基于深度学习模型的空中目标航迹预测算法的研究  

Research on the Trajectory Prediction Algorithm of Aerial Targets Based on Deep Learning Models

作　　者：王晶 段衍安 蒋明 Wang Jing;Duan Yan'an;Jiang Ming(PLA Army Academy of Artillery and Air Defense,Hefei,China)

机构地区：[1]中国人民解放军陆军炮兵防空兵学院,安徽合肥

出　　处：《科学技术创新》2025年第10期9-13,共5页Scientific and Technological Innovation

摘　　要：现有的武器系统中对目标航迹的预测通常都是基于常用的目标假定模型下进行状态估计和数据关联的。随着现代战场空中目标的机动性越来越强的特点,目标假定模型已很难适应目标运动的航迹特点,这就导致传统的目标航迹预测算法无法描绘目标的非线性动态特性,对于目标的复杂运动模式处理有限。本文提出了一种基于Cross Attention GRU(交叉注意力神经网络)算法的混合深度学习模型,用于空中目标的航迹预测。实验结果表明,该算法具备强大的非线性建模能力,可以适应现代战场目标的复杂性和机动性,大大提高空中目标航迹的预测精度。In the existing weapon systems,the prediction of target tracks is usually carried out for state estimation and data association under commonly used target assumption models.With the increasing maneuverability of aerial targets on the modern battlefield,the target assumption models have become very difficult to adapt to the track characteristics of target movements.This leads to the fact that the traditional target track prediction algorithms are unable to depict the nonlinear dynamic characteristics of targets and have limited capabilities in dealing with the complex motion patterns of targets.In this paper,a hybrid deep learning model of Cross Attention GRU is proposed for the short-term prediction of aerial target tracks.The experimental results show that this algorithm has powerful nonlinear modeling capabilities,can adapt to the complexity and maneuverability of targets on the modern battlefield,and significantly improves the prediction accuracy of aerial target tracks.

关 键 词：深度学习 Cross Attention GRU 航迹预测 MAE 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]