Mitigating sensor bias: a location-aware calibration framework for pedestrian analytics

Abstract

Abstract The promise of data-driven urbanism is critically undermined by the unreliability of passive sensors. Raw WiFi data systematically undercounts pedestrian volumes by approximately 50% across the eight study sites, rendering it fundamentally unsuitable for strategic decision-making. We present and validate a location-aware calibration framework that treats the scaling factor as an empirically estimated, site-specific coefficient capturing a socio-technical signature of place, shaped by urban morphology, human behaviour, and signal conditions. Derived and validated across eight diverse urban sites in Loughborough Town Centre, UK against multi-day ground-truth observations, our approach elevates data reliability from its unusable baseline to a robust 85.8% accuracy. The application of this calibrated data generates high-resolution pedestrian flow maps. These maps, in turn, provide a closer approximation of the city’s circulatory system by correcting distorted spatial narratives and revealing a functional hierarchy not evident in the raw data. Systematic anomalies in urban sensor records, including undercounting in social spaces and overcounting in congested areas, represent data signatures that reveal user–place interactions and inform evaluation of placemaking quality and infrastructure pressure. This research offers a scalable blueprint for converting low-cost sensor data into reliable intelligence, empowering evidence-based urban planning, design, and mobility management.

The authors' abstract, as published at the source. Computational Urban Science, 2026 · DOI ↗