Automatic Radio Map Database Maintenance and Updating Based on Crowdsourced Samples for Indoor Localization

Recently, the crowdsourcing-based fingerprint technique has been promoted as an assuring solution for indoor localization positioning systems to alleviate the burden of conducting a site survey. When reconstructing and updating an indoor localization radio map from crowdsourced data, the following challenges addressed; powers of some Access Point (APs) change, adaptive radio map updating, and inaccurate location Information. In this manuscript, we propose automatic APs Changed Powers Identification and Radio Map database Updating (ACIRMapU) technique based crowdsourcing for indoor localization to handle these challenges. The ACIRMapU consists of APs changed powers detection and identification, fingerprints updating, and online localization positioning algorithms. For each crowdsourcing sample, we first propose a changed APs power detection and identification algorithm. After that, we proposed a fingerprint Learning Vector Quantization (LVQ) algorithm to update indoor radio maps due to changed AP. In the online localization positioning phase, an AP feature selection is performed by the Pearson Correlation Coefficient (PCC) algorithm followed by the Weighted K-nearest Neighbor based on signal Euclidean distance (WKNN-SED) to estimate the target location point. We conduct extensive field measurements and experiments to investigate automatic radio map updating adaptation. Results demonstrate that the proposed ACIRmapU can outperform significant improvements over the state-of-art approaches in the changed APs powered removed, increased, and decreased, whereby the localization errors in updating the radio map have reduced more than 76.86%, 61.83%, and 36.53%, respectively. Meanwhile, achieving average localization accuracy around 0.6325 m, 1.0435 m, and 1.735 m, respectively.