Investigating the Interrelationships among Factors Associated with Automated Vehicle Crashes Using Analytic Hierarchy Process

Huang, Chunxi; Wang, Ange; Yan, Song; He, Dengbo · 2025 · Transportation Research Record

DOI: 10.1177/03611981241274152

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 interrelationships among factors associated with automated vehicle (AV) crashes to improve safety in mixed traffic environments. While AVs are expected to reduce human-error-related accidents, crashes still occur due to complex interactions between AVs and human-driven vehicles. Previous research relied on traditional regression models that failed to capture the topological dependencies among crash factors and utilized outdated data from the California Department of Motor Vehicles. To address these limitations, this study analyzes recent crash data to reveal up-to-date high-risk scenarios and the structural relationships between contributing factors, aiming to inform targeted interventions and technology advancements. The researchers analyzed 246 AV-involved crashes reported to the National Highway Traffic Safety Administration (NHTSA) between July 2021 and April 2023. The dataset included variables categorized into environmental factors (e.g., roadway type, weather), crash-procedure-related factors (e.g., crash type, pre-crash movements), and crash outcomes (e.g., injury severity). The study employed an Additive Bayesian Network (ABN) approach to construct a directed acyclic graph representing the topological relationships among these variables, overcoming the information loss associated with traditional Bayesian networks. This was followed by post-ABN regression analyses to quantify the relationships and account for interaction effects. The ABN modeling involved structure discovery, parameter learning, and validation through parametric bootstrapping to ensure robustness against overfitting. The results indicate that rear-end crashes remain the dominant type of AV-involved crash, likely due to discrepancies in driving behaviors between AVs and human-driven vehicles. The ABN model revealed that crash types are more strongly associated with the pre-crash movements of crash partners than with the pre-crash movements of the AVs themselves. Specifically, the pre-crash movement of the crash partner accounted for the most uncertainty in determining crash type. Furthermore, crash outcomes, such as injury severity, were found to be associated with environmental factors like the operating entity and crash-procedure factors like crash type. Regression analyses confirmed that AVs operated by Waymo were significantly more likely to be involved in crashes at intersections, and higher speed gap ratios were associated with AVs being stopped prior to the incident. The significance of this study lies in its use of a novel graphical network-based mixed approach to uncover the complex interdependencies among AV crash factors, which traditional statistical methods overlook. By identifying that crash partners' behaviors and specific environmental contexts drive crash types and outcomes, the findings provide actionable insights for refining AV control algorithms and designing mixed traffic regulations. The study highlights the need for targeted interventions, such as improving AV responses to intersection scenarios and addressing behavioral mismatches with human drivers, to enhance overall traffic safety as AV deployment expands.

Key finding

Rear-end crashes dominate automated vehicle incidents and are primarily driven by the pre-crash movements of crash partners, whereas injury severity is linked to environmental conditions and crash types.

Methodology

dataset

Sample size: 246

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 2026-06-10; verification: verified.

Topics

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

• pre crash contributing factors
• crash typology
• causation analyses
• naturalistic crash near crash
• motorcycle crash typology
• incidence prevalence

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