Search papers, labs, and topics across Lattice.
This paper introduces point-cloud-assisted tangent Gaussian splatting (PC-TGS), a novel framework that extrapolates angular power spectrum (APS) predictions to unmeasured locations by integrating sparse radio measurements with dense LiDAR-based geometry. By modeling environmental scatterers as anisotropic 3D Gaussians and employing a depth-aware electromagnetic splatting process, PC-TGS significantly enhances the accuracy of channel predictions in large-scale environments. Evaluations on a city-scale dataset reveal that PC-TGS outperforms existing methods in both APS and reference signal received power (RSRP) prediction while ensuring faster inference times.
PC-TGS can predict channel characteristics in unmeasured locations, transforming the landscape of wireless network optimization.
Accurate, site-specific channel information is crucial for optimizing next-generation wireless networks. Among various approaches, localized statistical channel modeling (LSCM), which models the channel multipath angular power spectrum (APS) from the reference signal received power (RSRP) measurement, has emerged as a state-of-the-art method tailored for efficient network optimization. However, despite its effectiveness, LSCM cannot predict APS at the vast majority of locations where no measurements are available, which significantly restricts its applicability in large-scale, real-world scenarios. To address this challenge, we present \emph{point-cloud-assisted tangent Gaussian splatting} (PC-TGS), the first framework to \emph{extrapolate} APS to unmeasured outdoor grids by integrating sparse radio measurements with dense LiDAR-based geometry. PC-TGS represents environmental scatterers as anisotropic 3D Gaussians, initialized and refined through a relaxed-mean reparameterization of the raw point cloud. A tangent-plane projection accurately maps each Gaussian into the local angular domain, while a depth-aware electromagnetic splatting process aggregates their contributions. To ensure practical deployment, we derive a closed-form Gaussian-weighted average (GWA) for APS bin integration and provide a provable error bound. { Evaluations on a LiDAR-scanned city-scale dataset (5M points, 6,310 RSRP samples) demonstrate that PC-TGS achieves better APS and RSRP prediction performance compared to state-of-the-art baselines and faster inference time for APS extrapolation task. These results highlight the potential of PC-TGS to enable geometry-aware and data-efficient channel prediction in large-scale wireless digital twins.
