Real-time combined safety-mobility assessment using self-driving vehicles collected data.

Area of Science:

• Transportation Engineering
• Traffic Safety Analysis
• Autonomous Vehicle Technology

Background:

• Assessing real-time safety and mobility for autonomous vehicles (AVs) is crucial for their integration into existing traffic systems.
• Traditional safety models struggle with limited and unevenly distributed conflict data, and unobserved spatial heterogeneity.
• Existing mobility assessment methods may not fully capture the dynamic nature of traffic conditions influenced by AVs.

Purpose of the Study:

• To develop and validate a real-time safety and mobility assessment approach for autonomous vehicles.
• To quantify real-time crash risk (RC) and return level (RL) using Bayesian hierarchical spatial random parameter extreme value models (BHSRP).
• To develop a Risk Exposure (RE) index and assess corridor mobility using Highway Capacity Manual (HCM) methodology.

Main Methods:

• Utilized a 440-hour dataset of autonomous vehicle (AV) data from a corridor in Palo Alto, California.
• Applied Bayesian hierarchical spatial random parameter extreme value model (BHSRP) for safety assessment, incorporating Time-To-Collision (TTC) as a conflict indicator.
• Assessed mobility using Highway Capacity Manual (HCM) methodology with real-time traffic data and developed a Risk Exposure (RE) index.

Main Results:

• Level of Service (LOS) E was identified as the most hazardous operating condition with the highest average crash risk after normalization.
• The Risk Exposure (RE) index demonstrated that accounting for vehicle exposure time is critical for accurate safety assessment.
• Autonomous vehicles face the highest chance of encountering extremely risky driving conditions at intersections (LOS D) and segments (LOS E) when exposure time is considered.

Conclusions:

• The proposed BHSRP model effectively handles data limitations and spatial heterogeneity for real-time safety assessment of AVs.
• Real-time safety and mobility assessments are essential for understanding and managing the impact of AVs on road networks.
• A nuanced understanding of risk, considering both operating conditions and exposure time, is necessary for safe AV deployment, particularly at intersections and road segments under specific LOS conditions.