Descriptive analysis of reports on autonomous vehicle collisions in California: January 2021–June 2022

Pokorny, Petr; Høye, Alena · 2022 · Crossref

DOI: 10.55329/xydm4000

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 provides a descriptive analysis of autonomous vehicle (AV) collisions reported to the California Department of Motor Vehicles (DMV) between January 2021 and June 2022. Motivated by the need to understand safety trends as AV technology advances and regulatory frameworks evolve, the authors examined whether recent data revealed new collision patterns compared to previous years. The research specifically compared collisions occurring while the AV was in autonomous mode, manual mode, or immediately following a disengagement from autonomous control. The methodology utilized publicly available DMV collision reports, analyzing 208 incidents (117 in 2021 and 91 in the first half of 2022). The authors employed descriptive statistics and calculated odds ratios with 95% confidence intervals to compare collision characteristics across different driving modes and time periods. Variables analyzed included collision type, location, severity, involved road users, and fault attribution. The findings indicate that AV collisions remained predominantly low-severity, with 88% involving no injuries and 85% resulting in minor or no property damage. Driving in autonomous mode was significantly associated with fewer instances of the AV being deemed at-fault; the odds of fault were reduced by 78% compared to other modes. Most collisions in autonomous mode were rear-end incidents at intersections, often caused by following vehicles striking the AV after it braked abruptly for a traffic signal. Conversely, single-vehicle collisions and collisions with vulnerable road users (VRUs) such as cyclists and pedestrians occurred primarily in manual mode or immediately after disengagement. Notably, the share of collisions involving VRUs doubled from 7% in 2021 to 13% in 2022, and the proportion of collisions occurring in autonomous mode increased from 38% to 53%. The study concludes that while AVs in autonomous mode are less likely to be at fault, their behavior—particularly abrupt braking—makes them unpredictable to human drivers, leading to frequent rear-end collisions. The increasing involvement of VRUs highlights specific detection and prediction challenges for AV systems. The authors suggest that as AVs operate more frequently in autonomous mode, the rising share of VRU collisions requires heightened attention due to the unpredictability of these road users. The results confirm historical trends regarding collision types and severity while highlighting emerging risks associated with increased AV deployment and VRU interactions.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• vru crash typology