Indoor Positioning Technology And Application Based On 5G Network Channel State Information

With the rapid development of cities and towns,people's living standards are constantly improving.At the same time,the demand for indoor positioning accuracy is also increasing.In daily life,positioning and navigation in various indoor environments such as shopping malls,subways,hospitals,office buildings and other public places are becoming more and more common.However,the accuracy of indoor positioning has not been high,and more in-depth research is urgently needed to improve the accuracy of indoor positioning.At the same time,with the development of the fifth-generation mobile communication technology,millimeter wave communication technology and the communication characteristics of low delay and high bandwidth have made highprecision indoor positioning possible.The main content of this thesis is to study static positioning and dynamic tracking for the indoor environment in 5G networks.In a general indoor scenario,multiple access points are required to achieve positioning during static positioning and dynamic tracking.In this study,a space-time two-dimensional spectrum estimation algorithm is used to process the channel state information to obtain the angle of arrival and flight time of a single anchor node.The method of constructing fingerprint library is adopted to realize static positioning,and the algorithm of Kalman filter is improved to carry out dynamic tracking.By studying the static positioning and dynamic tracking of a single anchor node,the feasibility of positioning a single anchor node is determined.Therefore,the number of devices configured in the indoor positioning process can be effectively reduced,and at the same time,the synchronization problem between different positioning devices is avoided.The main research contents are as follows:(1)Aiming at the problem that the results of space-time two-dimensional spectrum estimation are easily affected by the number of multipaths,a smoothing sequence algorithm is adopted.Before spectrum estimation,determine the number of multipaths contained in the channel state information data.Thereby improving the accuracy of the angle of arrival and flight time obtained by spectrum estimation.In order to prevent the contingency of the data obtained by single spectrum estimation,each sampling point is sampled multiple times,and the angle of arrival and time of flight obtained by multiple sampling of the same point are clustered.The maximum expectation algorithm of the Gaussian Mixture Model is used to cluster the data,and the obtained cluster centers are stored in the offline fingerprint database as the average value after multiple samplings,ensuring the accuracy of the offline fingerprint database.(2)Aiming at the problem of excessive matching amount in the online matching stage in the static positioning process,an improved K-nearest neighbor algorithm is proposed.In the arrival angle fingerprint database and the flight time fingerprint database,the flight time fingerprint database that causes a small error is selected for initial matching,and the matching range of the fingerprint database is reduced,and then the fingerprint database matching of the arrival angle is performed.On the premise of ensuring the matching accuracy,the amount of matching data is reduced and the matching efficiency is improved.(3)Aiming at the problem of inaccurate correspondence between virtual anchor nodes and multipath components in the dynamic tracking process,a data association algorithm based on the optimal sub-mode allocation metric is adopted.The redundant data that may exist in the multipath component data set is eliminated to ensure the accurate correspondence between the virtual anchor node and the multipath component data.It provides a more accurate input vector for the tracking equation of dynamic tracking.(4)Aiming at the problem that the Kalman filter algorithm is greatly affected by the nonlinear tracking equation,an improved Kalman filter algorithm is proposed.The nonlinear tracking equation is pseudo-linearized,and the deviation in this process is compensated to ensure the accuracy of dynamic tracking.