Research On Fast Convergence Detection Algorithm Based On SOR In Massive MIMO System

Massive MIMO technology is one of the core technologies of 5G.It equip with hundreds of signal to receive antennas at the base station which greatly improves the performance of channel capacity,spectrum efficiency,transmission rate and reliability,However,the complexity of signal processing also increase sharply when the number of antennas and the number of users increase.Therefore,it's urgent to discuss the matter how to reduce the complexity in the signal detection of large-scale MIMO systems.It can be found that the Minimum Mean Square Error(MMSE),the typical representative of linear detection algorithm,can achieve better detection performance with a relatively low computational complexity.But,the defect of it is the high dimensional matrix inversion.Experts and scholars have proposed a number of matrix inversion simplification algorithms,such as series expansion algorithm and iterative approximation algorithm.In this paper,the problem of the accelerated convergence of the Successive Over Relaxation(SOR)iterative algorithm and the high complexity of the Neumann series expansion term are studied.Firstly,in view of the slow speed of SOR algorithm,this paper proposes a low complexity hybrid iterative algorithm SDSOR based on the SD algorithm and SOR algorithm.Comparing with the poor initial convergence of the SOR,the SD algorithm has a better convergence direction at the beginning of the iteration.Therefore,combining the SD and SOR algorithm,the SD algorithm can provide an effective search direction for SOR algorithm.Then,it can speed up the convergence of the SOR algorithm.Eventually,the performance of the algorithm detection can be improved.The simulation results show that the hybrid iterative algorithm SDSOR possess three advantages.It can converges faster and achieve the same performance of MMSE detection with fewer iterations.Meanwhile,it can reduce an order of magnitude of algorithm.Secondly,considering the slow convergence speed of SOR and the high complexity of Neumann expansion term algorithm,this paper proposes a low complexity hybrid algorithm NESOR based on SOR algorithm and Neumann series expansion.Here,the NESOR algorithm avoids the problem of accurate inversion of high-dimensional matrices.Because the complexity of the second-order expansion term algorithm is relatively low,it can be used as initial value of SOR to accelerate the convergence speed.The simulation results indicate that NESOR achieves better detection performance than the traditional Neumann series expansion algorithm,and it can reduce one order of magnitude of complexity compare with MMSE.At the same time,the NESOR algorithm converges faster than the SOR algorithm,and can achieve the same performance of MMSE detection with fewer iterations.This paper introduces two approximate log-likelihood ratio calculation methods to further reduce the computational complexity of the algorithm implementation.