An optimized preprocessing Gauss-Seidel iterative detection algorithm for massive MIMO systems

The mini-mum mean square error(MMSE) algorithm nearly achieves optimal performance in massive Multiple-Input Multiple-Output (MIMO) detection, but its direct application is limited by the computational burden of high-dimensional matrix inversions. we propose an Optimized Preprocessing Gauss-Seidel (OPGS) iterative detection algorithm in this paper. The OPGS algorithm transforms the MMSE filter matrix into a linear equation, reducing complexity and eliminating the need for direct inversion. Additionally, by introducing a banded matrix to optimize the filter vector, we derived a novel iterative method that significantly improves both performance and convergence speed. We tested the proposed algorithm under various conditions, and simulations show that it requires fewer iterations to achieve detection performance similar to the MMSE algorithm under the same conditions. Specifically, in system with $256\times 32$, after two iterations, the performance between our algorithm and the MMSE method is only 0.04 dB. Notably, this result was accomplished with a minimal iteration.