Classifying crash causation patterns in 2-vehicle collisions between autonomous and conventional vehicles

Area of Science:

• Road safety research
• Autonomous vehicle technology
• Human-computer interaction in transportation

Background:

• Existing research on autonomous vehicle (AV) crashes predominantly focuses on AV system failures.
• There is a gap in understanding the causal factors of collisions between AVs and conventional vehicles (CVs), particularly concerning their interactions.

Purpose of the Study:

• To develop a classification framework for crash causation patterns in AV-CV collisions.
• To analyze the interactions between AVs and CVs leading to two-vehicle collisions.
• To identify the primary causes and contributing factors in AV-CV crashes.

Main Methods:

• Developed a 5-type classification framework for crash causation patterns.
• Analyzed 450 two-vehicle collisions involving AVs and CVs using a dedicated AV crash dataset.
• Categorized crashes based on AV system failure, takeover control issues, driver error post-takeover, CV adaptation failure, and other CV-related factors.

Main Results:

• The majority of AV-CV collisions (55%) are attributed to conventional vehicles failing to adapt to autonomous vehicle behavior (Pattern 4).
• Improper response of CVs to AVs stopping is a leading cause of these collisions.
• Crashes involving human driver errors (Patterns 3, 4, 5) result in more severe injuries, with Pattern 4 frequently leading to rear-end collisions.

Conclusions:

• Conventional vehicle behavior is a significant factor in AV-CV collisions.
• Recommendations include minimizing AV planning errors, enhancing AV-CV communication, and implementing specialized driver training for mixed traffic environments.
• Improving the adaptability and response of conventional vehicle drivers to AVs is essential for future road safety.