Research On 5G Terminal Fingerprint Location Algorithm And Fingerprint Database Construction Method

Nowadays,LBS has become a basic demand in people’s daily production and life,such as smart home,emergency search and rescue,express delivery and so on.The traditional wireless location technology will be affected by the multipath effect in complex NLOS environment,while fingerprint location technology can make full use of multipath information for effective location,so it has attracted extensive attention and research.Large scale MIMO,UDN and other technologies in 5G network can provide richer channel information for fingerprint location.Therefore,5G network not only puts forward higher requirements for the accuracy and speed of wireless location,but also provides a good opportunity for the development of fingerprint location technology.In 5G single station large-scale MIMO-OFDM scene,this paper studies the high-performance fingerprint location algorithm based on FRCPM and the construction method of "one click" fingerprint database based on PDR,mainly including the following contents:Firstly,in the two-dimensional scene,the location algorithm based on FRCPM is studied from the perspective of improving the localization accuracy,mainly including fingerprint similarity criterion and matching algorithm.In the experiment.ADCPM is used as the comparative fingerprint.In terms of fingerprint similarity criteria,the positioning accuracy of different fingerprints with different similarity criteria such as European distance and Manhattan distance is verified,and the optimal similarity criteria of different fingerprints are obtained.In terms of matching algorithm,KNN,WKNN,SVR,BPNN and other algorithms are verified.In order to further improve the positioning accuracy,WKNN algorithm is improved,and the average positioning error based on FRCPM is reduced by about 10%.Secondly,in the two-dimensional scene,the high-performance fingerprint location algorithm is studied from the perspective of improving the location speed,mainly including fingerprint matrix compression and fingerprint database clustering.In terms of fingerprint compression,The performance of TT compression and compressed fingerprint like ACPV,DCPV and ADF are studied for ADCPM,and FCPV,An CPV and An FF are studied for FRCPM.In the aspect of fingerprint database clustering,in order to solve the problem that K-means clustering,CAOA-LSH clustering and other algorithms will lose part of the location accuracy,a PBC clustering algorithm which can greatly improve the positioning speed without losing the positioning accuracy is proposed and verified.Thirdly,a three-dimensional system model is constructed.The high-performance fingerprint location algorithm based on FRCPM in two-dimensional scene is verified in this three-dimensional system model,the similarities and differences between the algorithm in three-dimensional scene and two-dimensional scene are obtained,and the algorithm in three-dimensional scene is adjusted and improved.Finally,aiming at the problem of heavy workload and strong professionalism in establishing fingerprint database in offline stage,a "one click" fingerprint database construction method based on PDR is designed.The distribution characteristics of horizontal and vertical errors when PDR technology is used to travel along a straight line are verified through experiments based on an Android mobile phone.Then,the influence of independent and identical distribution of reference node coordinate errors and cumulative errors on location accuracy is studied through simulation experiments,and the designed method of fingerprint database construction is improved accordingly.