Composite safety potential field for highway driving risk assessment

Area of Science:

• Vehicle Safety
• Human-Computer Interaction
• Transportation Engineering

Background:

• Driving risk assessment is crucial for advanced vehicle safety systems like collision warning and autonomous driving.
• Existing safety potential field models face limitations in calibration and generalization across longitudinal and lateral risks.
• Real-time, human-like risk assessment is needed for seamless vehicle-to-vehicle interaction.

Purpose of the Study:

• To propose a novel composite safety potential field (C-SPF) framework for comprehensive driving risk evaluation.
• To address the limitations of existing models in calibration and capturing both longitudinal and lateral risks.
• To develop a risk assessment metric that reflects human-like reasoning for improved vehicle safety.

Main Methods:

• Developed a C-SPF framework combining a subjective field for risk perception and an objective field for collision probability.
• Calibrated the C-SPF using extensive two-dimensional spacing data from vehicle trajectory datasets.
• Analyzed a naturalistic driving dataset to validate the framework's ability to capture safety maneuvers.

Main Results:

• The C-SPF effectively captures drivers' proximity risk perception and provides realistic explanations for driving behaviors.
• Demonstrated the framework's capability to identify both longitudinal and lateral risks triggering safety maneuvers.
• Case studies confirmed the C-SPF's effectiveness in explaining complex lateral behaviors like lane change adjustments.

Conclusions:

• The proposed C-SPF framework offers a more realistic and comprehensive approach to driving risk assessment.
• C-SPF overcomes the calibration and generalization limitations of previous safety potential field models.
• This advancement is vital for the development of safer and more interactive autonomous and semi-autonomous vehicles.