Extended Accelerated Over-Relaxation Algorithm for Low Complexity Massive MIMO Signal Detection

Abstract

This letter proposes an extended accelerated over-relaxation (EAOR) detector for massive multiple-input multiple-output (MIMO) systems that significantly improves convergence speed while maintaining low computational complexity. The EAOR method introduces a three-parameter optimization framework that extends classical two-parameter iterative schemes, providing enhanced flexibility for spectral radius minimization. A novel power-aware initialization strategy exploits channel strength information to address interference asymmetry and accelerate convergence. The proposed detector achieves near-minimum mean square error performance with $O\left(K^2\right)$ per-iteration complexity, substantially outperforming existing AOR and second-order Richardson iteration methods. Simulation results demonstrate that EAOR consistently achieves lower bit error rates with fewer iterations across various antenna configurations and modulation schemes. The enhanced convergence characteristics make EAOR an attractive solution for practical massive MIMO implementations in next-generation wireless systems.