Abstract

The paper analyzes the performance of Massive MIMO system with LDPC code. Utilizing the benefit of the power of model driven deep learning, which is a deep learning technique whose main characteristic is to maintain the mathematical structure of the model, converting the iterative detector into a neural network by making the parameters learnable rather than fixed. This method requires less data and has a faster training rate. This learning technique is used in order to enhance the output of the detector and thereby enhance the output of the LDPC decoder. The simulation outcomes reveal the strength of integrating both OAMP-NET detector with LDPC to scalable massive MIMO systems with different antenna configurations ranging from 4×4 to 64×64. The proposed system OAMP-NET+LDPC achieves a significant improvement in detection performance in both perfect and imperfect CSI scenarios. In addition, we conducted a comprehensive analysis of scalability and computational complexity to evaluate its practical feasibility for different MIMO configurations. In addition, we conducted generalization experiments under unseen SNR conditions and variable channel correlations, in addition to ablation studies at the depth of the unfolded network to prove the robustness and effectiveness of the proposed system.