AI-Based Prediction of Motion Sickness in Autonomous Vehicles Using Passenger Profiles and Driving Route Data

Autonomous Vehicles (AV) are expected to significantly reshape travel experiences. Drivers will no longer be responsible for operating the vehicle and will instead take on the role of occupants, free to engage in personal activities such as reading or using phones. Consequently, motion sickness arises from a mismatch between the motion perceived by the body and the motion expected by the brain. Motion sickness detection in AVs has typically been approached through sensor-driven methods that monitor vehicle dynamics as well as passenger conditions, requiring continuous streams of data for reliable predictions. However, while vehicle-mounted sensors can be seamlessly integrated into the system, passenger-oriented solutions often rely on wearable devices, which may not always be practical for wide use in everyday settings. To address this gap, we introduce a ProfileFusion model, a multimodal deep learning model providing a more user-friendly solution that requires minimal user input. It operates effectively using pre-planned driving routes obtained from GPS combined with insights from user profile information such as demographic data. Due to the absence of standardized benchmark datasets for motion sickness detection, the proposed method is evaluated using data collected in a designed study during an industrial research project. Our approach enhances predictive accuracy on four diverse driving scenarios. When a high likelihood of sickness is detected, the model can alert passengers so they can change their behavior, such as looking outside in direction of travel or taking short breaks from the screen. ProfileFusion contributes to more comfortable travel in AVs by providing timely warnings.