Optimization Of Hybrid Precoding Scheme For Large-scale Antennas

In recent years,the requirement of communication rate is increasing year by year.The existing microwave band can not meet the bandwidth requirement,so millimeter wave band will be used.Although this band has a large bandwidth,its wireless transmission loss is also very serious.In order to compensate for this path loss,precoding technology is needed.Full digital precoding technology has high computational complexity,and full analog precoding technology has low precoding accuracy,so a hybrid precoding technology has been developed,which achieves a good compromise between performance and complexity.In this paper,millimeter-wave channel characteristics based optimization of hybrid precoding for large-scale antennas is studied.Firstly,the design of hybrid precoding requires feedback of channel state information from the receiver.Under the large-scale antenna setting,the channel feedback is large.Considering the sparsity of millimeter-wave channel,this paper designs a multidimensional scaling algorithm to compress channel feedback information.In many cases,although the observed data samples are high-dimensional,perhaps only a low-dimensional distributions are closely related to the design task,that is,a low-dimensional embedding in high-dimensional space.Multidimensional scaling is to transform the original high-dimensional attribute space into a low-dimensional "subspace" through some mathematical transformation,in which the sample density is greatly increased to achieve channel compression.The simulation results show that the spectrum performance of the proposed scheme is good,and when the number of data streams is large,the spectrum efficiency of the proposed scheme is better than that of the finite feedback scheme based on the effective channel,and close to the full feedback scheme.Then,in order to reduce the complexity of the optimization problem of hybrid precoding,a low-complexity hybrid precoding algorithm is proposed based on Schmidt orthogonal hybrid precoding algorithm.According to the number of radio frequency chains,the algorithm finds one or more vectors closest to the optimal precoding vector in the initial codebook.The number of vectors is determined by the number of radio frequency chains.The selected vectors are taken as the sub-optimal precoding vectors,and then Schmidt orthogonalization is performed to get the precoding matrix.The simulation results show that the proposed algorithm has the lowest complexity compared with other algorithms,and its spectral efficiency is close to that of the high complexity algorithm when the number of radio frequency chains is large.Finally,considering that the unit modulus limit is the root cause of the non-convex problem of hybrid precoding.Convex relaxation and manifold optimization are used to solve this problem.Firstly,the constraints of unit amplitude are relaxed by convex relaxation algorithm.The original problem is transformed in to a convex optimization problem,and the global optimal solution is obtained.Then,the suboptimal solution of the original problem is obtained by compressing the unit modulus condition.In addition,the original non-convex problem is regarded as a manifold optimization problem.By calculating the gradient,the optimal solution of the manifold space can be obtained,and then the sub-optimal solution of the original problem can be obtained through the constraints.The simulation results show that the spectral efficiency of these two methods is superior to other algorithms,and very close to the full digital system.Aiming at manifold optimization algorithm,combined with one-dimensional accurate search,an modified manifold optimization algorithm is proposed.The simulation results show that the spectral efficiency of the modified manifold optimization algorithm is better than that of the original algorithm.