宇航学报

宇航学报 ›› 2019, Vol. 40 ›› Issue (2): 199-206.doi: 10.3873/j.issn.1000-1328.2019.02.009

• 制导、导航、控制与电子 • 上一篇    下一篇

再入飞行器自适应最优姿态控制

张振宁，张冉，聂文明，李惠峰   

  1. 北京航空航天大学宇航学院，北京100191
  • 收稿日期:2018-07-10 修回日期:2018-08-14 出版日期:2019-02-15 发布日期:2019-02-25
  • 基金资助:

    国家重点研发计划(2016YFB1200100)

Adaptive Optimal Attitude Control of Reentry Vehicles

ZHANG Zhen ning, ZHANG Ran, NIE Wen ming, LI Hui feng   

  1. School of Astronautics, Beihang University, Beijing 100191
  • Received:2018-07-10 Revised:2018-08-14 Online:2019-02-15 Published:2019-02-25

PDF

561

摘要：

针对再入飞行器姿态控制问题，应用自适应动态规划（ADP）理论设计了姿态控制器。将再入飞行器的姿态控制建模为非线性系统的最优控制问题，提出单网络积分型强化学习（SNIRL）算法进行求解，该算法简化了积分型强化学习（IRL）算法在迭代计算中的执行-评价双网络结构，只需要采用评价网络估计值函数就可以求得最优控制律，其收敛性得到了理论证明。基于SNIRL算法设计了自适应最优控制器，并证明了闭环系统的稳定性。通过数值仿真校验了SNIRL算法比IRL算法计算效率更高，收敛速度更快，并校验了自适应最优姿态控制器的有效性 。

关键词: 再入飞行器, 姿态控制, 自适应最优控制, 单网络积分型强化学习

Abstract:

An adaptive optimal controller is designed for the hypersonic vehicle attitude control employing adaptive dynamic programming (ADP). An optimal control problem of a nonlinear reentry attitude control system is formed and the single-network integral reinforcement learning (SNIRL) algorithm is proposed to solve this problem. SNIRL simplifies the actor-critic structure of the integral reinforcement learning (IRL) algorithm during iteration so that the optimal controller can be obtained using only one network which approximates the value function. The convergence of this algorithm is guaranteed. Based on the SNIRL algorithm, the adaptive optimal controller is designed and the stability of the closed-loop system is also proved. The simulation examples are provided to show that SNIRL converges faster and has higher calculation efficiency than IRL. The effectiveness of the adaptive optimal attitude controller is also shown in the results.

Key words:  Reentry vehicle, Attitude control, Adaptive optimal control, Single-network integral reinforcement learning

中图分类号: