Research On Intelligent Indoor Positioning Technology Based On 5G Fingerprint

With the development of science and technology,the application services based on location information are full of people’s daily life.These application services need the support of positioning technology,which makes the research on positioning more and more attention.At the same time,the related technologies of the Fifth-generation mobile communication(5G)also began to be widely integrated with positioning technology.At present,the outdoor positioning accuracy has basically met the users’ requirement,but the realization of indoor high-precision positioning still has serious challenges.With the standardization and application of 5G system,5G base stations(BSs)will be densely deployed and can effectively improve the accuracy of indoor positioning by using BS information.To improve the positioning accuracy and reduce the data acquisition cost,this thesis studies the positioning algorithm of 5G indoor environment.The relevant work is as follows:(1)The indoor positioning method based on 5G fingerprint is studied.Firstly,the data preprocessing method is studied to ensure the positioning accuracy.Then,the indoor positioning methods based on ranging/angle measurement and fingerprint are studied.Finally,the transfer learning indoor positioning algorithm based on 5G fingerprint are studied.(2)A weighted adaptive K-nearest neighbors(KNN)algorithm with historical information fusion is proposed to improve the indoor positioning accuracy of moving target.A positioning result of target is obtained by the improved weighted Euclide distance and the adaptive K value selection of KNN algorithm.According to the historical information fusion method,the positioning result is fused with the previous position of target to obtain the final positioning result.The proposed algorithm is tested by datasets form two different environments to fully verify its effectiveness.(3)A transfer learning indoor location algorithm based on derivative fingerprint is proposed to reduce the reconstruction overhead of fingerprint database after time-varying environment.A positioning result of target can be obtained by three types of derivative fingerprints and the optimal K value of the reference point.Combined with partition transfer learning,the offline fingerprint database after time-varying environment is reconstructed to obtain the positioning of target.The proposed algorithm is tested through experiments to verify its effectiveness.