High-Efficiency Urban 3D Radio Map Estimation Based on Sparse Measurements
Xinwei Chen, Xiaofeng Zhong, Zijian Zhang, Linglong Dai, Shidong Zhou
TL;DR
The paper tackles the difficulty of generating accurate 3D urban radio maps from sparse UAV data by leveraging a Gaussian Process Regression (GPR) framework that models the slow-fading residual $r(\mathbf{x}) = \gamma(\mathbf{p}) - \hat{\gamma}(\mathbf{p})$ with input $\mathbf{x}=[p_x,p_y,p_z,\hat{\gamma}(\mathbf{p})]$. It introduces a composite kernel $k_{\text{com}}(\mathbf{x},\mathbf{x'}) = k_{\text{const}}(\mathbf{x},\mathbf{x'}) \times k_{\text{Matérn}}(\mathbf{x},\mathbf{x'}) + k_{\text{WN}}(\mathbf{x},\mathbf{x'})$ and derives the GP posterior to estimate unknown RSRP across the 3D space. Two measurement-point selection strategies are proposed: online MAP-based variance maximization and offline clustering-based selection, both aimed at reducing data requirements. On a real urban dataset spanning $4.2\times 10^{6}\ \mathrm{m^3}$ with $4274$ samples, the method achieves RMSE improvements of over $2.5\ \mathrm{dB}$ and can recover full 3D maps with only $2\%$ of UAV measurements. This work enables data-efficient, high-fidelity 3D radio mapping for UAV-enabled urban operations and informs future 3D radio-map research.
Abstract
Recent widespread applications for unmanned aerial vehicles (UAVs) -- from infrastructure inspection to urban logistics -- have prompted an urgent need for high-accuracy three-dimensional (3D) radio maps. However, existing methods designed for two-dimensional radio maps face challenges of high measurement costs and limited data availability when extended to 3D scenarios. To tackle these challenges, we first build a real-world large-scale 3D radio map dataset, covering over 4.2 million m^3 and over 4 thousand data points in complex urban environments. We propose a Gaussian Process Regression-based scheme for 3D radio map estimation, allowing us to realize more accurate map recovery with a lower RMSE than state-of-the-art schemes by over 2.5 dB. To further enhance data efficiency, we propose two methods for training point selection, including an offline clustering-based method and an online maximum a posterior (MAP)-based method. Extensive experiments demonstrate that the proposed scheme not only achieves full-map recovery with only 2% of UAV measurements, but also sheds light on future studies on 3D radio maps.