Path Loss Map Construction Based on RadioResUNet
Wenhao Du, Wei Gong, Li Li, Zhenzhen Jiao, Jian Xun Jin · 2023
Propagation path loss prediction is critical in wireless communication, playing a significant role in wireless network planning, signal transmission optimization, and wireless coverage evaluation. However, traditional path loss prediction methods often rely on complex physical models and environmental parameters, leading to time-consuming and computationally intensive calculations. Several machine learning methods have been proposed to predict path loss, but their accuracy still needs improvement. To address these issues, this paper proposes a novel machine learning method called RadioResUNet, which is an improved version of ResUNet. RadioResUnet incorporates progressive multilevel encoder, attention mechanism, depthwise separable convolution, and fine-grained downsampling into ResUnet to adapt to the path loss prediction problem. Compared to traditional path loss prediction methods, our method directly predicts path loss from the input (city map) by learning map features and spatial relationships, reducing the reliance on complex physical models and environmental parameters, simplifying the computation process, and enhancing path loss prediction accuracy. Through experiments on the RadioMapSeer dataset, we demonstrate that our proposed method outperforms the existing methods in terms of prediction accuracy.