Development and classification of autonomous vehicle's ambiguous driving scenario

Area of Science:

• Autonomous Driving Systems
• Human-Computer Interaction
• Road Safety

Background:

• The increasing integration of autonomous vehicles (AVs) necessitates a clear understanding of their responses to complex driving situations.
• Ambiguous Driving Scenarios (ADS) pose significant challenges for AV safety, legal frameworks, and user acceptance.
• Current AV development requires a user-centric approach to address ethical and practical considerations.

Purpose of the Study:

• To develop a robust framework for classifying Ambiguous Driving Scenarios (ADS) in autonomous vehicles.
• To analyze user perceptions of various ADS categories, focusing on safety, moral, ethical, legal, and utility aspects.
• To inform policy and algorithm development for safer and more convenient autonomous driving.

Main Methods:

• Extensive literature review and in-depth interviews with industry experts.
• Comprehensive questionnaire survey administered to 548 autonomous vehicle users.
• Factor analysis applied to classify 28 diverse ambiguous driving scenarios based on user feedback.

Main Results:

• Ambiguous Driving Scenarios (ADS) were successfully grouped based on user perceptions, with all categories demonstrating high safety ratings.
• Scenarios were classified as: moral (AV yields), ethical (bottleneck), legal (cross-over), and utility-related (AV halts).
• User perceptions provided a clear basis for categorizing complex driving situations encountered by autonomous vehicles.

Conclusions:

• The study provides a novel classification framework for ADS grounded in user perceptions, crucial for AV development.
• Findings offer valuable insights for refining AV algorithms and informing regulatory policies to enhance safety and user trust.
• This research contributes to the advancement of self-driving car technology by addressing critical human-centric aspects of autonomous driving.