| The great demands for LBS have been generated in many fields with the advent of 5G era and the development of intelligent terminals and the Internet of Things,such as business,transportation and security.The precise positioning of mobile terminals is faced with greater challenges in complex environments such as modern urban streets and buildings.The multipath effects can be efficiently utilized by fingerprint positioning technologies to achieve better positioning accuracy.Therefore,the high precision fingerprint location algorithm for 5G network has become a popular research direction of LBS.There still exist many problems to realize the positioning accuracy required by 5G white paper.To realize real-time precise positioning of mobile terminals in practical applications,the rapid and accurate fingerprint localization algorithms for terminals are studied based on massive MIMO single-site systems in complex urban environments.The main researches are as follows.(1)The CAOA clustering algorithm based on the center angle of channel power is proposed according to the power distribution characteristics of ADCPM matrix in angle domain,which is aimed at improving the online localization speed.The algorithm is applied to different matching methods and the positioning speed is improved efficiently at online stage.The improvement of positioning speed for the CAOA clustering algorithm is limited with the increasing size of database.Therefore,the modified LSH-WKNN fast positioning algorithm based on CAOA clustering is proposed which maps the elements of each clustered subset into different buckets with LSH method and realizes the second clustering of database.The performance of improving the positioning efficiency of this algorithm is verified by comparing the positioning time under different size databases.(2)For the prominent problem of database storage load in practical application scenarios,the TT compression method which transforms the fingerprint matrix into three tuple without zero and the ACPV compressed fingerprints which are obtained by converting the ADCPM matrix into one dimensional vector of angle domain are both studied.The storage load can be significantly reduced and the positioning accuracy and speed are ensured simultaneously.The simulation results also show that the average positioning time can be further decreased with ACPV compressed fingerprints and the positioning accuracy is identical with ADCPM matrix.(3)The ScCU database update algorithm is proposed for the variability of database in actual dynamic environments,which sets a cut-off area for each changed scatterer and reconstructs the fingerprints of the reference points within the cutoff area.The updating time of the database increases with the radius of cutoff region.The WNIU algorithm is introduced to eliminate the effect of cutoff radius,which estimates the fingerprints of the non-feedback points with WNIU algorithm based on the database of feedback points.Since the influence of fingerprints correlation is ignored by WNIU algorithm,the KrIU algorithm is presented on the basis of Kriging interpolation method.The simulation results indicate that the updating time of the KrIU algorithm is greatly decreased and the positioning accuracy is close to the traditional updating algorithm,which can realize the precise localization in dynamic environments. |
PDF Full Text Request