Group Equivariant Convolutional Networks for Pathloss Estimation

TL;DR

This paper introduces RadioGUNet, a group equivariant UNet-based deep learning model that improves pathloss estimation accuracy in wireless communications by leveraging symmetries like rotations and reflections.

Contribution

It demonstrates how to extend UNet models with group equivariant convolutions and shows this extension enhances pathloss estimation performance.

Findings

Extended UNet models with group equivariance outperform standard UNet by up to 0.41 dB.

Group equivariant convolutions enable better generalization without data augmentation.

The approach is validated on the RadioMapSeer dataset.

Abstract

This paper presents RadioGUNet, a UNet-based deep learning framework for pathloss estimation in wireless communication. Unlike other frameworks, it leverages group equivariant convolutional networks, which are known to increase the expressive capacity of a neural network by allowing the model to generalize to further classes of symmetries, such as rotations and reflections, without the need for data augmentation or data pre-processing. The results of this work are twofold. First, we show that typical UNet-based convolutional models can be easily extended to support group equivariant convolution (g-conv). Secondly, we show that the task of pathloss estimation benefits from such an extension, as the proposed extended model outperforms typical UNet-based models by up to 0.41 dB for a similar number of parameters in the RadioMapSeer dataset. The code is publicly available on the GitHub page: https://github.com/EricssonResearch/radiogunet