Beam prediction model training method, beam prediction method and device

CN119301887APending Publication Date: 2025-01-10NEW H3C TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202380009041.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2023-05-09
Publication Date
2025-01-10

AI Technical Summary

Technical Problem

In mobile communication scenarios, it is difficult for the prior art to train a beam prediction model with relatively optimized performance under smaller computing power requirements, and traditional methods require a lot of pilot overhead and complex judgment logic.

Method used

By sending pilot signals in full between the base station and the terminal, receiving measurement report messages, filling in one-dimensional arrays as label data, sampling according to the preset sampling mode of different scenarios, and using multiple sets of training data to train the deep learning model, we obtain Beam prediction model.

Benefits of technology

It reduces the computing power requirements of base stations and terminals, reduces pilot overhead and data acquisition difficulty, and improves the generalization performance of the beam prediction model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119301887A_ABST
    Figure CN119301887A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides a beam prediction model training method and device and a beam prediction method and device, and relates to the technical field of wireless communication, and the method comprises the steps that a base station sends a pilot signal to a terminal by using a full-quantity sending beam, receives a measurement report message sent by the terminal, and sends the measurement report message to the terminal; the measurement results of the first preset number of beam pairs are filled into elements included in the one-dimensional array, the one-dimensional array serves as label data, and the number of the elements is the product of the maximum sending beam number and the maximum receiving beam number. And sampling the elements according to preset sampling modes corresponding to different scenes, filling the unsampled elements with invalid values to obtain training data corresponding to each preset sampling mode, and training a deep learning model by using multiple groups of training data and the label data to obtain a beam prediction model. The beam prediction model with better generalization performance can be obtained through training under the requirement of smaller computing power.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Cited By

  • Beam training and measurement pattern joint generation method based on deep learning

    CN122137437A