一种改进的双重损失判别函数生成对抗网络的航迹预测模型  

An Improved Track Prediction Model Based on GAN of Dual Loss Discriminant Function

作　　者：耿亚南 王硕 GENG Yanan;WANG Shuo(Huaxian Science and Technology Innovation Research Institute,Anyang 456480,China;Yutong Bus Co.,Ltd.,Zhengzhou 456000,China)

机构地区：[1]滑县科技创新研究院,河南安阳456480 [2]宇通客车股份有限公司,郑州450000

出　　处：《电子信息对抗技术》2025年第2期66-71,共6页Electronic Information Warfare Technology

基　　金：郑州市重大科技专项(2021KJZX0060)。

摘　　要：为提升复杂环境中无人机航迹预测精度,提出一种改进的双重判别损失生成对抗网络的航迹预测模型。首先,利用长短期记忆(Long Short-Term Memory,LSTM)网络提取航迹特性数据,并引入池化器获取无人机相对位移的交互向量,以提高模型环境抗干扰性。其次,构建生成对抗网络模型(Generative Adversarial Network,GAN)在相互博弈下不断优化改进。最后,引入对抗损失和位移损失的双重损失判别函数,提升模型整体预测精度。通过与3种较流行的预测模型实验对比结果表明,改进的GAN模型相比传统算法在航迹预测精度和稳定性上都有显著提升。In order to improve the accuracy of unmanned aerial vehicle(UAV)track prediction in complex environment,an improved dual discriminant loss function generating adversarial network track prediction model is proposed.Firstly,the characteristics of track data are extracted by long short-term memory network(LSTM),and the interaction vector of relative displacement of UAV is obtained by pooling model to improve the anti-jamming of model environment.Secondly,the dual loss discriminant function is made up of counter loss and displacement loss to improve the overall prediction accuracy of the model.Finally,generative adversarial network(GAN)is used to optimize the prediction model under adversarial game.Compared with three popular prediction models,the results show that the improved track prediction model of generative adversarial network has obvious improvement in prediction stability and prediction accuracy.

关 键 词：航迹预测 长短期记忆网络 生成对抗网络 双重损失判别 

分 类 号：TN971[电子电信—信号与信息处理]