Modeling vehicle behavior in two-dimensional driving scenarios considering collision risk stimulus and real-time interaction.

As the microscopic foundation of traffic dynamics, driving interactions and behaviors are critical for understanding the operational mechanisms of traffic systems and enabling safe, human-like autonomous driving. However, most existing physics-based driving behavior models are suitable for specific scenarios, and data-driven models also suffer from heavy data reliance and poor interpretability. These limitations restrict models' applicability across diverse two-dimensional scenarios. To address these limitations, this study proposes a two-stage Two-dimensional Desired Safety Margin (TDSM) model that unifies path estimation and velocity adjustment to model interactive driving behaviors in generalized two-dimensional scenarios based on the driver's risk perception quantification. This model could capture drivers' behaviors in yielding, preempting, and hesitation phases, then iteratively predict vehicle trajectories. Using real-world vehicle interaction trajectories, a generalized set of model parameters is calibrated, enabling the simulation of diverse situations: straight/turning maneuvers, preemption/yielding interactions, and vehicle-to-vehicle/multi-vehicle dynamics. Results demonstrate that TDSM achieves higher generalizability in describing interactive driving behaviors, with accuracy consistently matching or surpassing existing scenario-specific driving behavior models. Validations across external scenarios (car-following and crossing situations) and drivers' internal factors (varying driving styles) further demonstrate its robustness. The proposed model provides a low-complexity, high-accuracy baseline for applications in driving safety analysis, microscopic traffic simulation, human-like autonomous driving control, and scenario generation for automated driving testing, offering the potential for traffic accident prevention and autonomous driving safety.