Research On Pilot Multiplexing Algorithm In Massive MIMO-D2D System

Massive MIMO(multiple-Input multiple-output)one of the key technologies of 5G,which greatly improves the spatial multiplexing gain and spatial grading gain of the system,and meets the exponential growth of mobile services.While the number of terminals is growing rapidly,massive MIMO systems are prone to overloading at the base station side,and there are coverage blind areas at the cell edge.On the other hand,D2D(device-to-device)technology enables two nearby devices to establish a direct communication link through a base station,and introducing D2 D communication technology into a large-scale MIMO system under the control of a cellular system is beneficial to Solve the overload of the base station,reduce the communication delay,reduce the coverage blind area and improve the system spectrum efficiency.However,the multiplexing of cellular resources by D2 D users at the same time aggravates the pilot overhead of massive MIMO,resulting in serious pilot pollution.Therefore,how to design a reasonable pilot multiplexing method has become the first to be resolved in the introduction of D2 D technology in massive MIMO systems.problem.Based on the above problems and challenges,the paper first introduces the pilot multiplexing of large-scale MIMO single cells,and then focuses on the problems of spectral efficiency and pilot multiplexing after introducing D2 D into the massive MIMO system.The main contributions and innovations of the paper The points are summarized as follows:1.This paper introduces the situation of pilot multiplexing in a single cell in massive MIMO,laying a good foundation for pilot allocation in D2 D communication in massive MIMO.Channel modeling based on ray tracing channel modeling method shows that when different users of multiplexed pilots do not overlap with each other in their spatial angles of arrival,the sum of the mean square errors of the channel estimates can be minimized,thus confirming the large Feasibility of pilot multiplexing on large-scale MIMO channels.Provide corresponding theoretical basis for pilot multiplexing in massive MIMO-D2 D heterogeneous networks?2.We have studied the pilot multiplexing algorithm based on weighted graph in massive MIMO-D2 D heterogeneous network.In order to reduce the frequency overhead and mitigate the impact of pilot pollution,D2 D reuses the orthogonal pilot sequences of cellular users.Inspired by graph coloring,an interference graph is constructed based on the distance between cellular users.This is intended to limit reusable pilot sequences to each cellular user and determine the order in which the pilot sequences are allocated.In addition,D2 D users are assigned so that AOA does not overlap between users using the same pilot sequence,where the covariance matrix of the channel has information related to AOA.Therefore,our scheme is flexible in terms of combination and improves channel estimation accuracy.Simulation results show that the scheme is superior to the traditional scheme in terms of channel estimation accuracy and system spectral efficiency for the UE.3.A pilot multiplexing system based on a convolutional neural network in a massive MIMO-D2 D heterogeneous network is studied.In order to reduce the channel estimation error caused by pilot multiplexing.By minimizing channel interference strength as a cost function,a pilot allocation system for a convolutional neural network is established.First,the training data set is obtained by exhaustive search.Then,analyze the training data set.The proposed CNN-PAS can provide an optimal pilot multiplexing result for a given D2 D user position from the generated inference function.As the number of samples changes,the resulting inference function is used to provide approximately optimal pilot allocation results.Simulation results show that this scheme achieves nearly 98.78% of the theoretical upper limit performance.