Research And Application Of Key Techniques For Smartphone-Based Indoor Localization

With the great improvement of location-based service,Wi-Fi indoor fingerprinting position technology has become a key research of indoor localization schemes.Based on the in-depth theory study of the Wi-Fi indoor localization,this thesis compares the existing problems and proposes corresponding solutions for altered access points(including position movement or power adjustment)and heterogeneous device differences.The main work of the thesis is as follows:Firstly,this thesis analyzes the signal characteristics in the indoor environment using the RSS location fingerprint data collected in the actual Wi-Fi scenario.A detection algorithm is proposed in this thesis,whose name is APWAA,namely Adapted Positioning with Altered Aps,to estimate whether there is any altered AP signal or not in the target RSS vector and solve the fingerprint database degraded accuracy problems caused by the altered APs.For every location query,APWAA performs the detection by dividing the RSS vector and checking the dispersion degree of the RSS vector subset.If there is no altered AP,the positioning will be performed according to the existing original scheme.Otherwise,the altered APs will be effectively filtered,and according to the signal similarity,the dense cluster which has no need of the preset number of clusters will be identified and positioned by the cluster-based method,so that the positioning performance can be guaranteed.The experimental results show that there is 79.6% identify accuracy of AWPAA,and the localization error remains 95% as compared to the original one after the system runs AWPAA.What's more,it achieves a 20% average localization error for the case of different numbers of altered APs.Secondly,aiming at the RSS quantification problem of positioning errors caused by differences in heterogeneous devices,this thesis proposes an RSS fitting method based on support vector regression so that the positioning system is able to compensate heterogeneous devices.Compared with existing calibration methods,experiments are conducted in this paper.The experimental results show that the proposed algorithm can effectively reduce the RSS differences of different terminals and maintain more than 90% of the localization accuracy.Different phones such as XiaoMi 5 and MeiZu Pro6 Plus have reduced 0.90 and 1.81 meters respectively after performing SVR fitting scheme.Finally,this thesis designs and implements a C/S-based Wi-Fi location fingerprinting system based on the above theoretical analysis.Firstly,it gives the top ideas of system implementation,and then discusses the details of the client,server and database respectively.And the algorithms above are applied to a Wi-Fi measurement site.The results of our experiment show that the methods proposed in this thesis can effectively adapt the AP changes and the heterogeneous devices.