Diffusion Model-Based Virtual 3D City Augmentation for EM-Wave Propagation Analysis

We propose a method to generate a 3D model of virtual cities for electromagnetic-wave propagation data generation by ray tracing. In constructing propagation estimation models using machine learning, insufficient or biased training data may worsen the estimation performance. However, in cases where simulation-generated data are used, this effect could be suppressed if we prepare a variety of 3D models and generate propagation data. Therefore, the objective of this study is to generate a map of naturally occurring building layouts using an image generation model, and to create a 3D model on the basis of this map to ensure a larger number of practical propagation environment models.