Research On Multi-floor Indoor Positioning Technology Based On RSS With CSI

With the continuous growth of wireless communication network coverage in the indoor environment,wireless positioning technology has become a solution to the poor signal of indoor regional navigation satellites and Location Based Service(LBS).In recent years,the Wi-Fi location fingerprint location algorithm based on received signal strength(RSS)and channel state information(CSI)has been widely studied and applied in indoor positioning.At present,the vast majority of indoor positioning research is aimed at a single floor.Facing the widespread multi-floor scene,relying solely on the two-dimensional plane position cannot meet the existing positioning requirements.For this reason,this paper is based on the practical application of multi-floor identification and in-floor positioning.To solve the problem of low stability and high computational complexity of existing floor recognition technology,an improved K-means clustering method based on RSS fluctuation interval is used for floor recognition;for the problem of low accuracy of RSS positioning in the in-floor environment,the method of RSS assisted CSI positioning is used.Specific research work includes:(1)Based on the Ubuntu 14.04.3TLS system,a CSI collection platform is built using Inte15300 network card,Xiaomi4C router,and CSI-tools software.The advantages of CSI as fingerprint information are verified by comparing RSS and CSI physical quantities.(2)For the problem of multi-floor identification,a multi-floor identification method based on RSS fluctuation interval and clustering is designed.In the offline phase,gridding the location area,collecting Wi-Fi data in the grids,and building the location fingerprint database,according to the through-wall fluctuation characteristics of RSS,the time-varying intervals of signals at different access points are defined,and the fast clustering algorithm based on density peak is used to improve the defects of traditional K-means clustering;In the online stage,the floor number of the target is determined by combining the floor signal fluctuation interval,the shortest distance of the cluster center and the voting mechanism..In this algorithm,only a small amount of Wi-Fi fingerprint information is needed to get the floor information of the target,with an average recognition accuracy of 98.1%.(3)To compensate for the low positioning accuracy of the Wi-Fi single feature,this paper uses the in-layer positioning method of homologous and heterogeneous feature fusion.Firstly,the rough position of the target is determined by RSS information,and then the reference points around the target are filtered according to the radius threshold to construct a CSI sub-fingerprint library.Finally,the two signal features are fused using the kernel function to generate the location fingerprint for positioning.The results show that the average positioning error of the fusion method is 1.64 m,96.4%is within 2.5 m,and the average positioning accuracy is 21.2%higher than that of the RSS-only method.Figure[36]Table[10]Reference[83]...