Pattern recognition in crash clusters involving vehicles with advanced driving technologies

Area of Science:

• Road safety
• Automotive engineering
• Human-computer interaction

Background:

• Autonomous Vehicle (AV) technologies, including Advanced Driver Assistance Systems (ADAS) and Automated Driving Systems (ADS), offer potential safety benefits but are involved in increasing numbers of crashes.
• Existing research on AV crash factors has not fully elucidated critical risk patterns within ADAS- and ADS-specific crash clusters.
• Understanding these patterns is crucial for targeted safety improvements and policy development.

Purpose of the Study:

• To identify and analyze critical risk factor patterns in clusters of ADAS- and ADS-engaged AV crashes.
• To compare and contrast crash characteristics between ADAS- and ADS-engaged scenarios.
• To provide data-driven recommendations for enhancing AV safety and informing policy.

Main Methods:

• Utilized the Cluster Correspondence Analysis tool to cluster crash-related factors for ADAS and ADS engaged AVs.
• Analyzed crash data to identify distinct clusters and their associated characteristics.
• Compared crash outcomes, road conditions, and crash partners between ADAS and ADS engagement.

Main Results:

• Identified three distinct crash clusters for both ADAS and ADS engagement.
• ADAS-engaged crashes frequently involve fatal intersection collisions, especially with left-turning vehicles.
• ADS-engaged crashes commonly involve non-motorists in daylight; both systems show risks with animals on high-speed roads.

Conclusions:

• Infrastructural improvements and enhanced AV algorithms/sensors (especially for non-motorist and animal detection in low light) are recommended.
• Policymakers should focus on driver education for safe AV operation and mandate external human-machine interfaces to improve communication and reduce rear-end collisions.