Hybrid beam forming method under imperfect channel state information condition

Abstract

The invention discloses a hybrid beam forming method under an imperfect channel state information condition, which comprises the following steps of: (1) for a perfect channel matrix, destroying an original matrix by adding a noise matrix to obtain an imperfect channel matrix; (2) utilizing the principle of an auto-encoder, taking the obtained imperfect channel matrix as input, taking the obtained perfect channel matrix as a label, performing training through back propagation, and constructing a channel feature extraction network; (3) channel coding is carried out through one-time forward propagation of an encoder by using the trained feature extraction network based on the auto-encoder, and a feature matrix of a channel matrix can be obtained through extraction; and (4) inputting the channel characteristic matrix into a network for hybrid beam forming. According to the method, the sparse characteristic of the millimeter wave channel matrix is effectively utilized, channel feature extraction is carried out by utilizing the auto-encoder, and the influence of noise on the hybrid beam forming spectrum efficiency under the condition of imperfect channel state information is solved.

Description

technical field [0001] The present invention relates to the technical field of beamforming, in particular to a hybrid beamforming method under the condition of imperfect channel state information. Background technique [0002] The fifth generation mobile communication technology (5G for short) is the latest generation of cellular mobile communication technology. The performance goals of 5G are high data rates, reduced latency, energy savings, lower costs, increased system capacity, and large-scale device connectivity. The mmWave frequency band is critical for enabling fifth-generation mobile communication technology (5G), autonomous vehicles, and the Internet of Things. The millimeter-wave (mm-Wave) band, formally defined as the frequency range 30-300Ghz, combined with large-scale array antennas, has the potential to provide high data rates, improved spectral efficiency and signal coverage. However, since the wavelength of mmWave decreases with increasing frequency and the...

AI Technical Summary

Problems solved by technology

[0004] Based on the imperfect channel state information, the present invention proposes a deep learning beamforming algorithm to solve the problem of serious loss of hybrid beamforming spectral efficiency when the channel state information is imperfect

The present invention designs a millimeter-wave hybrid beamforming method combining autoencoder and ResNet-18 (18-layer residual network), which solves the problem that the existing millimeter-wave hybrid beamforming method based on deep learning is heavily dependent on perfect channel state information The technical problem of effectively implementing hybrid beamforming design under imperfect channel conditions

Images