Constructing Radio Maps for UAV Communications via Dynamic Resolution Virtual Obstacle Maps

Radio maps that characterize the air-to-ground communication channels play an important role in optimizing unmanned aerial vehicle (UAV) communications. Constructing a radio map is very challenging because a city may have a complicated building and vegetation topology that affects the air-to-ground radio propagation. Existing methods usually require a large amount of measurement data for training. This paper focuses on the small sample regime. A learning framework is developed to decompose the radio map into a structural component, which predicts the path loss via constructing a hidden virtual obstacle map, and a non-structural component, which captures the random scattering due to signal reflection and diffraction. It is found that constructing a virtual obstacle map with dynamic resolution improves the learning efficiency. This paper develops a simple grouping method that locally adjusts the map resolution according to the sample density and side information from a 2D street map. Numerical experiments show that the proposed method outperforms existing schemes in small to large sample regimes.