Extrapolation Principle-Based Low-Complexity Signal Detection in Massive MIMO Systems

We propose a novel detection algorithm for massive MIMO systems, countering the high complexity of MMSE detection due to large-scale matrix inversion. Our approach is based on the extrapolation principle and incorporates the generalized extrapolation method into the successive overrelaxation (SOR) technique for accelerated convergence. Additionally, we introduce a new matrix initializer to optimize algorithm performance and hasten convergence, improving overall detection efficiency. Complexity analysis and numerical results demonstrate the superiority of our proposed algorithm over the conventional SOR detector, approaching MMSE performance. Specifically, in a system with 128 receive antennas that serves 32 users, the difference in performance between our algorithm and the MMSE approach is only 0.07 dB after five iterations. Notably, this is achieved with one-order lower computational complexity. This advancement presents our algorithm as a promising and efficient solution for massive MIMO system detection.