首页|CNN-AC algorithm for hybrid precoding in millimeter-wave massive MIMO systems
CNN-AC algorithm for hybrid precoding in millimeter-wave massive MIMO systems
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Springer Nature
Abstract Hybrid precoding is one of the promising technologies for millimeter-wave communications. As a crucial structure of hybrid precoding, an adaptively-connected (AC) structure can achieve a trade-off between spectral efficiency (SE) and energy efficiency (EE). On account of a large quantity of phase shifters (PSs), the AC structure usually suffers from low spectral efficiency. To tackle this problem, this paper proposes a hybrid precoding algorithm based on a convolutional neural network (CNN) for the AC structure, named CNN-AC algorithm. The proposed CNN-AC algorithm translates the optimization problem from hybrid precoding to a predictive neural network problem. Firstly, a new CNN framework is constructed to predict the vectorized hybrid precoding matrix. Specifically, three unique network layers are constructed to meet the certain constraint. Then, using the fully digital precoder as a reference label, the CNN is trained to minimize the distance between the hybrid precoders and the digital precoders with the estimated channel gain as input. Finally, the estimated channel gain is input into the CNN, resulting in an output hybrid precoding matrix. Simulation results confirm that the CNN-AC algorithm offers satisfactory SE and high EE.
NETL
NSTL
Springer Nature
