3D-RadioDiff: An Altitude-Conditioned Diffusion Model for 3D Radio Map Construction

Abstract

Radio Maps (RMs) play a pivotal role in optimizing communication networks by providing critical insights into signal propagation and coverage. However, existing RM construction methods are predominantly tailored for the signal coverage of terrestrial networks, which significantly limits their applicability in emerging applications with aerial nodes. To address this limitation, this letter proposes 3D-RM Diffusion (3D-RadioDiff), which incorporates the altitude of Autonomous aerial vehicles (AAVs) and environmental topology as key conditioning factors, effectively extending the RMs construction from the ground to 3D space. Numerical evaluations demonstrate the feasibility of the proposed 3D-RadioDiff in generating RMs at different altitudes without received signal strength (RSS) sampling under different scenarios. We also show that the construction accuracy of RMs improves as the number of denoising steps increases.