@article{SONG2025, 
author = {Xieda SONG and Guoru DING and Haichao WANG and Jiangchun GU and Peng TANG and Yitao XU},
title = {DOA estimation of non-cooperative UAV beam signals based on BD-DOANet},
year = {2025},
journal = {Chinese Journal of Aeronautics},
volume = {38},
number = {12},
keywords = {BD-DOANet, Beam signals, Deep learning, Direction of arrival estimation, Unmanned aerial vehicle},
url = {https://www.sciopen.com/article/10.1016/j.cja.2025.103710},
doi = {10.1016/j.cja.2025.103710},
abstract = {In recent years, the proliferation of beamforming signals has made the electromagnetic environment more complex. Traditional spectrum sensing techniques mainly focus on the detection of omnidirectional signals and can no longer meet the needs of beamforming signals. Moreover, the next-generation spectrum sensing technologies must not only reliably detect the presence of beamforming signals but also accurately estimate the spatial information of these signals. This paper investigates the issue of Direction of Arrival (DOA) of non-cooperative Unmanned Aerial Vehicle (UAV) beamforming signals, where most of the prior information about non-cooperative transmitters, such as the transmission power and the communication time slots, is unknown. In such conditions, we consider two types of data models for UAV beamforming signals with different Signal-to-Noise Ratios (SNRs). Based on these data models, we develop a UAV Beamforming signals Detection-DOA Network (BD-DOANet), comprising convolutional modules, a channel attention module, and residual modules. Simulation results show that BD-DOANet effectively captures the angle information of non-cooperative UAV beamforming signals for both ideal and non-ideal data. At higher SNR levels, the average error is below 0.5 and its mean squared error is below 0.2. Even at lower SNR levels, BD-DOANet shows superior performance of DOA estimation.}
}