Patent · US Active

Composite beam generation using machine learning

US11996919B2 · kind B2 · utility

0Cited by
5References
20Claims
0Family size

Assignee

Inventors

Key dates

Filing dateJul 29, 2022
Grant dateMay 28, 2024
Priority date
Expiry dateJul 29, 2042

Classification

  • Technology area (CPC H)Electricity
  • CPC primaryH04B7/0617
  • WIPO fieldTelecommunications
  • WIPO sectorElectrical engineering

Abstract

Methods, apparatuses, systems, and computer-readable media for a composite beam generation using machine learning. A method includes identifying a narrow beam for a transmission, identifying a composite beam including the narrow beam based on an association between one or more composite beams and one or more narrow beams, identifying one or more beamforming weights for transmitting the composite beam, and transmitting the composite beam using the one or more beamforming weights. The one or more beamforming weights are determined based on machine learning. In some embodiments, the method includes generating a composite beam codebook including information indicating sets of one or more beamforming weights corresponding to composite beam indexes, respectively. The sets of one or more beamforming weights are determined using respective parameters of a machine learning algorithm that are separately updated for coverage regions of composite beams corresponding to the composite beam indexes, respectively.

Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.