A 3D precoding method based on model-driven deep learning

Abstract

The invention discloses a three-dimensional precoding method based on model-driven deep learning, which comprises the steps of obtaining statistical channel information of U users, and calculating statistical parameters and optimal horizontal and vertical direction vectors of each user; respectively constructing deep neural network models for judging the horizontal and vertical direction vectors of each user, inputting the statistical parameters of each user into the two models, and outputting the corresponding horizontal and vertical direction vectors; training the two models respectively, and enabling the direction vectors outputted by the two models to gradually approach to the optimal direction vectors so as to obtain model parameters; and calculating statistical parameters of a user with the precoding vector to be obtained by using statistical channel information thereof, inputting the statistical parameters into the two models, and determining the precoding vector of the user according to the horizontal and vertical direction vectors outputted by the models. The three-dimensional precoding method can reduce the calculation complexity of the statistical precoding vector, has high accuracy in predicting the optimal transmission direction, and can efficiently realize the download precoding design especially when the number of users and antennas is great.

Description

technical field [0001] The present invention relates to the field of communication technology, in particular to a three-dimensional precoding method based on model-driven deep learning. Background technique [0002] Massive multiple-input multiple-output (MIMO) transmission technology is one of the key technologies in 5G communication systems. This technology replaces multi-antenna arrays with large-scale antenna arrays to obtain higher spectral efficiency and transmission reliability. When the base station obtains the downlink channel state information (CSI), it can eliminate the interference between users through channel adaptive technologies such as precoding and beamforming, and at the same time use the three dimensions of space, time and frequency to transmit information, thereby Greatly increase the system capacity. However, it is very difficult to obtain complete channel state information in a real wireless communication system in time. For a time division duplex (...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above existing problems, the present invention provides a three-dimensional precoding method based on model-driven deep learning for the downlink transmission system of the base station using a uniform planar antenna array, which solves the problem of high computational complexity of the precoding vector, and can be based on the user's Statistical channel information is used to design the user's precoding vector to obtain higher accuracy with less computing resources. To achieve this goal, the present invention provides a three-dimensional precoding method based on model-driven deep learning, including the following steps:

Images