Risk Assessment of Roundabout Scenarios in Virtual Testing Based on an Improved Driving Safety Field

Area of Science:

• Autonomous Driving
• Intelligent Transportation Systems
• Vehicle Dynamics

Background:

• Scenario-based testing is crucial for autonomous vehicles.
• Traditional risk indicators are inadequate for complex roundabout scenarios.
• Accurate risk quantification is needed for safe autonomous navigation.

Purpose of the Study:

• To develop an improved driving safety field model for quantifying risks in roundabout scenarios.
• To enhance the accuracy and efficiency of dangerous scenario detection in autonomous driving.
• To provide guidance for virtual testing and improve autonomous vehicle safety.

Main Methods:

• Improved driving safety field model incorporating vehicle dynamics during merging/diverging.
• Model parameter calibration using the social force model with the rounD dataset.
• DENCLUE-like clustering for collision system extraction and risk degree calculation.

Main Results:

• The proposed model accurately quantifies driving risks in roundabout scenarios.
• Detection efficiency for dangerous scenarios increased by 175% compared to Time to Collision (TTC).
• Temporal and spatial risk analysis provides valuable insights for virtual testing.

Conclusions:

• The improved driving safety field model effectively addresses limitations of traditional methods in roundabouts.
• The method enhances the generation of critical scenarios for robust autonomous vehicle testing.
• This contributes to advancing the safety and reliability of autonomous driving systems.