Automated vehicle collisions in California: Applying Bayesian latent class model
DOI: 10.1016/j.iatssr.2020.03.001
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 investigates the safety outcomes and collision patterns of automated vehicles (AVs) in California to inform regulatory strategies and improve AV safety. Motivated by the rapid advancement of AV technology and the California Department of Motor Vehicles’ (CA DMV) mandate for public reporting of AV collisions, the authors aim to uncover heterogeneity effects in roadway features, human interactions, and other attributes associated with these incidents. The research addresses the critical need for policymakers to understand specific safety concerns beyond general disengagement trends. The researchers analyzed 151 AV collision reports filed by manufacturers between September 2014 and May 2019. They employed a Bayesian latent class model (LCM) using variational inference to cluster the collision data into distinct patterns, allowing for the identification of hidden trends within the heterogeneous dataset. Additionally, the study utilized exploratory data analysis to examine spatial distributions, temporal frequencies, and variable characteristics, alongside text mining of police collision narratives to identify significant linguistic patterns associated with different collision classes. The analysis revealed six distinct classes of collision patterns. Class 6 was the largest cluster, characterized by conventional two-vehicle collisions occurring while the AV was stopped, with minor vehicle damage and cloudy or foggy weather. Classes associated with turning maneuvers, multi-vehicle collisions, dark lighting conditions with streetlights, and sideswipe or rear-end collisions were linked to higher proportions of injury severity. Specifically, Class 3 involved severe operator injuries and left/right turns, while Class 5 showed the highest percentages of multi-vehicle collisions occurring while moving, with operator severity more than twice as high as the no-severity condition. Temporally, collisions peaked in late 2018, particularly in November, and were most frequent on Thursday and Friday afternoons. Spatially, collisions were concentrated in the Mountain View and northeastern San Francisco areas. The findings provide a detailed characterization of AV collision dynamics, highlighting that specific scenarios, such as turning and multi-vehicle interactions in low-light conditions, pose greater risks for injury severity. By identifying these latent structures, the study offers critical insights for developing targeted safety strategies and regulations. The results contribute to the broader understanding of AV safety outcomes, suggesting that while AVs may reduce human-error-induced accidents, specific operational contexts require enhanced monitoring and regulatory attention to mitigate risks effectively.
Key finding
Bayesian latent class modeling of 151 automated vehicle collisions in California identified six distinct collision patterns, with classes involving turning, multi-vehicle interactions, and dark lighting conditions showing higher proportions of severe injuries.
Methodology
dataset
Sample size: 151
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 author_sweep_intake on 2026-05-28.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | author_sweep | — | — | 2 | 2026-05-28 |
| archive | success | canonical_url | — | — | 11 | 2026-06-06 |
| extract | success | cached | — | — | 3 | 2026-06-10 |
| clean | success | clean | — | — | 1 | 2026-06-04 |
| chunk | success | chunk | — | — | 1 | 2026-06-04 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-04 |
| enrich | success | semantic_scholar | — | — | 2 | 2026-06-04 |
| promote | success | — | — | — | 1 | 2026-06-04 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 2 | 2026-06-10 |
| tag | success | vector_similarity | — | — | 15 | 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.
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