Learning to uncover human discomfort from vortex-induced bridge vibrations through immersive experiments

Vortex-induced vibrations of long-span bridges are increasingly triggering pronounced user discomfort and public concern. Here, we introduce an immersive motion-visual platform that synchronizes a vertically actuated motion base with a virtual-reality bridge-driving scene, enabling controlled human-subject experiments under vibration conditions that are impractical and unethical to study in the field. Using time-resolved discomfort reports across a broad range of vibration amplitudes, frequencies, and traveling scenarios, we demonstrate that the ISO-style dose-based Motion Sickness Index systematically fails to reproduce the observed evolution of discomfort for moving users, as it cannot represent relief and recovery during node-antinode transitions. We therefore propose a time-weighted motion-sickness formulation with a learned temporal “memory” kernel identified by symbolic regression and linked to discomfort via probabilistic regression. The resulting index is validated to generalize well across scenarios and enhance the prediction of discomfort for real-world bridge VIV events, supporting refined serviceability assessment and operational decision-making.