A pressure-based intelligent seat systems: A novel approach to driver behavior recognition

As intelligent vehicles become increasingly prevalent, ensuring their operational safety has emerged as a paramount challenge. Vehicles equipped with advanced intelligent systems offer the potential to enhance safety by continuously monitoring and analyzing the driver’s condition and behavior in real time, thereby aligning vehicle control more closely with the driver’s intentions. While visual detection methods within the cockpit are often impeded by obstructions such as other passengers or interior elements, the driver’s seat provides a unique advantage due to its constant, unobstructed contact with the driver throughout the driving process. This positioning enables the seat to detect subtle changes in its morphology, offering a reliable means to map the driver’s state with minimal interference. This paper introduces a novel pressure-based intelligent seat system integrated with an array of polyvinylidene fluoride (PVDF) sensors and proposes a tailored Pressure Mat Vision Transformer (ViT) model to accurately classify the driver’s operational behaviors. The system demonstrates the capability to predict the precise depth of throttle, brake, and clutch pedal engagement, as well as the steering wheel’s control state. Validation of the model’s performance was conducted using a custom-built dataset. Furthermore, the feasibility of the pressure recognition technology and the efficacy of the fine-tuned architecture for real-time application were substantiated through engineering deployment on an operational platform. The development of this intelligent seat represents a significant advancement in addressing safety concerns related to human–machine interactions and establishes an innovative framework for leveraging data in the evolution of next-generation intelligent vehicles.