Improving Multi-floor WiFi-based Indoor positioning systems by Fingerprint grouping

Indoor localization has an important role in different applications of Internet of Things (IoT). Wi-Fi fingerprinting indoor positioning systems (IPSs) have drawn a lot of attention due to their low Cap-Ex, but these systems suffer severe signal fluctuations which lead to accuracy reduction. In this paper Fingerprint (FP) grouping method is presented to improve Wi-Fi fingerprinting IPSs in multi-floor buildings. The proposed method consists of three sub-schemes. In each sub-scheme, FPs are grouped based on a different parameter, then in the online phase, the positioning algorithm is confined to specific FP groups determined by the method. Using this method, FPs which could reduce accuracy are filtered, and positioning is done on a subset of FP-database. In the first sub-scheme, FPs are grouped based on the received signal strength of Access Points (APs). In the second sub-scheme, FP grouping is performed based on the last estimated location of the user. Using the map-constrained graph, which is matched to the environment map, is the third sub-scheme. Not only these methods have improved the accuracy but also decreased the execution time spent in the positioning algorithm. The results of the practical tests indicate accuracy improvement of 6%, 47%, and 67% for each of the sub-schemes respectively and execution time reduction of 1.5 to 10 times.