Joint Channel Estimation and Feedback in Multiple-Input Multiple-Output Systems by using Pilot Feedback Network with Adaptive Compression Scheme

Nowadays, Massive Multiple-Input Multiple-Output (MIMO) has become important in 5G which enhances channel capacity by developing large antenna arrays at the Base Station (BS). Traditional approaches had faced several challenges which include high pilot overhead and increased computational complexity. Therefore, this research proposes Pilot Feedback Network with Adaptive Compression Scheme (PFnet with Adaptive Compression Scheme) for Channel Estimation (CE) and feedback in MIMO systems. Initially, data is generated from COST 2100 model. Then, a massive MIMO system with OFDM subcarriers and N_t antennas are considered. For channel estimation, the Least Squares (LS) estimator with interpolation is employed because of its low complexity by making it suitable for scenarios where computational efficiency is prioritized. However, to enhance estimation accuracy, a deep learning-based Channel Estimation and Feedback Network (CEFnet) is integrated by improving the robustness of estimation. Unlike conventional deep learning-based CSI compression methods like CsiNet, PFnet directly compresses pilot signals rather than CSI by reducing overhead. The proposed Adaptive Compression Scheme further optimizes feedback efficiency by dynamically adjusting the compression ratio based on the Signal-to-Noise Ratio (SNR). The proposed PFnet with Adaptive Compression Scheme achieved better results in terms of Normalized Mean Square Error (NMSE) when compared with existing PFnet.