Channel Sensing and Vision-Aided Distributed Beamforming for 6G Ultra-Dense Networks

This paper presents a beamforming strategy designed to optimize distributed beamforming (DSBF) in 6G ultra-dense networks. DSBF is a key enabler for high-frequency communications by spatially coordinating transmissions from geographically distributed base stations (BSs), but its practical deployment faces critical challenges, most notably accurate BS selection and precise beam alignment across distributed nodes. To address these limitations, we propose a two-phase solution that tightly integrates advanced radio signal processing with computer vision techniques. In the first phase, multi-BS angle-delay channel power matrices (ADCPMs) are constructed under multipath propagation and processed using a Vision Transformer to estimate user equipment (UE) positions. Leveraging observations from multiple BSs provides a broader spatial context compared to relying on a single BS, where the environmental visibility is limited and spatial cues are often insufficient for accurate positioning. These estimates are then used to identify the nearest BSs for DSBF node selection. In the second phase, RGB-depth cameras mounted on selected BSs leverage computer vision techniques to perform direct UE detection and line-of-sight path identification. These outputs guide accurate downlink beamforming while avoiding the quantization loss and search complexity of codebook-based methods. This vision-assisted approach avoids quantization loss and complexity from codebook-based beam searches. By bridging radio signal processing with computer vision, the proposed system significantly improves DSBF performance, addressing the limitations of conventional codebook-based approaches.