A Serial Maximum-likelihood Detection Algorithm for Massive MIMO Systems

As an important part of massive Multi-Input Multi-Output (MIMO) technologies, signal detection has been studied in the literature in recent years. The detection complexity grows significantly as the number of antennas increases in the system. Maximum-likelihood (ML) has the optimal performance with the highest complexity, which is prohibitive for implementation. In this work, we propose a serial ML (SML) algorithm, which changes the way of detection from parallel multi-dimensional searching to serial single-dimensional searching to reduce detection complexity. Besides, we employ a valid initial value for the proposed algorithm to obtain a faster convergence. Based on the simulation results, for the system with 128 receive antennas, the proposed SML algorithm outperforms the Minimum Mean Square Error (MMSE) method under different numbers of users and modulation schemes. When achieving a similar performance, the complexity of serial ML is almost a half of that of low complexity Message Passing Detection algorithm in the system with 16QAM and 16 or 32 users. It is demonstrated that our proposed SML method is more suitable for signal detection when the system adopts low order modulation schemes and serves larger number of users.