Beamforming Inferring by Conditional WGAN-GP for Holographic Antenna Arrays

TL;DR

This work tackles the high overhead of beamforming for holographic MIMO by inferring full high‑dimensional beamforming matrices from limited CSI using a conditional WGAN‑GP framework. The generator learns to map low‑dimensional channel/beamforming inputs to $V^{high}$ while the discriminator enforces alignment with the real high‑dimensional distribution through Wasserstein distance and gradient penalties, with an auxiliary $L_2$ loss for stability. Training leverages ground-truth $H^{real}$ and $V^{real}$ obtained via WMMSE on WAIR‑D data, and prediction operates on scarce CSI to yield real‑time beamforming with over 50% overhead reduction. The results show comparable spectral efficiency to WMMSE and notable speedups, highlighting the practicality of holographic antenna arrays under reduced CSI requirements.

Abstract

The beamforming technology with large holographic antenna arrays is one of the key enablers for the next generation of wireless systems, which can significantly improve the spectral efficiency. However, the deployment of large antenna arrays implies high algorithm complexity and resource overhead at both receiver and transmitter ends. To address this issue, advanced technologies such as artificial intelligence have been developed to reduce beamforming overhead. Intuitively, if we can implement the near-optimal beamforming only using a tiny subset of the all channel information, the overhead for channel estimation and beamforming would be reduced significantly compared with the traditional beamforming methods that usually need full channel information and the inversion of large dimensional matrix. In light of this idea, we propose a novel scheme that utilizes Wasserstein generative adversarial network with gradient penalty to infer the full beamforming matrices based on very little of channel information. Simulation results confirm that it can accomplish comparable performance with the weighted minimum mean-square error algorithm, while reducing the overhead by over 50%.

Figures (8)

• Figure 1: The diagram shows the common system model. The function $\Phi_{\mathcal{M}_1 \rightarrow \mathcal{M}_2}(\cdot)$ transforms the low-dimensional beamforming matrix from the antenna set $\mathcal{M}_1$ to the high-dimensional beamforming matrix at $\mathcal{M}_2$.
• Figure 2: Network architecture of the proposed conditional WGAN-GP approach.
• Figure 3: Training and prediction stages.
• Figure 4: Top view of the scenario-2 case 1 of the WAIR-D. (a) shows 100 users among the total 10000 users, (b) shows all the 10000 users.
• Figure 5: Pseudo-color plots for real components of (a) low-dimensional beamforming matrix $\textcolor{black}{\mathbf{V}^{low}}$, (b) target high-dimensional beamforming matrix $\textcolor{black}{\mathbf{V}^{real}}$ and (c) predicted high-dimensional beamforming matrix $\textcolor{black}{\mathbf{V}^{high}}$.