A precoding method and system based on multi-dimensional permutation equivariant neural network

By constructing a multidimensional permutation equivariant neural network, the problems of high complexity and poor reusability in existing precoding techniques are solved, realizing a low-complexity, high-performance precoding scheme and improving the performance of wireless communication systems.

CN117353780BActive Publication Date: 2026-06-26SOUTHEAST UNIV
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
SOUTHEAST UNIV
Filing Date
2023-09-28
Publication Date
2026-06-26

AI Technical Summary

Technical Problem

Existing precoding techniques fail to effectively utilize the multidimensional permutation and other variability characteristics of channel information, resulting in high network complexity and poor reusability, which affects the performance of wireless communication systems.

Method used

A multidimensional permutation equivariant neural network is constructed to optimize the design of the precoding matrix using channel state information. The equivariance of multidimensional permutation reduces network complexity and improves reusability. A multidimensional permutation equivariant neural network is used to design the precoding scheme.

Benefits of technology

It significantly improves the performance of wireless communication systems, reduces the number of network parameters, and enhances the reusability of precoding schemes, achieving high-performance precoding with low complexity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117353780B_ABST
    Figure CN117353780B_ABST
Patent Text Reader

Abstract

The application discloses a kind of pre-coding method and system based on multidimensional permutation invariance neural network, the basis on known channel state information, with each distributed access point pre-coding power as constraint, with maximum system performance index as objective function, construct pre-coding optimization problem;The real part and imaginary part of channel information tensor and noise variance are connected in series to obtain tensor dataset;According to the obtained dataset and pre-coding optimization problem, construct multidimensional permutation invariance neural network and train;The trained multidimensional permutation invariance neural network is used to solve pre-coding optimization problem to obtain pre-coding scheme.The application uses the inherent permutation invariance in communication problem to construct neural network and solve the corresponding pre-coding optimization scheme, greatly reduces the required parameter amount and improves the reusability of network, with very low complexity to solve the pre-coding optimization problem, and then improves the performance of wireless communication system.
Need to check novelty before this filing date? Find Prior Art