摘要
为消除导引头天线罩引入的瞄准误差对制导系统稳定性和精度造成的负面影响,提出了一种基于神经网络和无迹卡尔曼滤波融合的天线罩误差斜率估计方法。考虑先验模型知识,分别建立导引头、自动驾驶仪、弹目相对运动系统和弹体动力学系统的动态模型,选取真实视线角、视角和天线罩误差斜率作为状态变量,根据视线角观测值建立测量模型。考虑到模型的不确定性,基于神经网络技术学习非线性滤波模型中的动力学方程,结合无迹卡尔曼滤波技术,根据所学习的代理模型和带噪声的系统量测,对天线罩误差斜率等状态进行实时在线估计。与传统采用非线性滤波技术的天线罩误差斜率估计方法相比,本方法基于数据驱动思想,减少了对精确动力学模型的依赖,能有效消除模型不确定性的影响。与单纯采用离线训练构造的神经网络相比,本方法结合贝叶斯滤波理论,对实时数据具有更强的适应性。经多次仿真实验,证实该方法能够有效控制预测误差,具有较高精度。
In order to eliminate the negative impact of the aiming error introduced by the seeker radome on the stability and accuracy of the guidance system,a radome error slope estimation method based on the fusion of neural network and unscented Kalman filter is proposed.Considering prior model knowledge,the dynamic models of seeker,autopilot,projectile relative motion system and projectile body dynamics system are established respectively.The real sight angle,viewing angle and radome error slope are selected as state variables,and the measurement model is established according to the observation value of sight angle.Then considering the uncertainty of the model,the dynamic equation of the nonlinear filtering model is learned based on the neural network technology,combined with the unscented Kalman filter technology,according to the learned proxy model and the system measurement with noise,the radome error slope and other states are estimated online in real time.Compared with the traditional radome error slope estimation method using nonlinear filtering technology,this method is based on data-driven idea,which reduces the dependence on accurate dynamic model,and can effectively eliminate the influence of model uncertainty.Compared with the neural network constructed by offline training alone,this method combined with Bayesian filtering theory has stronger adaptability to real-time data.The simulation results show that this method can effectively control the prediction error and has high precision.
作者
虞卞雨萱
陆科林
符启恩
张宁
Yu Bianyuxuan;Lu Kelin;Fu Qien;Zhang Ning(School of Automation,Southeast University,Nanjing 210096,China;Beijing Electro-mechanical Engineering Institute,Beijing 100074,China)
出处
《战术导弹技术》
北大核心
2023年第1期121-131,161,共12页
Tactical Missile Technology
基金
国家自然科学基金(61903084)。