A Low-Overhead Incorporation-Extrapolation based Few-Shot CSI Feedback Framework for Massive MIMO Systems

TL;DR

The paper tackles the challenge of prohibitive CSI feedback overhead in FDD massive MIMO-OFDM systems by introducing IEFSF, a low-overhead framework that combines an incorporation-extrapolation scheme with few-shot data augmentation. It forms a low-dimensional eigenvector matrix $\overline{W}$ via group-wise subcarrier incorporation, then uses a Transformer-based compressor/reconstructor and a lightweight frequency extrapolation network to recover the full eigenvector matrix $W$, significantly reducing feedback and computation. To enable few-shot learning, two augmentation pathways are proposed: KDDA leverages frequency-domain correlations to synthesize additional $\overline{W}$ samples, and EGAN-based AIGC augments data by learning distributions at both low- and full-dimensional scales, trained with Wasserstein distance and gradient penalties. Extensive DeepMIMO-based simulations show IEFSF achieving high CSI feedback accuracy with only hundreds of collected samples and up to $64\times$ lower feedback overhead than baselines, validating its practicality for indoor and outdoor scenarios and its potential for real-world deployment.

Abstract

Accurate channel state information (CSI) is essential for downlink precoding in frequency division duplexing (FDD) massive multiple-input multiple-output (MIMO) systems with orthogonal frequency-division multiplexing (OFDM). However, obtaining CSI through feedback from the user equipment (UE) becomes challenging with the increasing scale of antennas and subcarriers and leads to extremely high CSI feedback overhead. Deep learning-based methods have emerged for compressing CSI but these methods generally require substantial collected samples and thus pose practical challenges. Moreover, existing deep learning methods also suffer from dramatically growing feedback overhead owing to their focus on full-dimensional CSI feedback. To address these issues, we propose a low-overhead Incorporation-Extrapolation based Few-Shot CSI feedback Framework (IEFSF) for massive MIMO systems. An incorporation-extrapolation scheme for eigenvector-based CSI feedback is proposed to reduce the feedback overhead. Then, to alleviate the necessity of extensive collected samples and enable few-shot CSI feedback, we further propose a knowledge-driven data augmentation (KDDA) method and an artificial intelligence-generated content (AIGC) -based data augmentation method by exploiting the domain knowledge of wireless channels and by exploiting a novel generative model, respectively. Experimental results based on the DeepMIMO dataset demonstrate that the proposed IEFSF significantly reduces CSI feedback overhead by 64 times compared with existing methods while maintaining higher feedback accuracy using only several hundred collected samples.

Figures (13)

• Figure 1: The incorporation-extrapolation CSI feedback scheme which consists of the incorporation process, the proposed Transformer-based CSI compression and reconstruction network, and the frequency extrapolation network.
• Figure 2: An illustration of the knowledge-driven data augmentation method ($N_{gr}=4$ and $N_{sgr}=2$). scs: subcarriers.
• Figure 3: The diagram depicting the connection between the knowledge-driven data augmentation method and the AIGC-based data augmentation method.
• Figure 4: The network architecture of the EGAN. (a) The overall network architecture of the EGAN. The sizes of key tensors and neural block outputs are given near them. (b) Detailed architectures and configurations of neural blocks constituting the EGAN. The detailed architecture and configuration of the frequency extrapolation network have been given in Fig. \ref{['NN']} and are therefore neglected here.
• Figure 5: Top views of the indoor scenario and the outdoor scenario.