Robust Access Points Selection Strategies for Dynamic Indoor Localization

The proliferation of mobile devices has fueled the demand for indoor location-based services. Consequently, a plethora of techniques have emerged to facilitate object and device localization in indoor environments. Among these, fingerprint-based indoor localization systems, which leverage machine learning, stand out as a promising solution for providing accurate localization. Nonetheless, their performance is inherently reliant on the accuracy of the underlying database, and any changes in the indoor layout can significantly impact the wireless signals, subsequently affecting the localization accuracy. To circumvent this issue, this work proposes a novel access point (AP) selection framework to enhance the robustness of fingerprint-based indoor positioning systems in dynamic indoor environments. More specifically, a hybrid Wi-Fi and BLE fingerprint database is constructed by collecting received signal strength (RSS) from the pre-defined reference points (RPs). To ensure that the fingerprint database remains relevant over time, some RPs are designated as known points so that the system can periodically collect new RSS at the known points. Subsequently, the proposed scheme computes the differences between the RSS of the database and the updated RSS from the new layout to account for the changes occurred. The building-based, floor-based, and zone-based implementation modes determine which APs are reliable to be utilized during localization based on the RSS discrepancies observed in building, floor, and zone, respectively. Results demonstrate that the proposed building-based, floor-based, and zone-based AP selection schemes could achieve reduction in positioning error up to 28.86%, 33.53%, and 39.66%, respectively, compared to the baseline technique without AP selection schemes.