[Paper Review] AVOID: Autonomous Vehicle Operation Incident Dataset Across the Globe

Ou Zheng, Mohamed Abdel‐Aty+4 more|arXiv (Cornell University)|Mar 22, 2023

Traffic and Road Safety13 citations

TL;DR

This paper introduces AVOID, an open, manually verified autonomous vehicle crash dataset compiled from NHTSA, CA DMV, and global incident news, with added land use, weather, and geometry data.

ABSTRACT

Crash data of autonomous vehicles (AV) or vehicles equipped with advanced driver assistance systems (ADAS) are the key information to understand the crash nature and to enhance the automation systems. However, most of the existing crash data sources are either limited by the sample size or suffer from missing or unverified data. To contribute to the AV safety research community, we introduce AVOID: an open AV crash dataset. Three types of vehicles are considered: Advanced Driving System (ADS) vehicles, Advanced Driver Assistance Systems (ADAS) vehicles, and low-speed autonomous shuttles. The crash data are collected from the National Highway Traffic Safety Administration (NHTSA), California Department of Motor Vehicles (CA DMV) and incident news worldwide, and the data are manually verified and summarized in ready-to-use format. In addition, land use, weather, and geometry information are also provided. The dataset is expected to accelerate the research on AV crash analysis and potential risk identification by providing the research community with data of rich samples, diverse data sources, clear data structure, and high data quality.

Motivation & Objective

• Motivate the creation of a comprehensive AV crash dataset to advance safety research.
• Provide an open, high-quality resource that aggregates AV/ADAS crash data from multiple sources.
• Include contextual information (land use, weather, geometry) to support richer analyses.
• Ensure data is manually verified and formatted for ready-to-use research use.

Proposed method

• Data collection from multiple sources: NHTSA, California DMV, and worldwide incident news.
• Manual verification and summarization of crash data into a ready-to-use format.
• Augmentation with land use, weather, and geometric information.
• Clear data structure aimed at facilitating AV crash analysis and risk identification.

Experimental results

Research questions

• RQ1 What is the scope and structure of an open, globally sourced AV crash dataset?
• RQ2 How can diverse data sources be integrated and verified to support AV safety research?
• RQ3 What contextual features (land use, weather, geometry) enhance analysis of AV crashes?
• RQ4 How does AVOID improve data quality and usability for crash analysis and risk identification.

Key findings

• An open AV crash dataset (AVOID) is introduced, combining multiple data sources.
• Data are manually verified and summarized into a ready-to-use format.
• Contextual attributes such as land use, weather, and geometry are included.
• The dataset aims to accelerate AV crash analysis and risk identification with rich samples and high data quality.