Research On Localization Algorithm For Massive MIMO Single Base Station

With the continuous development of the mobile Internet,people’s demand for location based services is increasing rapidly.Nevertheless,the high-rise buildings in the city lead to complex and changeable signal propagation paths,which causes the fact that traditional positioning methods can not meet people’s needs for real-time and accurate positioning technology.As one of the key technologies of 5G,massive MIMO technology can extract finegrained features of multipath channels by adding antenna arrangements,which makes it possible to achieve accurate positioning in massive MIMO single base station systems.The single base station positioning technology based on massive MIMO system in two-dimensional scenarios is studied in this paper,and the main contents are as follows:1.From the perspective of improving the positioning accuracy,this study proposes a geometric positioning method combining the Gram matrix of multi-carrier signals and virtual nodes to solve the problem of large parameter errors faced by traditional estimation algorithms.This algorithm uses the single-sample snapshot information of the receiver to estimate channel parameters in the angle and time delay domains respectively,and combines with the VB-GMM algorithm to further reduce the parameter error.Then,the multipath parameters are corresponding to the virtual nodes,and the least weighted squares method is used to realize the positioning.The simulation shows that compared with the traditional method,this algorithm has smaller estimation error and higher positioning accuracy,and can achieve a positioning error within 2 meters with 87% probability in the experimental scenario.2.From the perspective of reducing the positioning time,this paper studies the positioning algorithm based on ADCPM fingerprints,and proposes two fast positioning schemes based on spectral clustering algorithm and improved VGG algorithm.Firstly,based on ADCPM fingerprints,the influence of different similarity criteria on localization performance is verified,and Manhattan distance is selected as the measure of similarity.Secondly,the spectral clustering algorithm is used to divide the subsets,reducing the number of samples in the matching operation and improving the time-consuming of system positioning.Finally,the fitting and regression ability of the improved VGG algorithm is used to train the network model parameters and improve the positioning speed.Simulations show that when the total number of samples is small,the spectral clustering algorithm can achieve high-efficiency positioning with similar positioning errors.For a large-capacity fingerprint database,the positioning system using the improved VGG algorithm has obvious advantages.Under the condition of satisfying the positioning accuracy,the online positioning time can be shortened to tens of milliseconds.