Model Transmission-Based Online Updating Approach for Massive MIMO CSI Feedback
Boyuan Zhang, Haozhen Li, Xin Liang, Xinyu Gu, Lin Zhang
Abstract
Deep learning has been widely applied in frequency division duplexing (FDD) massive multiple-input multiple-output (MIMO) systems to achieve the accurate and effective channel state information (CSI) feedback. The challenges of the network models applied in real-world systems has also obtained more attention, especially the problem of generalization. In this letter, a model transmission-based online updating approach is proposed to achieve the real-time adaptation of the model and solve the problem of model mismatch when unseen data occurs. The feedback model will be updated using the real-time data with limited overhead, and the updating procedure is designed considering the feedback accuracy, requirement on training data, and storage usage. Experimental results indicate that the proposed approach can achieve quick model adjustment in changing scenarios and achieve comprehensive accuracy with limited training cost and storage usage, contributing to the feasibility of AI-based schemes.