An Unsupervised Machine Learning Scheme for Index-Based CSI Feedback in Wi-Fi
arxiv(2023)
摘要
With the ever-increasing demand for high-speed wireless data transmission,
beamforming techniques have been proven to be crucial in improving the data
rate and the signal-to-noise ratio (SNR) at the receiver. However, they require
feedback mechanisms that need an overhead of information and increase the
system complexity, potentially challenging the efficiency and capacity of
modern wireless networks. This paper investigates novel index-based feedback
mechanisms that aim at reducing the beamforming feedback overhead in Wi-Fi
links. The proposed methods mitigate the overhead by generating a set of
candidate beamforming vectors using an unsupervised learning-based framework.
The amount of feedback information required is thus reduced by using the index
of the candidate as feedback instead of transmitting the entire beamforming
matrix. We explore several methods that consider different representations of
the data in the candidate set. In particular, we propose five different ways to
generate and represent the candidate sets that consider the covariance matrices
of the channel, serialize the feedback matrix, and account for the effective
distance, among others. Additionally, we also discuss the implications of using
partial information in the compressed beamforming feedback on the link
performance and compare it with the newly proposed index-based methods.
Extensive IEEE 802.11 standard-compliant simulation results show that the
proposed methods effectively minimize the feedback overhead, enhancing the
throughput while maintaining an adequate link performance.
更多查看译文
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要