摘要
再入滑翔目标的轨迹预测是一项困难且具有意义的技术,现有利用简单函数拟合控制参数进行轨迹预测的方法,拟合精度不高且对数据的关联性不强。针对该问题,本文结合长短期时序网络提出了基于控制参数估计的智能轨迹预测算法。首先,通过设计快速轨迹生成算法,结合攻角走廊模型快速生成大量机动轨迹,构建数据集。然后,建立了包含末点修正网络、控制参数修正网络及预测网络的智能轨迹预测框架,利用数据集对关键控制参数的变化规律进行学习。最后,结合目标运动模型积分外推实现轨迹的准确预测。仿真结果表明,所设计的预测算法在不同机动模式下的预测平均误差不超过1.4 km,最大误差不超过2.5 km,能够实现轨迹的快速预测,且对大气扰动造成的模型不确定性具有一定的鲁棒性。
Trajectory prediction of reentry glide target is a difficult and significant technology.The existing prediction methods obtain predicted trajectory by fitting control parameters’temporal variation using simple function,by which fitting accuracy is not high and data association is insufficient.Aiming at that problem,an intelligent trajectory prediction algorithm based on control parameter estimation is proposed combined with long and short term memory networks.Firstly,through designing the fast trajectory generation algorithm,combined with the attack angle corridor model,a large number of maneuvering trajectories are quickly generated to consist of data sets.Then,the intelligent trajectory prediction framework is built,which includes the last point modification network,control parameters modification network and parameters prediction network.The variation laws of key control parameters are studied through data sets.Finally,the trajectories are predicted accurately by extrapolation combined with target’s dynamic models.The simulation results show that the average error of spatial distance and max error of spatial distance of the proposed prediction algorithm are less than 1.4 km and 2.5 km under different maneuvering modes with a fast prediction speed,and the method has robustness to model uncertainties caused by atmospheric disturbances.
作者
李明杰
周池军
雷虎民
邵雷
骆长鑫
LI Mingjie;ZHOU Chijun;LEI Humin;SHAO Lei;LUO Changxin(Air Defense and Anti-Missile College,Air Force Engineering University,Xi’an 710051,China)
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2023年第1期221-233,共13页
Systems Engineering and Electronics
基金
国家自然科学基金(62173339,61873278)资助课题。
关键词
再入滑翔目标
控制参数
长短期记忆网络
轨迹预测
reentry glide target
control parameters
long and short term memory networks
trajectory prediction
