Understanding driver cognition and decision-making behaviors in high-risk scenarios: A drift diffusion perspective

Area of Science:

• Autonomous Systems
• Human-Computer Interaction
• Traffic Safety

Background:

• Mixed traffic systems with autonomous vehicles (AVs) and human drivers present safety challenges, especially in complex, high-risk scenarios.
• Understanding and modeling human driver behavior, including risk cognition and decision-making, is crucial for safe AV integration.

Purpose of the Study:

• To develop a cognition-decision framework integrating individual driver variability and commonalities to model risk cognition and dynamic decision-making.
• To enhance the safety of human-AV interactions in mixed traffic environments.

Main Methods:

• Developed a risk sensitivity model using a multivariate Gaussian distribution to characterize individual risk cognition.
• Introduced a cognitive decision-making model based on the drift diffusion model (DDM) to capture common decision-making mechanisms.
• Simulated high-risk scenarios (lateral, longitudinal, multidimensional) in a driving simulator to test the model's predictive accuracy.

Main Results:

• The proposed DDM framework accurately predicts cognitive responses and decision behaviors during emergency maneuvers.
• Incorporating driver-specific risk sensitivity allowed for dynamic adjustments of DDM parameters, enabling personalized decision-making representations.
• Comparative analysis showed DDM more precisely captures human cognitive processes and adaptive decision-making than existing models (IDM, Gipps, MOBIL).

Conclusions:

• The DDM-based framework provides a robust method for modeling human driving behavior in high-risk scenarios.
• Findings offer critical insights for developing safer AVs and improving AV-human interaction in real-world traffic.
• This research lays a theoretical foundation for personalized AV decision-making algorithms.