Approximative Matrix Inversion Based Linear Precoding for Massive MIMO Systems

The linear zero-forcing (ZF) precoding has been considered as one of the practical techniques for massive multipleinput multiple-output (MIMO) systems. However, it involves complicated matrix inversion of high dimension. To approximate the channel matrix inversion of ZF precoding, we propose a joint Chebyshev iteration and Neumann series (Joint-CINS) precoding scheme, which is essential to improve the convergence rate of the Neumann series (NS) precoding. In addition, the Chebyshev iteration (CI) method is exploited to efficiently search direction for the Neumann series. Moreover, the iteration result of CI method is employed to reconstruct the Neumann series expansion to further speed up the convergence rate. Theoretical analysis demonstrates that the proposed Joint-CINS precoding scheme exhibits a faster convergence rate than the existing NS precoding. Simulation results indicate that the proposed JointCINS precoding scheme can achieve the satisfactory bit-errorrate (BER) performance and sum-rate approaching performance with smaller number of iterations than some previously proposed schemes.