Research On Indoor Fingerprinting Localization Techniques Based On Low Rank Matrix Recovery

In recent years,with the rapid development of mobile communication technology and the wide spread of intelligent mobile terminals,location-based services have gained wide attention,and the positioning needs of people are increasing gradually.At present,the indoor fingerprint localization technology based on WLAN has gradually become a research focus due to its low deployment cost,implementation simplicity,extensibility and high positioning accuracy.The indoor localization algorithm based on fingerprint utilizes the real-time Received Signal Strength(RSS)to match with the fingerprint database to estimate the location information.Therefore,the fingerprint database construction in the offline phase is crucial for accuracy positioning.However,RSS is easily affected by multi-path,shadow effect and complex environment changes,which leads to the fingerprint database contaminated by outliers and noise,and decreases the positioning accuracy.In the meanwhile,the fingerprint database construction in the offline phase needs to collect RSS at each reference point,which needs a huge workload,and seriously affects the practicability and promotion of the method.For this reason,this paper has carried out the following research work by referring to the relevant literature and practical experimental analysis:(1)Through theoretical analysis and data experiments,it is proved that the spatial correlation and temporal correlation between RSS,which results in the low rank attribute of fingerprint database.Therefore,in order to solve the problem that fingerprint database contaminated by outliers and noise and improve the accuracy of the fingerprint database and online localization,we proposed to utilize Robust PCA technique to recover the fingerprint database.The advantages and disadvantages of the traditional Robust PCA model are analyzed in detail.We propose an improved Robust PCA optimization problem that replaces the nuclear norm with weighted nuclear norm and introducing 2,1 norm regularization term.In the meanwhile,ADMM method is applied to design an effective algorithm for the improved optimization model.The simulation data and real environment data experiments verify that the proposed algorithm can effectively remove outliers and noise of fingerprint database,and improve the accuracy of fingerprint database and online localization.(2)In order to solve the problem that it takes a lot of work to construct the fingerprint database in the offline phase,and some reference points in the location area can not be measured,an efficient fingerprint database construction algorithm based on matrix completion is proposed.Utilizing the spatial correlation of RSS and part of known RSS to construct the complete fingerprint database efficiently.In order to improve the accuracy of construction and the robustness to noise,an improved matrix completion model is proposed.And the corresponding effective algorithm is also designed in detail.At the same time,to solve the problem that matrix completion can not deal with all zero rows and columns,we propose to combine the matrix completion with KNN algorithm.Finally,the simulation data and real environment data experiments show that the proposed algorithm can effectively fill the data of unknown points,and reduce the workload in the offline phase.And the proposed algorithm has better performance than other algorithms.