Perturbed Utility Based Flow Estimation from Link-Based Person-Time Measures

Abstract

Indoor pedestrian route choice analysis in large three-dimensional spaces is difficult because WiFi-based observations are noisy and path-level map matching is often unreliable. This paper proposes a unified framework based on a probabilistic link-based person-time measure derived from uncertain Wi-Fi observations. The model jointly estimates pseudo-OD-specific directed flow and perturbed-utility-based behavioral parameters, while LLM-derived semantic attributes are included as contextual covariates. Using the Shibuya Station pedestrian network, the framework closely reproduces the observed person-time field and yields stable, interpretable route choice parameters, indicating that person-time-based flow estimation provides a practical basis for indoor pedestrian behavior analysis under observational uncertainty.