Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CommUNext: Deep Learning-Based Cross-Band and Multi-Directional Signal Prediction

Chi-Jui Sung, Fan-Hao Lin, Tzu-Hao Huang, Chu-Hsiang Huang, Hui Chen, Chao-Kai Wen, Henk Wymeersch

TL;DR

6G networks require cross-band cognition across FR1–FR3, but physics-based ray tracing and inter-frequency measurement gaps impede scalable planning and operation. CommUNext presents a data-driven framework that predicts high-frequency ($7$ GHz), multi-directional signal-strength maps from low-frequency coverage maps and sparse high-frequency measurements, formalized via a Bayes-optimal predictor $f^*(\mathcal{X})=\mathbb{E}[\mathcal{Y}|\mathcal{X}]$. It introduces Full CommUNext and Partial CommUNext to handle complete and incomplete reference-band data, respectively, leveraging an automated data-generation pipeline (OSM/Blender/Sionna) to produce aligned $128\times128$ maps and LoS/NLoS priors. Across extensive RT-based experiments, the Seg architecture improves accuracy, Partial CommUNext demonstrates robust reconstruction under incomplete coverage with tunable sampling strategies, and the approach promises scalable, real-time cross-band beam prediction and resource management for 6G deployments.

Abstract

Sixth-generation (6G) networks are envisioned to achieve full-band cognition by jointly utilizing spectrum resources from Frequency Range~1 (FR1) to Frequency Range~3 (FR3, 7--24\,GHz). Realizing this vision faces two challenges. First, physics-based ray tracing (RT), the standard tool for network planning and coverage modeling, becomes computationally prohibitive for multi-band and multi-directional analysis over large areas. Second, current 5G systems rely on inter-frequency measurement gaps for carrier aggregation and beam management, which reduce throughput, increase latency, and scale poorly as bands and beams proliferate. These limitations motivate a data-driven approach to infer high-frequency characteristics from low-frequency observations. This work proposes CommUNext, a unified deep learning framework for cross-band, multi-directional signal strength (SS) prediction. The framework leverages low-frequency coverage data and crowd-aided partial measurements at the target band to generate high-fidelity FR3 predictions. Two complementary architectures are introduced: Full CommUNext, which substitutes costly RT simulations for large-scale offline modeling, and Partial CommUNext, which reconstructs incomplete low-frequency maps to mitigate measurement gaps in real-time operation. Experimental results show that CommUNext delivers accurate and robust high-frequency SS prediction even with sparse supervision, substantially reducing both simulation and measurement overhead.

CommUNext: Deep Learning-Based Cross-Band and Multi-Directional Signal Prediction

TL;DR

6G networks require cross-band cognition across FR1–FR3, but physics-based ray tracing and inter-frequency measurement gaps impede scalable planning and operation. CommUNext presents a data-driven framework that predicts high-frequency ( GHz), multi-directional signal-strength maps from low-frequency coverage maps and sparse high-frequency measurements, formalized via a Bayes-optimal predictor . It introduces Full CommUNext and Partial CommUNext to handle complete and incomplete reference-band data, respectively, leveraging an automated data-generation pipeline (OSM/Blender/Sionna) to produce aligned maps and LoS/NLoS priors. Across extensive RT-based experiments, the Seg architecture improves accuracy, Partial CommUNext demonstrates robust reconstruction under incomplete coverage with tunable sampling strategies, and the approach promises scalable, real-time cross-band beam prediction and resource management for 6G deployments.

Abstract

Sixth-generation (6G) networks are envisioned to achieve full-band cognition by jointly utilizing spectrum resources from Frequency Range~1 (FR1) to Frequency Range~3 (FR3, 7--24\,GHz). Realizing this vision faces two challenges. First, physics-based ray tracing (RT), the standard tool for network planning and coverage modeling, becomes computationally prohibitive for multi-band and multi-directional analysis over large areas. Second, current 5G systems rely on inter-frequency measurement gaps for carrier aggregation and beam management, which reduce throughput, increase latency, and scale poorly as bands and beams proliferate. These limitations motivate a data-driven approach to infer high-frequency characteristics from low-frequency observations. This work proposes CommUNext, a unified deep learning framework for cross-band, multi-directional signal strength (SS) prediction. The framework leverages low-frequency coverage data and crowd-aided partial measurements at the target band to generate high-fidelity FR3 predictions. Two complementary architectures are introduced: Full CommUNext, which substitutes costly RT simulations for large-scale offline modeling, and Partial CommUNext, which reconstructs incomplete low-frequency maps to mitigate measurement gaps in real-time operation. Experimental results show that CommUNext delivers accurate and robust high-frequency SS prediction even with sparse supervision, substantially reducing both simulation and measurement overhead.

Paper Structure

This paper contains 29 sections, 2 theorems, 17 equations, 11 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Data Generation and Preprocessing
  4. Automated Data Generation Toolchain
  5. OSM and Blender
  6. NVIDIA Sionna
  7. Data Generation
  8. Building Maps
  9. 3.5 GHz Coverage Maps
  10. 7 GHz Multi-Directional SS Maps
  11. Data Preprocessing
  12. Dataset Partitioning
  13. Sparse 7 GHz SS Maps
  14. LoS and NLoS Masks
  15. Sampled 3.5 GHz Maps
  16. ...and 14 more sections

Key Result

Theorem 1

Under the MSE criterion, the Bayes-optimal predictor of $\mathcal{Y}$ given $\mathcal{X}$ is

Figures (11)

  • Figure 1: Automated data generation toolchain integrating OSM, Blender, and Sionna.
  • Figure 2: Example of importing OSM building data into Blender.
  • Figure 3: Example of (a) a building map $\mathbf{B}$ and (b) the corresponding 3.5 GHz coverage map $\mathbf{S}_c$.
  • Figure 4: Eight generated 7 GHz SS maps $\{\mathbf{S}_{d1}, \mathbf{S}_{d2}, \dots, \mathbf{S}_{d8}\}$, showing beam concentration along pointing directions and variations caused by blockages.
  • Figure 5: (a) LoS and (b) NLoS masks corresponding to the building map in Fig. \ref{['fig:building_coverage']}(a).
  • ...and 6 more figures

Theorems & Definitions (3)

  • Theorem 1: Bayes-Optimal Predictor under MSE bishop2006pattern
  • Corollary 1: Incomplete Reference-Band Case
  • Remark 1: Structured Prior via NLoS Mask