Combined pilot frequency, feedback and multi-user hybrid coding method based on deep learning

Abstract

The invention discloses a combined pilot frequency, feedback and multi-user hybrid coding method based on deep learning. The problems that in a traditional channel estimation feedback method, the pilot frequency and feedback cost is too large, and in a traditional hybrid precoding method, performance is lost are solved. The method comprises the following steps: modeling a downlink pilot frequency transmission process as a single-layer linear activation function full-connection neural network without bias, namely a pilot frequency transmission network; constructing a pilot feedback network at a user side, compressing a received pilot signal into a bit vector, and feeding back the bit vector to a base station; constructing a hybrid precoding network at a base station end, wherein the hybrid precoding network is used for designing a hybrid precoder according to bit vectors fed back by all users; combining the pilot frequency transmission network, the pilot frequency feedback network and the hybrid precoding network to form an end-to-end neural network, and performing data-driven end-to-end training on the end-to-end neural network by taking a system sum rate as indexes; and applying the trained end-to-end deep learning neural network to pilot frequency transmission, channel feedback and hybrid precoding.

Description

technical field [0001] The present invention relates to the fields of channel estimation, channel feedback, and precoding in wireless communication, and in particular to a joint pilot, feedback, and multi-user hybrid coding based on deep learning under the hybrid analog-digital precoding architecture of a millimeter-wave massive MIMO system method. Background technique [0002] Massive MIMO (Multiple-Input Multiple-Output) is a key enabling technology for next-generation communication systems. In the downlink massive MIMO system, the base station needs to perform multi-user precoding, so as to eliminate the interference among multiple users that cannot perform joint signal processing. In low-band MIMO systems, precoding is usually performed at baseband in an all-digital manner. However, this all-digital processing method requires each antenna to be equipped with dedicated RF link hardware, which leads to excessive hardware cost and power consumption in massive MIMO systems...

AI Technical Summary

Problems solved by technology

[0006] The invention discloses a joint pilot, feedback, and multi-user hybrid encoding method based on deep learning, which solves the problems of excessive pilot and feedback overhead and performance loss of the traditional hybrid precoding method in the traditional channel estimation feedback method

Images