US 11,742,901 B2
Deep learning based beamforming method and apparatus
Seung Eun Hong, Daejeon (KR); Hoon Lee, Busan (KR); Seok Hwan Park, Jeonju-si (KR); and Jun Beom Kim, Gwangju (KR)
Assigned to ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE, Daejeon (KR); and PUKYONG NATIONAL UNIVERSITY INDUSTRY-UNIVERSITY COOPERATION FOUNDATION, Busan (KR)
Filed by ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE, Daejeon (KR); and Pukyong National University Industry-University Cooperation Foundation, Busan (KR)
Filed on Jul. 20, 2021, as Appl. No. 17/380,826.
Claims priority of application No. 10-2020-0092834 (KR), filed on Jul. 27, 2020; and application No. 10-2021-0091256 (KR), filed on Jul. 12, 2021.
Prior Publication US 2022/0029665 A1, Jan. 27, 2022
Int. Cl. H04B 7/0426 (2017.01); G06N 3/08 (2023.01); H04B 7/08 (2006.01)
CPC H04B 7/043 (2013.01) [G06N 3/08 (2013.01); H04B 7/086 (2013.01)] 15 Claims
OG exemplary drawing
 
1. A beamforming method using a deep neural network, the deep neural network comprising an input layer, L hidden layers (L is a natural number greater than or equal to 1), and an output layer, and the beamforming method comprising:
obtaining channel information h between a base station and K terminals (K is a natural number greater than or equal to 1) and a transmit power limit value P of the base station, and inputting h and P into the input layer; and
performing beamforming on signals to be transmitted to the K terminals using beamforming vectors derived using the output layer and at least one activation function,
wherein the base station transmits the signals to the K terminals using M transmit antennas (M is a natural number greater than or equal to 1), and
wherein the deep neural network is trained:
in an unsupervised learning scheme in which parameters of the deep neural network are trained to minimize a loss function defined by multiplying a sum data rate calculated according to input-output mapping of the deep neural network by (−1); or
in a supervised learning scheme in which the parameters of the deep neural network are trained to minimize a loss function defined as a difference between a sum data rate calculated according to optimal beamforming vectors corresponding to the channel information h and the transmit power limit values P and a sum data rate calculated according to the input-output mapping of the deep neural network.