Modeling the Driver's Lane-Changing Decision Under the Influence of the Blind Zone Image of an Intelligent Vehicle: A Practice Based on Cognitive-Driven Large Language Model

Lane-change blind zone image in intelligent vehicles can significantly improve driving safety, reducing the accident rate by 14%. However, inappropriate blind zone image design can lead to driver distraction and cognitive overload. This paper establishes a cognitive theory-driven large language model to address the interaction between the driver, blind zone image, and the road environment. The model is based on cognitive frame-works, integrating the powerful reasoning capabilities of large language models and a physiologically-based working memory module. This approach better simulates the cognitive decision-making processes in complex driver-machine-environment inter-actions, with improved interpretability and accuracy. The results show that the model achieves prediction accuracies of 96.59 % for driver decisions and 94.13 % for gaze points, effectively replicating driver cognitive decision-making. Additionally, the model can efficiently evaluate and optimize blind zone image design, revealing that improper lane-change blind zone image design significantly increases driver cognitive load and safety risks. Furthermore, it aids in optimizing HMI design to reduce cognitive load and enhance safety. This study lays the foundation for designing safer and more user-friendly intelligent vehicle HMI systems, promoting high-quality development in the automotive industry.