Performance Research Of Transmission Scheme Based On User Grouping In FDD Massive MIMO System

At present,the fifth-generation mobile communication technology(5G)is in an important stage of standard updating and industrial configuration.It will open the era of Internet of Everything with a new network architecture and key technologies.Joint Spatial Division and Multiplexing(JSDM)can significantly reduce the feedback overhead of Chanel State Information at Transmitter(CSIT),which will be used for Massive MIMO(Massive multiple-input multiple-output,Massive MIMO)implementation in FDD mode has great practical significance.Therefore,this paper deeply studies the two-stage precoding design,user scheduling strategy and the specific implementation scheme of user grouping key technologies in JSDM transmission scheme.Firstly,aiming at the high complexity of greedy user scheduling method in JSDM system,a joint scheduling strategy based on improved channel spatial correlation is proposed.According to the orthogonality between user channel vectors,the optimal orthogonality is selected.The user acts as the first scheduled user,and then uses greedy scheduling to maximize the system and rate to complete the scheduling of other candidate users.The simulation results show that the performance of the improved strategy can approach the greedy user scheduling strategy at lower computational complexity.Then,for the two-stage precoding,the"energy penalty"performance limitation problem occurs after the traditional linear precoding method.The vector perturbation(VP)strategy is introduced to improve the regular zero forcing(RZF)precoding VP-RZF.A perturbation vector search algorithm ML-K-best improved based on K-best tree is proposed to further solve the problem.This method no longer needs to perform exhaustive search in the full space of the K layer,only exhaustive search in the first N layer,and only retain few paths with the smallest branch metric value after each layer search is completed to continue to the next layer.Searching ensures the search accuracy of the perturbed vector while effectively reducing the complexity of the spherical coding algorithm.The simulation results show that the improved JSDM system sum rate based on VP-RZF precoding algorithm are further improved.Finally,for the K-means grouping in the JSDM scheme,there are preset packet numbers,and the problem of randomly selecting the packet center proposes a DP-K-means grouping algorithm based on the improved density concept:firstly,the users are divided into cores according to the user's distribution density.Users,edge users,and orphaned users.The core users can be connected to the density to form G core user groups,and then calculate the centers of the G core user groups as the initial grouping center of the K-means algorithm.The improved scheme has three major advantages:the first selection is a more representative initial grouping center,which can accelerate convergence while avoiding falling into a local optimal solution.Second,the optimal number of packets can be adaptively determined according to the geometric cluster of user distribution,and the preset is no longer fixed.Thirdly,the problem of exacerbating inter-group interference caused by the excessive difference in the number of users between groups is avoided when the density clustering is simply used.The simulation results show that the improved scheme only needs three iterations to complete the grouping,and the grouping effect is better,and the system sum rate are greatly improved.