Indoor Path Loss Modeling for 5G Communications in Smart Factory Scenarios Based on Meta-Learning

Abstract

Millimeter waves (mmWaves) of the 28 GHz frequency bands have been selected for the 5G communications with special usage scenarios such as smart factories. Indoor path loss prediction plays an important role in configuring a base station to be able to utilize the full capacity of the new technology. Although machine learning has attracted much attention recently in path loss modeling thanks to its ability to make accurate predictions, its performance can be limited by the size of available measurement data set used for training. In this paper, we propose a new training strategy to train path loss models based on convolutional neural network (CNN). The proposed strategy is based on meta-learning which performs well in few-shot learning scenarios with multiple tasks comprising a meta-task. It is shown that the indoor path loss model based on a CNN configured as a metatask of multiple beams can outperform the CNN models by a conventional training algorithm as well as empirical models.