Radio Environment Map Construction Based on Graph Convolutional Network

Constructing radio environment maps(REMs) from spatial sparse measurements finds many applications in the field of wireless communication, including network planning, power control, dynamic spectrum access and emitter localization to name a few. However, it is still a challenging task to construct a high-precision radio environment map, especially in urban environments or scenarios with multiple emitters under few sampling nodes conditions. In this paper, a method for constructing high-precision REMs based on graph convolutional network regressor is proposed, which can extract patterns between electromagnetic data while considering spatial correlation. The comparative simulation experiments with benchmark methods in both urban and open areas show the effectiveness of the method proposed in this paper.