UrMap: UAV-assisted Spatio-Temporal Radio Mapping using Sparse Cellular Samples
Journal
IEEE International Conference on Communications
Start Page
1
End Page
6
ISBN
[9798319542090]
Date Issued
2026-05-24
Author(s)
Abstract
The rapid development of Unmanned Aerial Vehicles (UAVs) and wireless communication technologies has fueled the rise of the low-altitude economy, supporting applications such as logistics delivery, precision agriculture, and disaster response. Reliable UAV communication and navigation rely on accurate radio maps that characterize the spatial distribution of signal strength and channel quality. However, constructing such radio maps is challenging, as extensive data collection in complex 3D environments is constrained by flight regulations and energy limits. In addition, pre-constructed radio maps may degrade over time due to environmental dynamics. To address these challenges, this work presents UrMap, a radio map estimation system based on multi-task spatio-temporal modeling. In UrMap, a Gaussian Process Regression (GPR) model is developed to jointly model spatial and temporal correlations, enabling the generation of complete radio maps that remain robust under time-varying environmental conditions. Furthermore, the proposed GPR model generates diverse connectivity maps by exploiting their cross-modality correlations, thereby providing a comprehensive understanding of the radio environment. Extensive experiments show that UrMap outperforms baseline methods, achieving a Mean Absolute Error (MAE) of 1.1 dB in the best case and maintaining 4.2 dB even in the worst case for radio map estimation.
Event(s)
2026 IEEE International Conference on Communications, ICC 2026
Subjects
Gaussian process regression
multi-task learning
radio map estimation
spatio-temporal modeling
UAV communications
Publisher
IEEE
Type
conference paper