基于SVRGD的机载预警雷达自适应波束形成算法  

Adaptive beamforming algorithm for airborne early warning radar based on SVRGD

作　　者：彭芳[1] 吴军[2] 王帅 向建军[1] PENG Fang;WU Jun;WANG Shuai;XIANG Jianjun(Aeronautics and Astronautics Engineering School, Air Force Engineering University, Xi’an 710038, China;Air Traffic Control and Navigation College, Air Force Engineering University, Xi’an 710051, China)

机构地区：[1]空军工程大学航空工程学院,陕西西安710038 [2]空军工程大学空管领航学院,陕西西安710051

出　　处：《系统工程与电子技术》2021年第1期83-90,共8页Systems Engineering and Electronics

基　　金：航空基金(20175596020)资助课题。

摘　　要：自适应波束形成是机载预警雷达数字信号处理的一个关键环节。针对传统最小均方误差(least mean square,LMS)算法在短快拍数条件下的波束形成性能下降以及因迭代震荡易收敛于局部最优值的问题,提出了一种基于机器学习的随机方差减小梯度下降(stochastic variance reduction gradient descent,SVRGD)自适应波束形成方法。首先,建立面阵列接收信号数据模型。其次,基于随机梯度下降原理,引入方差缩减法通过内外循环迭代方式进行梯度修正,以减小随机梯度估计的方差,建立算法模型与实现流程。最后,通过设置平面阵列仿真场景,分析SVRGD自适应波束形成算法在波束形成、抗干扰、收敛速度等方面的性能,验证了该算法在低快拍数、强干扰和强噪声背景下具有的优良能力。Adaptive beamforming is a key step of digital signal processing in airborne early warning radar.To solve the problem that the beamforming performance of traditional least mean square(LMS)algorithms is reduced under the condition of short snapshot and the algorithm tend to converge to local optimal value because iterative oscillation,an adaptive beamforming approach for stochastic variance reduction gradient descent(SVRGD)based on machine learning is proposed.Firstly,the data model of planar array receiving signal is established.Secondly,based on the stochastic gradient descent principle,the variance reduction method is introduced to modify the gradient through internal and external iteration for reducing the variance of the stochastic gradient estimation,and the algorithm model and implementation process are established.Finally,by setting up a planar array simulation scene,the performance of the SVRGD algorithm in the aspects of beamforming,anti-jamming and convergence speed is analyzed,and the excellent capability of algorithm in the background of short snapshot number,strong interference and noise is verified.

关 键 词：机载预警雷达 自适应波束形成 随机梯度下降 随机方差减小梯度下降 机器学习 

分 类 号：TN959.73[电子电信—信号与信息处理]