A Planar Array Synthesis Method Based on Deep Learning and Radiation Pattern Superposition Method

A dual-branched convolutional neural network (CNN) integrated with hybrid training criteria is proposed for real-time multibeam synthesis in planar uniform antenna arrays. For higher speed in dataset generation, the pattern superposition method is introduced in this article, which significantly impr...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on antennas and propagation Vol. 73; no. 8; pp. 5265 - 5277
Main Authors: Huang, Jianming, Liu, Rui, Zhang, Naibo, Cui, Yansong, Ren, Weizheng, Guo, Qiuquan, Du, Yanjun, Zhao, Jiayu
Format: Journal Article
Language:English
Published: New York IEEE 01.08.2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects:
Antenna arrays
Antenna design
Antenna radiation patterns
Antennas
Approximation
Arrays
Artificial neural networks
Beam steering
Beamforming
convolutional networks
Convolutional neural networks
Criteria
Deep learning
Machine learning
Neural networks
Optimization
Planar arrays
Real time
Real-time systems
Reconfiguration
Sidelobe reduction
Sidelobes
Stochastic processes
Synthesis
Training
ISSN:0018-926X, 1558-2221
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:A dual-branched convolutional neural network (CNN) integrated with hybrid training criteria is proposed for real-time multibeam synthesis in planar uniform antenna arrays. For higher speed in dataset generation, the pattern superposition method is introduced in this article, which significantly improves data generation speed (~0.7 ms per each). With a new hybrid training strategy integrating both "data-driven approximation" and "physics-informed approximation" criteria, flexible training while maintaining model synthesis efficiency is achieved simultaneously. Based on the approaches mentioned above, the proposed dual-branched CNN enables a dynamic multiobjective balance between its two branch outputs, accomplishing amplitude-phase synthesis for antenna array design. In the numerical simulation, consistent sidelobe suppression (<inline-formula> <tex-math notation="LaTeX">20\sim 40 </tex-math></inline-formula> dB) is demonstrated for dual-beam steering across <inline-formula> <tex-math notation="LaTeX">10^{\circ }{\sim }20^{\circ } </tex-math></inline-formula> off-axis angles, with high synthesis speeds (~0.05 s per each) on suitable computing platforms. In comparison experiments, the proposed method outperforms typical population-based stochastic optimization algorithms and deep neuron network (DNN) structure in both synthesis efficiency and pattern regularity. The model's real-time capability is verified for beam reconfiguration scenarios requiring subsecond responses, suggesting a new pathway for physics-embedded deep learning in antenna array synthesis.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0018-926X
1558-2221
DOI:10.1109/TAP.2025.3564705