Potential of machine learning/Artificial Intelligence (ML/AI) for verifying configurations of 5G multi Radio Access Technology (RAT) base station
Imran Ullah, Muhammad (2023-07-04)
Imran Ullah, Muhammad
M. Imran Ullah
04.07.2023
© 2023 Muhammad Imran Ullah. Ellei toisin mainita, uudelleenkäyttö on sallittu Creative Commons Attribution 4.0 International (CC-BY 4.0) -lisenssillä (https://creativecommons.org/licenses/by/4.0/). Uudelleenkäyttö on sallittua edellyttäen, että lähde mainitaan asianmukaisesti ja mahdolliset muutokset merkitään. Sellaisten osien käyttö tai jäljentäminen, jotka eivät ole tekijän tai tekijöiden omaisuutta, saattaa edellyttää lupaa suoraan asianomaisilta oikeudenhaltijoilta.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:oulu-202307042836
https://urn.fi/URN:NBN:fi:oulu-202307042836
Tiivistelmä
The enhancements in mobile networks from 1G to 5G have greatly increased data transmission reliability and speed. However, concerns with 5G must be addressed. As system performance and reliability improve, ML and AI integration in products and services become more common. The integration teams in cellular network equipment creation test devices from beginning to end to ensure hardware and software parts function correctly. Radio unit integration is typically the first integration phase, where the radio is tested independently without additional network components like the BBU and UE. 5G architecture and the technology that it is using are explained further. The architecture defined by 3GPP for 5G differs from previous generations, using Network Functions (NFs) instead of network entities. This service-based architecture offers NF reusability to reduce costs and modularity, allowing for the best vendor options for customer radio products. 5G introduced the O-RAN concept to decompose the RAN architecture, allowing for increased speed, flexibility, and innovation. NG-RAN provided this solution to speed up the development and implementation process of 5G. The O-RAN concept aims to improve the efficiency of RAN by breaking it down into components, allowing for more agility and customization. The four protocols, the eCPRI interface, and the functionalities of fronthaul that NGRAN follows are expressed further. Additionally, the significance of NR is described with an explanation of its benefits. Some benefits are high data rates, lower latency, improved spectral efficiency, increased network flexibility, and improved energy efficiency. The timeline for 5G development is provided along with different 3GPP releases. Stand-alone and non-stand-alone architecture is integral while developing the 5G architecture; hence, it is also defined with illustrations. The two frequency bands that NR utilizes, FR1 and FR2, are expressed further. FR1 is a sub-6 GHz frequency band. It contains frequencies of low and high values; on the other hand, FR2 contains frequencies above 6GHz, comprising high frequencies. FR2 is commonly known as the mmWave band. Data collection for implementing the ML approaches is expressed that contains the test setup, data collection, data description, and data visualization part of the thesis work. The Test PC runs tests, executes test cases using test libraries, and collects data from various logs to analyze the system’s performance. The logs contain information about the test results, which can be used to identify issues and evaluate the system’s performance. The data collection part describes that the data was initially present in JSON files and extracted from there. The extraction took place using the Python code script and was then fed into an Excel sheet for further analysis. The data description explains the parameters that are taken while training the models. Jupyter notebook has been used for visualizing the data, and the visualization is carried out with the help of graphs. Moreover, the ML techniques used for analyzing the data are described. In total, three methods are used here. All the techniques come under the category of supervised learning. The explained models are random forest, XG Boost, and LSTM. These three models form the basis of ML techniques applied in the thesis. The results and discussion section explains the outcomes of the ML models and discusses how the thesis will be used in the future. The results include the parameters that are considered to apply the ML models to them. SINR, noise power, rxPower, and RSSI are the metrics that are being monitored. These parameters have variance, which is essential in evaluating the quality of the product test setup, the quality of the software being tested, and the state of the test environment. The discussion section of the thesis explains why the following parameters are taken, which ML model is most appropriate for the data being analyzed, and what the next steps are in implementation.
Kokoelmat
- Avoin saatavuus [34589]