Research On Heterogeneity Of Device In Indoor Positioning Based On WiFi Location Fingerprint

With the development of information society towards more universal and intelligent direction,people's demand for indoor location services is increasingly urgent.However,indoor propagation environment is more complex than outdoor,which makes it difficult for outdoor satellite positioning mature technology to continue brilliant.People have pioneered the use of different wireless technology standards for indoor positioning research and exploration.WiFi technology has the characteristics of wide deployment,seamless indoor coverage and low cost.Therefore,indoor positioning method based on WiFi location fingerprint has attracted wide attention.However,the large-scale fluctuation of position fingerprint caused by device heterogeneity seriously affects the stability and robustness of indoor positioning system based on WiFi position fingerprint.For the problem of device heterogeneity,this paper studies fingerprint feature deformation and fingerprint feature extraction for fingerprint feature mining,and proposes two calibration-free positioning methods.The main research work is as follows.Firstly,aiming at the large-scale fluctuation of position fingerprint caused by heterogeneous devices,proposed a method named CLAS-STDRSS-ELM,a standardized sub-fingerprint database is constructed by combining the strongest AP classification and Prussian analysis,and a systematic calibration-free positioning method based on extreme learning machine is proposed.Then,aiming at small-scale fluctuation of position fingerprint caused by noise interference,CLAS-SDAE-WKNN method was proposed,which is a systematic calibration-free location method based on fingerprint feature mining,using stacked denoising autoencoder(SDAE)to learn standardized sub-fingerprint database to obtain depth feature fingerprints,constructing sub-fingerprint database of depth feature and using WKNN to estimate position.Finally,typical experimental floors as experimental environments.Experiments show that the two kinds of calibration-free methods can effectively reduce the impact of heterogeneous devices.The average positioning error of CLAS-STDRSS-ELM method is 2.89m for 200 points to be positioned,which is 16.7%higher than signal strength difference,that is a traditional calibration-free method;the average positioning error of CLAS-SDAE-WKNN method for the same 200 points is 2.72m,which is 5.9%higher than that of CLAS-STDRSS-ELM method.