GS3D: Signal Recovery in 3D Wireless Networks With Gaussian Splatting-Synergized Stable Diffusion

Three-dimensional (3D) wireless networks, that integrate terrestrial base stations (BSs), aerial platforms, low earth orbit (LEO), and geostationary (GEO) satellites, can offer global coverage, enabling ubiquitous and reliable communication in sixth-generation (6G) wireless networks. Despite this potential, one key challenge in such networks is managing interference from densely deployed nodes at varying altitudes, which can significantly degrade the overall performance and disrupt seamless communication. To address this challenge, in this work, we propose a novel Gaussian splatting-synergized stable Diffusion (GS3D) model to perform signal recovery in downlink 3D wireless networks. Specifically, we consider a scenario, where LEO satellite constellation (SatCon) serves earth stations in motion (ESIMs), while terrestrial BSs, aerial platforms, including unmanned aerial vehicles (UAVs), high altitude platforms (HAPs), and GEO satellites introduce interference to these ESIMs. Our results show that the proposed GS3D model can significantly improve the signal recovery accuracy compared to the state-of-the-art denoising diffusion probabilistic model (DDPM) and the regularized zero-forcing (RZF) scheme.