分布式IRS辅助大规模MIMO上行链路分组迁移学习检测算法OA

Centralized single intelligent reflecting surface(IRS)-aided multiuser massive multiple input multiple output(MIMO)systems face significant challenges.The uplink cascaded channel suffers from extensive crosstalk and degraded detection performance at the base station(BS)due to the large number of users,leading to high bits error rates(BER)and reduced spectral efficiency.Further,the repetitive and frequent design and control of the IRS phase imposes a heavy load on the BS.To address these is-sues,this study constructs a distributed IRS-aided uplink multiuser massive MIMO communication archi-tecture and a detection algorithm based on grouped transfer learning(GTL).Simulation results validate the correctness and efficacy of the proposed approach.Notably,even when the BS lacks knowledge of the IRS phase and the channel state information from users to the IRS,the proposed GTL-based detection method in the distributed IRS-aided architecture achieves the BER and spectral efficiency close to theo-retical bounds,evidently outperforming those centralized single-IRS-assisted systems.Under limited training data,the proposed GTL-based detection algorithm features a faster convergence rate and attains a lower BER and a higher spectral efficiency compared to traditional convolutional neural network-based detection methods.