Simplified matrix polynomial-aided block diagonalization precoding for massive MIMO systems

In the downlink of Massive Multiple-Input-Multiple-Output (MIMO) systems, the high computational cost of precoding is a major challenge for real-time data transmission. In this paper, we propose a simplified matrix polynomial-aided block diagonalization (SMP-BD) precoding scheme to simplify the conventional block diagonalization (BD) type precoding schemes for users with multiple antennas. By replacing the singular value decomposition (SVD) operation with a matrix inversion, which is then approximated by matrix polynomials with optimized coefficients, SMP-BD is shown to be hardware-efficient, reduce the transmission overhead and obtain high performance. The optimized coefficients can be calculated offline and the convergence of the matrix polynomials is guaranteed. Simulations show that SMP-BD with polynomial order L = 3 performs close to previously reported algorithms based on matrix inversions, while being much simpler.