Improved LAS detection based on grouped genetic algorithm for massive MIMO system

Massive Multiple-input multiple-output (MIMO) system shows promising future due to the high spectral efficiency and improved capacity. However, one major issue in massive MIMO detection is the balance between complexity and performance. The LAS detection has low complexity, however, it shows difficulty to escape local optimal due to the greedy strategy which leads to performance loss. In this paper, an improved algorithm combining likelihood ascent search (LAS) and grouped genetic algorithm (GA), termed as GGALAS, is proposed. We take fully advantage of the global search ability of the genetic algorithm and the local search ability of the LAS. By changing MMSE detection to generate initial solution, designing several groups for search diversity and applying complexity reduction method, the simulation results demonstrate that the proposed algorithm can achieve near SISO AWGN performance in massive MIMO system. It has the same complexity order as 1-stage LAS but outperforms the 3-stage LAS in BER performance. The proposed algorithm also has much shorter time delay than 3-stage LAS. For example, the time required for 4GGA only accounts for 0.18% of the 3-stage LAS in 64 x 64 system at 4QaM.