Towards Floor Identification and Pinpointing Position: A Multistory Localization Model with WiFi Fingerprint

The era of Internet of Things (IoT) has stimulated the diversification of wireless applications, and a pragmatic way is to adopt and leverage WiFi to pinpoint the position of a mobile device. However, there still exist significant challenges in this field, such as heterogeneous crowd-sourced data distribution, external scene interference, etc. We focus on indoor WiFi fingerprint localization in multistory buildings. To confine the search scope to a specific floor, we propose a novel floor identification module. In this module we construct a WiFi fingerprint graph representation to fully explore the correlations of reference points (RP). Furthermore, a fingerprint graph attention mechanism is introduced to capture the importance of adjoining fingerprints for a more accurate floor identification. In addition, a two-panel fingerprint homogeneity graph is adopted to gauge the resemblance of localization fingerprints, and the estimated 2-D location is predicted by the integration of panel results. By comprehensively analyzing the fingerprint attributes of a crowd-sourced database, we have conducted experiments to demonstrate the localization algorithm's performance. Compared with other algorithms, the results show that the proposed method can achieve the best performance in floor identification, reaching 96.93%; In the aspect of 2-D geometric positioning, the proposed method also has better performance.