航空学报 > 2022, Vol. 43 Issue (8): 325433-325433   doi: 10.7527/S1000-6893.2021.25433

基于预测校正的落角约束计算制导方法

刘子超1,2, 王江1,2, 何绍溟1,2, 李宇飞3   

  1. 1. 北京理工大学 宇航学院, 北京 100081;
    2. 北京理工大学 无人机自主控制技术北京市重点实验室, 北京 100081;
    3. 北京理工大学 信息与电子学院, 北京 100081
  • 收稿日期:2021-03-01 修回日期:2021-07-19 出版日期:2022-08-15 发布日期:2021-06-18
  • 通讯作者: 何绍溟,E-mail:shaoming.he@bit.edu.cn E-mail:shaoming.he@bit.edu.cn
  • 基金资助:
    空军装备部预先研究项目(3030209)

A computational guidance algorithm for impact angle control based on predictor-corrector concept

LIU Zichao1,2, WANG Jiang1,2, HE Shaoming1,2, LI Yufei3   

  1. 1. School of Aerospace Engineering, Beijing Institute of Technology Beijing 100081, China;
    2. Beijing Key Laboratory of UAV Autonomous Control, Beijing Institute of Technology, Beijing 100081, China;
    3. School of Information and Electronics Beijing Institute of Technology, Beijing 100081, China
  • Received:2021-03-01 Revised:2021-07-19 Online:2022-08-15 Published:2021-06-18
  • Supported by:
    Advanced Research Program of Air Force Equipment Department (3030209)

摘要: 针对带有落角约束的导弹制导问题,提出一种基于预测校正思想的计算制导方法。根据导弹飞行状态与落角之间的映射关系设计了深度监督学习网络,在导弹飞行过程中基于观测信息实时预测落角;根据落角误差与控制指令之间的映射关系设计了深度强化学习模型,在比例导引律的基础上增加偏置项,基于预测落角误差实时输出偏置项,校正落角误差。引入深度监督学习网络解决了传统强化学习方法中的稀疏奖励问题。设计了仿真飞行实验训练深度监督学习网络与深度强化学习模型,然后测试了提出的计算制导方法的性能。仿真结果表明所设计的计算制导方法准确地实现了落角控制,在嵌入式计算机的计算精度高,实时性好,具备工程实用价值。

关键词: 落角约束, 计算制导, 预测校正制导, 偏置比例导引律, 深度学习

Abstract: To solve the problem of missile guidance with constraint of terminal impact angle, a learning-based computational guidance algorithm is proposed based on the general predictor-corrector concept. A deep neural network is designed based on the relationship between the flight state and impact angle, to predict the exact terminal impact angle under proportional navigation guidance with realistic aerodynamic characteristics. A biased command to nullify the impact angle error is developed based on the relationship between impact angle error and the acceleration command, and the deep reinforcement learning techniques is utilized. The deep neural network is augmented into the reinforcement learning block to resolve the issue of sparse reward that has been observed in traditional reinforcement learning formulation. Extensive numerical simulations are conducted to verify the proposed algorithm. The simulation results show that the designed computational guidance method can realize impact angle control accurately. The guidance algorithm has high precision and low delay in embedded computer, which shows that the algorithm can be applied to engineering.

Key words: impact angle constraint, computational guidance, predictor-corrector guidance, bias proportion navigation, deep learning

中图分类号: