Multi-Base Station Radio Map Prediction Based on Residual Enhanced Dual Path UNet

Radio maps are used for various applications such as resource allocation, interference control, fingerprint positioning, and network planning. With the development of 5G and subse-quent mobile communication technologies, establishing detailed radio maps has become crucial to effectively utilize limited spec-trum resources. Traditional methods for constructing radio maps typically rely on complex physical models and environmental parameters, resulting in time-consuming and computationally in-tensive processes. Although some machine learning methods have been proposed to construct radio maps, their accuracy still needs improvement. This paper proposes a novel radio map prediction method based on Residual Enhanced Dual Path UNet (RED-UNet), an improved version of RadioUNet. By adding residual connections and batch normalization after each convolutional layer, RED-UNet accelerates model convergence, enhances model generalization, promotes gradient flow, and mitigates gradient disappearance, thereby improving the accuracy of radio map prediction. Additionally, we utilize this method for predicting radio maps in areas of interest containing multiple base stations. Compared to traditional prediction methods, our method demonstrates better accuracy. Experimental results validate the effectiveness of our approach.