Abstract—The technological advancements and demand of high speed communication has led to evolvement of Multiple-Input and Multiple-Output (MIMO) and massive MIMO (mMIMO) communication systems. However, the increased number of antennas lead to an increase in computational complexity and implementation cost. Moreover, achieving the performance to meet the communication demand also remains a challenging task. The current researches have reported that the precoding scheme can help to minimize the computational complexity and increase the performance of mMIMO system. Hybrid precoding schemes have gained huge attention due to their significant nature to improve the overall efficiency of the system but the traditional schemes usually focus on optimization or greedy mechanism which suffer from the complexity issues and provide the sub-optimal performance. Moreover, the performance of these systems is directly affected by the quality of channel data. Therefore, we present a Deep Learning (DL) based approach using Deep Neural Network (DNN) model which uses limited feedback mechanism to handle the compression and reconstruction error. It aims to minimize the reconstruction error by providing the transmitter with sufficient information about the Channel State Information at the Receiver (CSIR) despite using a reduced amount of feedback compared to full feedback systems. This scheme uses encoder and decoder based module for limited feedback modelling. In order to prove the robustness of proposed DL based approach, we have presented extensive experimental analysis where the proposed DL based mechanism achieves average performance as 16.85 bits/s/Hz, 12.45 bits/s/Hz, and 8.028 bits/s/Hz in terms of achievable rate, spectral efficiency and average sum rate respectively. In contrast to this, the existing Simultaneous Orthogonal Matching Pursuit (SOMP) achieves the average sum rate as 6.042 bits/s/Hz.
 
Keywords—deep learning, limited feedback, precoding, compression, massive Multiple-Input and Multiple-Output (mMIMO)


Copyright © 2025 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).