Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Data-Driven Radio Environment Map Estimation Using Graph Neural Networks

Ali Shibli, Tahar Zanouda

TL;DR

Problem: estimate Radio Environment Maps (REMs) in telecom networks using sparse geo-located measurements. Approach: a Graph Convolutional Network (GCN) framework over a temporal graph of hexagonal tiles (via H3) that fuses measurements with physical cell information to predict REMs, capable of regression ($R^2$) and classification. Contributions: (i) a GCN-based REM estimation architecture, (ii) comprehensive comparisons against XGBoost, FCN, and TabNet showing superior performance, (iii) a data fusion pipeline integrating measurements, cell parameters, and geographic context, and (iv) validation on multi-city data with temporal dynamics. Findings: the GCN achieved $R^2 = 0.83$ for regression and $0.92$ accuracy for classification, demonstrating the importance of leveraging spatial relationships in REMs. Impact: enables more accurate REM construction for network configuration, spectrum estimation, and proactive resource management in dynamic telecom networks, with potential for incorporating additional modalities in future work.

Abstract

Radio Environment Maps (REMs) are crucial for numerous applications in Telecom. The construction of accurate Radio Environment Maps (REMs) has become an important and challenging topic in recent decades. In this paper, we present a method to estimate REMs using Graph Neural Networks. This approach utilizes both physical cell information and sparse geo-located signal strength measurements to estimate REMs. The method first divides and encodes mobile network coverage areas into a graph. Then, it inputs sparse geo-located signal strength measurements, characterized by Reference Signal Received Power (RSRP) and Reference Signal Received Quality (RSRQ) metrics, into a Graph Neural Network Model to estimate REMs. The proposed architecture inherits the advantages of a Graph Neural Network to capture the spatial dependencies of network-wide coverage in contrast with network Radio Access Network node locations and spatial proximity of known measurements.

Data-Driven Radio Environment Map Estimation Using Graph Neural Networks

TL;DR

Problem: estimate Radio Environment Maps (REMs) in telecom networks using sparse geo-located measurements. Approach: a Graph Convolutional Network (GCN) framework over a temporal graph of hexagonal tiles (via H3) that fuses measurements with physical cell information to predict REMs, capable of regression () and classification. Contributions: (i) a GCN-based REM estimation architecture, (ii) comprehensive comparisons against XGBoost, FCN, and TabNet showing superior performance, (iii) a data fusion pipeline integrating measurements, cell parameters, and geographic context, and (iv) validation on multi-city data with temporal dynamics. Findings: the GCN achieved for regression and accuracy for classification, demonstrating the importance of leveraging spatial relationships in REMs. Impact: enables more accurate REM construction for network configuration, spectrum estimation, and proactive resource management in dynamic telecom networks, with potential for incorporating additional modalities in future work.

Abstract

Radio Environment Maps (REMs) are crucial for numerous applications in Telecom. The construction of accurate Radio Environment Maps (REMs) has become an important and challenging topic in recent decades. In this paper, we present a method to estimate REMs using Graph Neural Networks. This approach utilizes both physical cell information and sparse geo-located signal strength measurements to estimate REMs. The method first divides and encodes mobile network coverage areas into a graph. Then, it inputs sparse geo-located signal strength measurements, characterized by Reference Signal Received Power (RSRP) and Reference Signal Received Quality (RSRQ) metrics, into a Graph Neural Network Model to estimate REMs. The proposed architecture inherits the advantages of a Graph Neural Network to capture the spatial dependencies of network-wide coverage in contrast with network Radio Access Network node locations and spatial proximity of known measurements.
Paper Structure (16 sections, 3 equations, 3 figures, 3 tables, 1 algorithm)

This paper contains 16 sections, 3 equations, 3 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Previous work
  3. Data
  4. Method
  5. Problem Definition
  6. Method Overview
  7. Overview of the process
  8. Data pre-processing
  9. Coverage Map Representation using a Temporal Graph
  10. Feature Engineering
  11. Coverage Map Estimation Model using Graph Neural Networks
  12. Experiments
  13. Evaluation Metrics
  14. Regression Task
  15. Classification Task
  16. ...and 1 more sections

Figures (3)

  • Figure 1: A figure showing a coverage area, divided in hexagons, where UEs receive signal from cells
  • Figure 2: Representing Coverage Areas using H3 Hexagons
  • Figure 3: REMs estimation using GCN model.