AHMAD, M. Predikce CQI v mobilních sítí 5G NR [online]. Brno: Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií. 2025.
The master thesis submitted by Ahmad Md Nayeem focuses on the research area related to the CQI (Channel Quality Indicator) prediction for MCS (Modulation and Coding Schemes) adaption in cellular systems. The theoretical part describes the CSI (Channel State Information) mechanism, CSI report components (including CQI), CQI parameter importance in mobile networks, and CSI/CQI aging/inaccuracy due to user mobility. In the practical part, the NS3 (Network Simulator 3) was used as the tool for CQI data generation, namely, LENA-5G module was selected and utilized. The dataset obtained in the first step was further utilized as the input for the creation of the CQI prediction method utilizing the LSTM (Long short-term memory) and Random Forest machine learning models. The original research goal of the practical part was to develop a CQI prediction module for accurate downlink scheduling in 5G New Radio cellular systems, aiming to reduce the negative impact of the outdated CQI and MCS, leading to the degradation of the network performance (especially in the case of high-speed scenarios). From the supervisor’s perspective, the level of submitted work is at the border of acceptability. This work was conducted under the double degree study program of Tampere University (Finland) and Brno University of Technology (Czech Republic). The student extended the period of the thesis to two years to achieve the expected results. Unfortunately, the provided results are generic, without an explicit description of the relationship and influence between the key (communication) parameters, that is, the CQI prediction, RSSI, RSRP, SINR, and SNR. It must also be stated that artificial intelligence has been used to generate this work, namely, ChatGPT and Copilot. To conclude, I recommend the thesis for defense, but I encourage the committee to thoroughly discuss the technical aspects of the thesis. The evaluation score is 50/E.
The presented thesis addresses a highly relevant and timely topic for today’s mobile network operators, focusing on CQI prediction in 5G NR using NS-3 simulations combined with machine learning models. The selection of the NS-3 simulator, particularly the use of the LENA-5G module, is appreciated because it provides a practical and widely accepted framework for network-level evaluation. The approach of generating synthetic data through simulation and applying machine learning techniques to predict CQI values is appropriate and aligns well with ongoing research trends in the field. However, there are several areas where this thesis could be improved. In particular, the NS-3 simulation component lacks sufficient detail and visual representations. For instance, a clearer illustration of the node arrangement, mobility patterns, and simulation topology would help readers better understand the experimental setup. Including relevant screenshots or figures from the simulation would enhance clarity. Furthermore, the evaluation and comparison of machine learning models remain quite general. Although the thesis mentions several algorithms, the performance comparison could be made more rigorous through well-structured tables and figures that highlight the key metrics (e.g., MSE, R). Such visualizations would significantly strengthen the results and discussion section, which currently lacks depth in interpreting specific outcomes or drawing strong conclusions from the data. It is also recommended that students consider publishing their source code and datasets in a public GitHub repository. From a presentation standpoint, the thesis suffers from formatting inconsistencies, including LaTeX issues, such as incorrect spacing, inconsistent use of abbreviations, and several grammatical errors throughout the text. A more careful proofreading and formatting review would greatly improve the overall quality and readability of the document. The topic is well chosen, the methods are appropriate, and the work demonstrates the ability to conduct simulation-based research and apply data-driven techniques in a 5G+ context. Therefore, I consider this thesis to be a valid contribution at the master's level.
eVSKP id 168206