Deep Learning-Powered Equalization with Autoencoders for Improved 5G Communication
No Thumbnail Available
Date
2024-05
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
The Fifth Generation (5G) wireless technology has significant advancements in communication
speed, capacity and latency, revolutionizing various industries and enabling
transformative applications. However, these benefits are challenged by the
complexities of the wireless environment, characterized by multipath propagation,
fading, and interference.
This thesis address the challenge of mitigating errors within 5G communication
systems. The multipath propagation and fading present in wireless channels often
lead to Inter Symbol Interference (ISI) and other forms of distortion. As a mitigation
for this cases an autoencoder-based equalizer tailored for 5G communication systems
is proposed and thoroughly evaluated. Leveraging the power of deep learning, the
autoencoder architecture is adept at extracting complex features from received signals,
thus enabling equalization in the presence of channel impairments. Our focus
is on mitigating errors within the context of the International Telecommunication
Union (ITU) 2020 channel model and Quadrature Amplitude Modulation (QAM)
schemes (16-QAM and 64-QAM). Through simulation the performance of the proposed
equalizer is assessed using constellation plots, Symbol Error Rate (SER), Bit
Error Rate (BER) and convergence rate. Results indicate that the designed autoencoder
achieved an SER of approximately 10−4 and a BER of 10−5 for the 16-QAM
and an SER of approximately 10−3 and a BER of 10−4 for the 64-QAM. Our comparison
analysis reveals the efficacy and competitiveness of the autoencoder-based
equalizer in mitigating the effects of the channel for 5G downlink outdoor to indoor
communication system.
Description
Keywords
Autoencoder, 5G, ITU/IMT 2020 channel model, Equalization, Neural networks.