Integrating Virtual Reality-based eye-tracking with urban digital twin for unveiling pedestrian visual attention in wayfinding tasks

Urban navigation relies heavily on signage, yet pedestrian visual interaction with these cues in complex environments is not yet fully quantified. This study developed an immersive Virtual Reality platform integrating eye-tracking with a detailed, photogrammetry-based 3D model that captures real-world geometric and textural features to investigate the gazing behavior and visual attention to support wayfinding signage design. A methodological framework is proposed to extract participants’ gazing behavior and estimate visual attention from eye-tracking data. Leveraging this integrated methodology, VR experiments were then designed to investigate pedestrian gaze behavior under varying signage information densities and different task conditions. An explainable machine learning approach combining XGBoost and SHAP was employed to model and interpret how geometric features like distance, incidence angle, pitch, and sign size influence specific gaze behaviors, distinguishing between focusing and scanning actions. Results revealed that high-information signage enhanced wayfinding efficiency but may increase cognitive load. For sign design, gaze distance and incidence angle were identified as key factors influencing attention, exhibiting non-linear effects and task-dependent thresholds, with distance below 30 m, incidence angle below 46°, and pitch angle below 4.2°associated with increased visual attention. Based on the attentional mechanisms observed, two recommended sign design strategies were proposed according to the task context. The study demonstrates the application of VR and digital twin technology for analyzing visual behavior, providing evidence-based insights for context-aware urban signage design.