Wayfinding simulation based on virtual reality experiment data: Evaluating guide sign layout in underground space

Wayfinding difficulties, such as disorientation, remain persistent challenges in complex indoor and underground environments. Well-designed guide sign layouts offer an effective and cost-efficient solution. This study introduces a vision-driven pedestrian agent model, VD25, developed to quantitatively evaluate sign layouts in three-dimensional (3D) spaces. The model incorporates a refined sign detection mechanism applying visual catchment contour and simulates wayfinding behaviors on 3D walkable surface by switching between free-exploratory and sign-following movements. Validation using a multi-level metro station environment with a walking distance of approximately 100 m, shows strong correlations between sign read rates and those measured in virtual reality and real-space experiments (correlation coefficients of 0.899 and 0.814, respectively). The model also shows good agreement in walking distance and path selection. Moreover, simulations using alternative sign layouts and sensitivity analyses across diverse pedestrian populations highlight the VD25 model's potential to assist architects and facility managers in improve guide sign designs.