基于递归神经网络的再入飞行器最优姿态控制
Recurrent neural network-based optimal attitude control of reentry vehicle
摘要点击 146  全文点击 75  投稿时间:2020-03-12  修订日期:2020-10-02
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2020.00134
  2021,38(3):329-338
中文关键词  再入飞行器  最优控制  自适应动态规划  递归神经网络  姿态跟踪
英文关键词  reentry vehicle  optimal control  adaptive dynamic programming  recurrent neural networks  attitude tracking control
基金项目  天津市教委科研计划项目(2017KJ249)资助.
作者单位E-mail
吉月辉 天津理工大学 jiyuehuitju@163.com 
周海亮 天津市计量监督检测科学研究院  
车适行 天津理工大学  
高强 天津理工大学  
中文摘要
      针对再入飞行器的姿态跟踪问题, 基于递归神经网络提出最优跟踪控制. 采用反步法和递归神经网络, 设 计自适应前馈控制, 将再入飞行器的最优姿态跟踪问题转化为等价的姿态角误差/角速率误差最优调节问题. 采用 自适应动态规划技术, 解决最优调节问题. 引入神经网络估计最优控制中的代价函数, 推导最优反馈控制律, 同时保 证Hamilton–Jacobi–Isaacs(HJI)方程估计误差最小化. 采用Lyapunov理论, 保证闭环系统中所有信号, 包括姿态角跟 踪误差是一致最终有界的. 在MATLAB/Simulink中仿真验证了所提出控制策略的有效性.
英文摘要
      An optimal control is proposed based on recurrent neural networks (RNNs) for the attitude tracking problem of reentry vehicle. Firstly, backstepping and RNNs are introduced to accomplish the adaptive feedforward control. The optimal attitude tracking problem of the reentry vehicle is transformed into the equivalent optimal regulation problem for attitude angle error/angular rate error. Then, adaptive dynamic programming is adopted to fulfill the optimal regulation problem. The neural network is utilized to estimate the cost function in the optimal control, subsequently the optimal feedback control law is constructed, and the estimation error in HJI equation is minimized. The stability analysis based on Lyapunov theory can ensure that all the signals in the closed-loop system, especially attitude angle error, are uniformly ultimately bounded. The effectiveness of the proposed control strategy is verified by numerical simulation in MATLAB/Simulink environment.mulation in MATLAB/SIMULINK environment.