宇航学报

宇航学报 ›› 2017, Vol. 38 ›› Issue (3): 279-286.doi: 10.3873/j.issn.1000-1328.2017.03.008

• 制导、导航、控制与电子 • 上一篇    下一篇

基于神经网络的RLV再入段有限时间自适应姿态控制

陈辰，吕跃勇，马广富，王润驰   

  1. 哈尔滨工业大学控制科学与工程系，哈尔滨150001
  • 收稿日期:2016-05-21 修回日期:2016-11-29 出版日期:2017-03-15 发布日期:2017-03-25
  • 基金资助:

    国家自然科学基金（61673135, 61603114,61403103）

Neural Network Based Finite Time Stable Adaptive Attitude Control for RLV Reentry

CHEN Chen, LV Yue yong, MA Guang fu, WANG Run chi   

  1. Department of Control Science and Engineering, Harbin Institute of Technology, Harbin 150001, China
  • Received:2016-05-21 Revised:2016-11-29 Online:2017-03-15 Published:2017-03-25

PDF

1491

摘要：

针对可重复使用运载器（RLV）再入段的姿态控制问题，提出了一种基于神经网络的有限时间自适应姿态跟踪控制方法。首先，在传统RLV建模的基础上将模型不确定性、耦合及扰动力矩分离作为复合扰动；然后，利用径向基神经网络（RBFNN）对其在线估计并在标称控制器中进行动态前馈补偿；最后，利用终端吸引子改进控制器实现了对期望状态的有限时间跟踪，并通过引入鲁棒项降低了RBFNN估计误差对控制精度的影响。设计的姿态控制器无需获知精确的气动数据与扰动范围而仅需某飞行状态下的标称值。仿真结果表明提出的控制方法对RLV再入姿态跟踪具有较好的控制效果。

关键词: 可重复使用运载器, 神经网络, 自适应估计, 有限时间稳定

Abstract:

A neural network based finite-time-stable adaptive attitude control strategy for a reusable launch vehicle (RLV) reentry is proposed. Firstly, the traditional RLV dynamic model is improved by separating out the uncertainty, coupled dynamic and disturbance together as combined disturbance. Then, adaptive estimation for the combined disturbance based on radical basis function neural network (RBFNN) is introduced into a nominal controller as feed-forward compensation. Moreover, the terminal attractor is used to improve the controller so that the desired system state could be tracked in finite time, and a robust control function is also introduced so as to reduce the impact on control accuracy from the error of RBFNN estimation. Only the nominal parameters of the system rather than the precise value and bounds of disturbance are utilized for the proposed controller. Finally, the effectiveness of the controller is demonstrated by the numerical simulations.

Key words: Reusable launch vehicle, Neural network, Adaptive estimation, Finite time stable

中图分类号: