Accurate Fifth Generation Mobile Network Coverage Prediction in Smart Cities with Machine Learning

This study uses machine learning in smart city environments to address challenges in predicting the Received Signal Reference Power (RSRP) performance in Fifth Generation (5G) networks. We applied and validated several established Machine Learning (ML) models to train extensive 5G drive test datasets representing urban and sub-urban environments in Malaysia. Drive tests were conducted around Putrajaya (urban) and UKM (sub-urban) areas to collect 5G Non-Standalone (NSA) datasets to develop a Random Forest-based machine learning model by integrating both 4G LTE and 5G network datasets and design a machine learning-based online coverage estimating application for 5G networks. This study is the second improved version of the Machine Learning-based Online Coverage Estimator (MLOEv2), which was developed with a MATLAB-based graphical user interface facilitating online RSRP predictions for teaching and learning purposes. This study aids in understanding 5G coverage in Malaysia, laying a foundation for practical mobile network deployment.

I. Introduction

As the 5G mobile networks rapidly evolve, deploying mobile network infrastructure must navigate a complex environment in a smart city where signal propagation, interference, and resource allocation become increasingly intricate due to complex and dense building structures [1]. Therefore, enhancing the transformative power of 5G is crucial to overcoming these challenges, and this is where the implementation of Artificial Intelligence (AI) comes into play.