Deep Learning-Enabled CSI Estimation and Detection in Modern Multi-Antenna Systems

The remarkable success of deep learning (DL) techniques has established them as a promising approach for channel estimation and symbol detection in 6th-generation (6G) multi-antenna systems. These systems, including massive multiple-input multiple-output (mMIMO), reconfigurable intelligent surfaces (RIS)-supported mMIMO, and their advanced variants, pose diverse challenges and dependencies, in which estimation inaccuracies can significantly degrade performance. Traditional methods, such as least-squares (LS) and minimum mean-square error (MMSE) estimators, exhibit significant limitations, including the requirement for a channel covariance matrix and matrix inversion. Moreover, conventional techniques often achieve lower accuracy than DL-based methods, which can learn complex channel mappings, an advantage that makes DL-based methods promising for 6G systems. In the realm of symbol detection, 6G multi-antenna systems face several challenges, including high computational complexity and the escalating difficulty of managing signal interference. These factors can significantly degrade the performance of traditional detection methods, particularly in environments characterized by multiple interfering sources and noise. The data-driven and model-driven DL-aided estimation and detection methods have unlocked a new research paradigm that is now under rigorous investigation by academia and industry. In this paper, we provide a comprehensive review of recent advancements in DL-aided channel estimation and detection for multi-antenna systems, with a specific focus on massive MIMO and RIS-enabled massive MIMO systems. We conclude by discussing the challenges inherent in DL-assisted channel estimation and detection, and propose potential future directions to improve efficiency and accuracy.

IEEE Open Journal of the Communications Society

IEEE Open Journal of the Communications Society

N/A

2644-125X

N/A

N/A

38