Beamforming optimization using deep neural networks for 5G wireless communication

• Main Author: Ong, Jie Chung
• Other Authors: A S Madhukumar
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science
• Online Access: https://hdl.handle.net/10356/175369
• Institution: Nanyang Technological University

Beamforming is an advanced signal processing technique employed by wireless systems that manipulates signals from antenna arrays to create focused transmission beams. This technique enhances signal quality, strength, and network performance, particularly in 5G networks. Multiple Input Multiple Output (MIMO) technology in 5G enables complex beamforming strategies, facilitating simultaneous communication with multiple devices. Traditional beamforming methods, such as Singular Value Decomposition (SVD) beamforming, face challenges in the complex and dynamic environments of 5G MIMO networks. Deep Neural Networks (DNN) offer a promising solution with their ability to learn from vast data and adapt to changing conditions. This report explores a DNN model to optimise an SVD-based hybrid beamforming system. A parameterised millimeter-wave (mmWave) MIMO dataset that can map realistic channels from outdoor environment conditions is explored and used to train and test the DNN model. The trained DNN model is also tested against a substantially wider and more diverse dataset to display the robustness of the model. The performance of the model is also demonstrated to effectively outperform a traditional non-deep learning-based beamforming algorithm by 72-114% upon comparing the average achieved rates of both algorithms.