Probabilistic surface correlation for robust indoor localization using RSSI vector sequence

Abstract

Accurate indoor localization remains a challenging task due to the inherently unstable nature of RSSI signals in complex environments. To address this challenge, we previously introduced the Surface Correlation (SC) technique, which estimates a user’s location by comparing the spatial distribution of accumulated RSSI vectors—referred to as the User RSSI Surface (URS)—with a pre-constructed radio map. SC captures the user’s signal pattern over space, improving localization accuracy. However, it relies on deterministic similarity calculations, which can be affected by noise and signal fluctuations commonly observed in indoor environments. In this study, we extend the SC framework by introducing a probabilistic formulation based on the Naive Bayes approach. We construct the URS using a sequence of BLE-based RSSI vectors synchronized with the user’s estimated trajectory and compute the posterior probability of each reference point (RP) given the observed URS. By modeling the conditional likelihood of RSSI values under each RP, our method accounts for the uncertainty inherent in wireless signal measurements and enables more robust location estimation. Experiments conducted on the L3 floor of KIST demonstrate that the proposed probabilistic SC significantly improves localization accuracy, particularly in environments with high RSSI variability, compared to the original SC method.