GSBF: 3D Gaussian Splatting for Environment-Aware Beamforming

A new arXiv preprint (2608.05896) introduces GSBF, a 3D Gaussian splatting pipeline for environment-aware beamforming in MIMO communication systems. The method leverages multi-modal data to create a persistent 3D Gaussian representation of the environment, modeling scattering with reciprocity-preserving bidirectional spherical Gaussian (Bi-SG) kernels. Two-sided electromagnetic rasterization renders an angular propagator map, which is aggregated through an over-complete array-manifold dictionary and projected to constant-modulus beamformers. This approach aims to reduce pilot overhead and computational complexity compared to conventional beamforming that requires accurate instantaneous CSI and iterative optimization. The work is based on the principle that radio propagation is governed by physical geometry. The paper is available on arXiv.