.2336.1048
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This study addresses the challenge of comparing crash rates between automated vehicles and human-driven vehicles by utilizing naturalistic driving data. The research is motivated by the need to establish a baseline for safety performance, specifically investigating how many crashes go unreported to police or insurance and whether these unreported rates vary by location. The authors aim to determine if automated vehicles, particularly those in autonomous mode, exhibit lower crash rates than human drivers across different levels of automation (Level 1, Level 2, and Level 3). The methodology involves analyzing data from the Strategic Highway Research Program 2 (SHRP 2) Naturalistic Driving Study, which provides a robust dataset of human driving behavior. The researchers calculated crash rates per million miles driven for various demographic groups, adjusting for age to ensure fair comparisons. They compared these human baseline rates against reported crash data for self-driving cars. The analysis segmented data by multiple factors, including driver age groups (e.g., ≤25, 26-35, >65), speed zones, locality types (such as urban, interstate, and residential areas), traffic density levels of service (LOS A through F), and specific crash characteristics like precipitating events, driver behavior, and fault attribution. Statistical formulas were applied to weight crash rates based on the distribution of the U.S. licensed driver population versus the SHRP 2 participant sample to derive age-adjusted metrics. The findings indicate that the crash rate for self-driving cars in autonomous mode is 5.6 crashes per million miles. In comparison, the age-adjusted crash rates for human drivers in the SHRP 2 dataset varied by automation level context, with specific values noted at 3.3, 2.5, and 1.6 crashes per million miles for different comparative categories. The study highlights that crash rates fluctuate significantly based on environmental and behavioral factors. For instance, crash rates were analyzed across different localities, showing variations between urban centers and interstates. Additionally, the data examined the prevalence of driver impairment, distraction, and illegal maneuvers across different automation levels, providing a detailed profile of human error versus automated system performance. The sample sizes for various age groups were explicitly detailed, ranging from hundreds to over a thousand participants, ensuring statistical relevance. The significance of this research lies in providing a data-driven framework for evaluating the safety of automated vehicles against human baselines. By using naturalistic data and applying age-adjustments, the study offers a more accurate comparison than raw police-reported statistics, which often miss minor incidents. The results suggest that while automated vehicles have measurable crash rates, understanding the context—such as location, speed, and driver demographics—is crucial for interpreting safety performance. This work contributes to the broader field of autonomous vehicle safety by establishing methodological standards for crash rate comparison and highlighting the importance of unreported crash data in safety assessments.
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.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | OpenAlex-citations | — | — | 1 | 2026-06-25 |
| archive | success | openalex | — | — | 5 | 2026-06-26 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-25 |
| chunk | success | chunk | — | — | 1 | 2026-06-25 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-25 |
| promote | success | — | — | — | 1 | 2026-06-25 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-25 |
| verify | partial | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified_with_issues.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.
- naturalistic crash near crash
- incidence prevalence
- sex gender
- induced exposure
- crash typology
- causation analyses
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