基于DNN的无人机数据OFDM传输技术  

OFDM data transmission technology of UAV based on deep neural network

作　　者：刘步花 丁丹 杨柳 薛乃阳 刘仲谦 LIU Buhua;DING Dan;YANG Liu;XUE Naiyang;LIU Zhongqian(Department of Graduate Management, Space Engineering University, Beijing 101416, China;Department of Electronic and Optical Engineering, Space Engineering University, Beijing 101416, China;Chongqing Aerospace Rocket Electronic Technology Co. Ltd., Chongqing 400039, China)

机构地区：[1]航天工程大学研究生院,北京101416 [2]航天工程大学电子与光学工程系,北京101416 [3]重庆航天火箭电子技术有限公司,重庆400039

出　　处：《系统工程与电子技术》2022年第2期696-702,共7页Systems Engineering and Electronics

基　　金：国家高技术研究发展计划(“863”计划)(2015AA7026085)资助课题。

摘　　要：针对无人机信道含多径、多普勒频移,还易受到外来干扰和高功率放大带来的非线性失真影响的问题,提出一种基于深度神经网络(deep neural network,DNN)的无人机正交频分复用(orthogonal frequency division multiplexing,OFDM)数据传输技术。下行链路采用OFDM系统,在接收端进行解调之后利用最小二乘(least square,LS)信道估计和迫零(zero forcing,ZF)算法做初步信号检测,然后输入到由BiLSTM和全连接层组成的DNN中进行信道估计和信号检测优化,并恢复数据流。仿真结果表明,与传统插值方法相比,所提方法在无人机起降、飞行和盘旋等3种状态下,信道估计性能具有明显优势,比特误差率性能提升至少一个数量级;在非线性失真和外来干扰的影响下,所提方法仍具有显著性能优势,不仅简化了系统对非线性失真和干扰的处理模块,而且提高了系统的可靠性。The unmmand aerial vehicle(UAV)channel contains multipath,Doppler frequency shift.It is easily affected by external interference and nonlinear distortion caused by high power amplification.To solve those problems a data transmission technology of UAV orthogonal frequency division multiplexing(OFDM)based on deep neural network(DNN)is proposed.In the downlink,OFDM system is used.After demodulation and borrowing at the receiver,the least square(LS)channel estimation and zero forcing(ZF)algorithm are used for preliminary signal detection,and then input into the DNN composed of BiLSTM and full connected layer for channel estimation and signal detection optimization,and data stream recovery.Simulation results show that compared with the traditional interpolation method,the proposed method has obvious advantages in channel estimation performance under three states of UAV takeoff and landing,flight and hovering,and the bit error rate performance is improved by at least one order of magnitude;under the influence of nonlinear distortion and external interference,the proposed method still has significant performance advantages,which not only simplifies the processing module of nonlinear distortion and interference,but also improves the stability of the system qualitative.

关 键 词：无人机 时变多径信道 正交频分复用 深度神经网络 干扰 信道估计 信号检测 

分 类 号：TN914.34[电子电信—通信与信息系统]