Artificial Intelligence Enabled Joint Channel Estimation and Signal Detection for Massive MIMO Systems

Effective joint channel estimation and signal detection in dynamic environments has always been a significant challenge in massive multiple-input multiple-output (MIMO) systems. This paper presents a comprehensive review of state-of-the-art methods, including data-driven approaches like deep neural networks, convolutional neural networks, and recurrent neural networks, which exhibit exceptional adaptability and feature extraction capabilities. However, these methods often require extensive training datasets and substantial computational resources. Model-driven techniques, on the other hand, integrate domain knowledge with deep learning and offer improved interpretability and computational efficiency but depend heavily on the accuracy of prior models. Recently, diffusion models (DM) have emerged as a promising solution, addressing traditional limitations through adaptive noise modeling and stochastic sampling techniques. Moreover, based on DM, a novel algorithm named as enhanced diffusion models is also proposed in this paper. By incorporating precondition matrix optimization, dynamic temperature setting, adaptive momentum, and learnable step sizes, it achieves enhanced performance over existing methods. Finally, this paper discusses the key challenges and future research directions, emphasizing the need for interpretable models, generalization to dynamic environments, and lightweight frameworks for practical deployment.