Multiple-compression-rate CSI feedback for mm-wave massive MIMO based on federated learning
Bo Hu,Hongbo Xu,Ze Wang,Li Zhao,Aizhi Zhou
DOI: https://doi.org/10.1016/j.phycom.2024.102305
IF: 2.379
2024-01-28
Physical Communication
Abstract:In mm-wave massive multiple-input-multiple-output (MIMO) systems, accurate channel state information (CSI) at the base station (BS) is the key knowledge to obtain the performance gain. Consequently, the user is not only required to complete the channel estimation, but also to feedback CSI to the BS. However, large-scale antenna arrays result in a substantial feedback overhead, which poses a challenging issue. Furthermore, the accuracy requirements of channel estimation and CSI feedback depend on the computing capacity of the user. In this paper, we propose a joint channel estimation and multiple-compression-rate feedback (JCEMF) scheme, and adopt centralized learning (CL) and federated learning (FL) strategies for the scheme. According to the limited computational resources available to users, the JCEMF scheme enables various lengths of feedback bits to change the feedback overhead. Additionally, the users in FL train the local models using their own datasets and upload the local model updates to the BS, thereby reducing communication overhead and protecting data privacy. Specifically, an estimation network is designed for the user to estimate the channel from the received signal. In the CSI feedback process, we introduce an MCRF network, which can achieve CSI compression and reconstruction with different numbers of feedback bits. Simulation results verify that the proposed approach shows good performance of joint channel estimation and multiple-compression-rate CSI feedback in different channel conditions.
telecommunications,engineering, electrical & electronic