Estimation of the coverage area with ResNet-based Conditional Variational Autoencoder

Abstract

The rapid expansion of wireless communication networks, especially with 5G, has made the accurate and fast determination of base station locations critical. With the increase in network traffic and the number of users, it has become important to optimize the network infrastructure. This process is complex, taking into account technical and environmental factors. The study investigates machine learning (ML) and deep learning (DL) techniques to optimize base station placement, emphasizing that traditional methods require manual calculations and field measurements, whereas ML and DL-based approaches are more efficient and faster. Furthermore, a ResNet-based Conditional Variational Autocoder (ResNet-CVO) model for coverage map estimation is proposed and its performance is evaluated. With the proposed generative model, a more efficient approach for coverage map estimation is presented.