What is it about?
Getting a strong 5G signal inside tall buildings is difficult because modern construction materials, such as reinforced concrete and metal-coated glass, induce severe signal attenuation. Typically, mapping this indoor network coverage requires labor-intensive manual walk tests. This study introduces a new method that uses Unmanned Aerial Vehicles (UAVs) to measure radio frequency signals from the exterior of a building. Using an artificial intelligence framework called a Dynamic Graph Attention Network (DGAT), the system takes these exterior measurements and accurately predicts the indoor signal strength (RSSI) and channel quality (CQI). It achieves this by learning a connectivity structure that captures the non-linear way signals propagate around architectural barriers, rather than relying on simple straight-line distances. The system then generates interactive 3D maps to visually highlight coverage gaps and dead zones.
Featured Image
Photo by Sylwia Bartyzel on Unsplash
Why is it important?
Over 80% of mobile data consumption occurs indoors, making consistent in-building coverage essential for 5G and future 6G networks. This framework solves a major challenge by bridging the gap between non-invasive exterior aerial sensing and accurate interior coverage estimation. By effectively reconstructing complete 3D indoor coverage maps from sparse external data, it minimizes the need for exhaustive manual testing. This provides network engineers and urban planners with an actionable, robust foundation for designing next-generation wireless infrastructure.
Perspectives
Standard geometric models and distance-based graphs often fail in complex indoor environments because architectural barriers create unpredictable signal paths. Our approach abandons static geometric assumptions in favor of learned topological associations. By optimizing the model to understand these latent propagation dynamics, we achieved a 76% reduction in Root Mean Square Error (RMSE) for RSSI and an 89% reduction for CQI compared to existing location-aware baselines.
David HASON RUDD
University of Technology Sydney
Read the Original
This page is a summary of: Spatial-Temporal Inference of Indoor RSSI and CQI for 3D Coverage Mapping using Dynamic Graph Attention Networks, May 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3774905.3795607.
You can read the full text:
Contributors
The following have contributed to this page
