Deep Learning-Empowered Movable-Antenna Position Optimization with Partial CSI

2026-06-16

AI Summary

This work addresses the challenge of optimizing reconfigurable antenna positions to enhance wireless data rates, which traditionally relies on full channel state information (CSI) and incurs substantial estimation overhead. To overcome this limitation, the authors propose deep learning–based approaches: in single-user scenarios, a supervised learning framework learns the nonlinear mapping between antenna positions and performance using only local received power measurements; in multiuser MISO systems, they introduce a novel unsupervised attention mechanism that directly maximizes the sum rate without requiring explicit CSI or ground-truth optimal labels. Experimental results demonstrate that the proposed method nearly achieves the theoretical optimum in single-user settings and significantly outperforms conventional CSI-based alternating optimization algorithms in multiuser environments.

Abstract

Movable antennas (MAs) are a promising technology to improve wireless data rates by dynamically adjusting their positions to avoid deep fading. However, finding the optimal MA positions requires full channel state information (CSI) for all possible locations within the movement region, creating massive channel estimation overhead. This paper proposes a deep neural network (DNN)-based learning framework to predict the optimal positions of multiple transmit MAs in a multi-user multiple-input single-output (MISO) system, entirely bypassing explicit channel estimation.First, we analyze a single-user MISO case, revealing a complex, highly nonlinear mapping between the optimal MA positions and the channel power gains from a specific subset of locations in the transmit region to the user. Because this mapping cannot be mathematically characterized for practical channel models, we train a DNN via supervised learning to capture it. The pre-trained DNN can then determine optimized MA positions in real-time relying only on partial power measurements from the transmit region.Extending this to multi-user scenarios is challenging due to complex rate expressions and the lack of globally optimal position solutions to use as training labels. To overcome this, we develop an unsupervised training framework that directly maximizes the multi-user sum-rate. This framework utilizes an attention-based architecture to extract latent features from the partial channel measurements and effectively manage inter-user interference. Simulation results show that our proposed approach achieves near-optimal performance in single-user systems and surpasses conventional CSI-based alternating optimization algorithms in multi-user environments.

Problem

• movable antennas
• position optimization
• partial CSI
• channel state information
• wireless communication

Innovation

• movable antennas
• deep learning
• partial CSI
• unsupervised learning
• attention mechanism