Research On Millimeter Wave MIMO Channel Estimation Algorithm

With the commercial use of 5G,more and more attention has been paid to millimeter wave channel research.Millimeter-wave communication uses MIMO technology,and the expansion of antenna scale and complex precoding techniques bring challenges to channel estimation.Traditional channel estimation methods have high training sequence overhead and high complexity of channel estimation algorithms.In order to solve the above problems,this thesis conducts corresponding research on millimeter-wave MIMO channel estimation.Aiming at the high overhead of training sequences for downlink channel estimation in millimeterwave frequency division duplex systems,this thesis proposes to use the low-rank characteristics of millimeter-wave channels and matrix completeness theory to transform the channel estimation problem into a channel matrix rank minimization problem.This thesis proposes a channel estimation method based on non-convex Schatten-p norm factorization.This method uses the Schatten-p norm to decompose the channel matrix into two matrix factors,and reduces the complexity of channel estimation by introducing auxiliary variables and solving the sub-optimization problem of the matrix factors alternately.The simulation results show that the algorithm can obtain better channel estimation performance with less training sequences.Compared with the kernel norm,the smooth function is closer to the rank function of the channel matrix,but the smooth function is non-convex,and it is more complicated to solve it directly.This paper proposes to use the arctangent function to approximate the matrix rank function,change the optimization problem to the augmented Lagrangian form,and use the gradient descent method to solve the optimization problem,and finally obtain the channel estimation matrix.The convergence proof of the algorithm is given.The simulation results show that the algorithm has better convergence effect,less channel estimation error,and better anti-noise ability.