PATH-U: A data-driven agent-based wayfinding model incorporating perceived path uncertainty and cognitive strategies in unfamiliar indoor environments

As built environments become more complex, indoor wayfinding challenges increase, especially for first-time visitors. Effective wayfinding design and signage are crucial for helping people reach their destinations. Occupant simulations can analyze these features before construction and identify potential issues. However, current models for human wayfinding in unfamiliar environments are limited and rarely predict continuous experiences like perceived path uncertainty. This study developed an integrated agent-based model called “PATH-U”, which simulates multi-floor wayfinding tasks without prior knowledge of the environment and provides feedback on uncertainty levels. This model is based on an observational study with 39 participants completing 273 wayfinding tasks in a complex university building. We developed a path-planning model incorporating visual perception, natural movements, short-term memory, heuristic strategies, and a data-driven multiple linear regression model for uncertainty prediction based on data from 28 participants. Validation with data from 11 participants under a different signage condition shows that the model mostly mirrors human wayfinding behavior and perceived uncertainty, with a few notable discrepancies. The findings suggest that wayfinding design should consider spatial dimensions, confirmational signage, and enhanced cues at crucial intersections to reduce uncertainty and improve performance. Future simulations should incorporate on-route behaviors and environmental reasoning.