中国空间科学技术

中国空间科学技术 ›› 2018, Vol. 38 ›› Issue (1): 36-43.doi: 10.16708/j.cnki.1000-758X.2018.0003

• 研究探讨 • 上一篇    下一篇

RLV再入神经网络自适应姿态控制器设计

余光学*，程兴，杨云飞   

  1. 北京宇航系统工程研究所，北京  100076
  • 收稿日期:2017-08-08 录用日期:2018-01-15 发布日期:2020-02-12 出版日期:2018-02-25
  • 通讯作者: 余光学(1986-),男,工程师,yuguangxue123@126.com,研究方向为飞行器制导、控制与动力学

Design ofreentry neuralnetwork adaptive attitude controller for reusable launch vehicle

YU Guangxue*, CHENG Xing, YANG Yunfei   

  1. Beijing Institute of Astronautical Systems Engineering, Beijing 100076, China
  • Received:2017-08-08 Accepted:2018-01-15 Online:2020-02-12 Published:2018-02-25

PDF (PC)

592

摘要： 可重复使用运载器（RLV）大包线再入过程中，广泛存在模型不确定与外界干扰，会给姿态控制器的设计带来不利影响，为此提出了一种神经网络自适应控制器设计方案。基于时标分离原理设计了快、慢双回路控制结构。在此基础上设计了径向基神经网络（RBFNN）自适应律，用于在线估计模型不确定和外界干扰力矩，并在控制器中进行补偿。仿真验证表明，RBFNN 自适应控制器能良好地完成姿态跟踪控制，有效地抑制干扰力矩对姿态控制的影响。自适应律能够在线估计真实的飞行器动态和外界干扰力矩，控制器具有抗扰动能力。

关键词: 可重复使用运载器, 再入, 自适应控制, RBF神经网络, 抗干扰, 不确定性

Abstract:

The control of reusable launch vehicle (RLV) is challenging due to the changes in the dynamics as the vehicle flies through large flight envelopes. There are uncertainties and disturbances in the reentry phase of RLV, influencing attitude control performance. Based on radical basis function neural network (RBFNN), an adaptive attitude controller design scheme was presented. Firstly, an RLV control model was developed. The fast and slow loops control system was designed based on time-scale separate theory. Then an RBFNN was implemented to generate the estimation of model uncertainty and disturbance. The adaptive controller based on RBFNN was used to compensate for the effect of the modeling error and disturbance torque. Results show that the control scheme meets the attitude tracking performance requirements. Simulation demonstrates that the RBFNN can estimate the modeling uncertainty and disturbance torque effectively.

Key words:

reusable launch vehicle (RLV), reentry, adaptive control, radical basis function neuralnetwork (RBFNN), disturbance-rejection, uncertainty