How left-turning vehicles deal with conflicts at intersections: A driving behavior model based on relative motion risk quantification

Unprotected left turns at intersections in right-hand traffic are a critical factor affecting traffic safety. Traditional risk assessment indicators, which typically rely on vehicle relative positions, fall short in supporting yield/go decisions by left-turning drivers across different types of conflicts, and the corresponding driving behavior models struggle to capture the underlying behavioral mechanisms. To address these limitations, this paper introduces an improved risk assessment indicator based on risk field theory. By quantifying the relative motion risk between interactive vehicles, the proposed indicator offers a unified standard for intuitively determining whether a conflict has been resolved. Building on this, a Perception-decision-action behavioral framework, grounded in the preview-follower theory and risk homeostasis theory, is employed to model decision-making behaviors. This behavioral mechanism-driven model is validated through numerical simulations of vehicle trajectories, achieving a 92.59 % accuracy rate in replicating the decision-making behavior of left-turning vehicles, comparable to the performance of previous data-driven classification models. Furthermore, several cases are analyzed and discussed under different risk preferences and preview times, demonstrating that the model has potential for personalized trajectory planning. Overall, this paper provides a valuable reference model for enhancing intersection safety and advancing trajectory planning in autonomous driving systems.