Autonomous Vehicle Collision Reports from California DMV, 776 Incidents

776 publicly available autonomous vehicle collision reports from the California Department of Motor Vehicles, analyzed by Liu Yang. The dataset, last updated in April 2026, categorizes risk factors into vehicle information, collision details, and road/environmental characteristics. It employs statistical and Bayesian network analysis to explore causal chains and collision severity.

Use Cases

• Modeling causal relationships between risk factors and collision severity based on the Bayesian network analysis described.
• Assessing the significance of single factors like vehicle mode or brand on collision outcomes using the Chi-square test methodology mentioned.
• Identifying high-risk scenarios for property damage or casualties based on described factors like collision type and roadway type.
• Comparing collision risks between autonomous and conventional human driving modes as referenced in the study.

Strengths

• Contains 776 real-world collision reports from an official source (California DMV).
• Analysis identifies six specific variables that independently affect collision severity.
• Employs both statistical (Chi-square) and probabilistic (Bayesian network) analytical methods.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown for the underlying data files, which may limit suitability assessment.
• Data may reflect geographic bias inherent to its source (California only).

Provenance

Source

California Department of Motor Vehicles

Collection Method

Analysis of publicly available collision reports.

Time Range

null

Freshness

Last updated 2026-04-27 16:39:43; freshness should be verified.

Geography

California, USA