ACT-WTGAN: A Dynamic Radio Map Repair Method for UAVs Via Spatial-Frequency Domain Fusion

Abstract

Radio maps are used to describe the distribution of signal strength in wireless environments and play a crucial role in spectrum management and the optimization of communication efficiency, especially in the emerging 6G networks. Traditional methods struggle to handle the complex changes posed by dynamic obstacles such as Unmanned Aerial Vehicles, and highly complex networks are not suitable for platforms with limited computing capabilities, such as drones. This paper proposes a method for constructing dynamic radio maps in urban environments suitable for drone deployment (ACT-WTGAN). ACT-WTGAN addresses these challenges by only repairing the areas affected by dynamic obstacles and using the static radio map as a benchmark. This method includes an aggregation context transformation module to capture context features at different spatial distances, thereby improving the accuracy and rationality of the generated content. A frequency domain convolution module is also introduced to effectively capture high-frequency details and ensure the generation of fine texture features in the DRM. Experimental results show that ACT-WTGAN significantly outperforms existing advanced methods in terms of accuracy, structural similarity, and peak signal-to-noise ratio, highlighting its potential application value in future 6G wireless communication networks.