基于强化学习的航天器姿态敏捷机动控制
Spacecraft Attitude Agile Maneuver Control Based on Reinforcement Learning
张宏 1吴云华 1毛雨荷 1曾占魁2
作者信息
- 1. 南京航空航天大学航天学院,江苏 南京 211106
- 2. 上海埃依斯航天科技有限公司,上海 201108
- 折叠
摘要
针对航天器姿态敏捷机动任务中存在模型参数不确定和外界未知干扰的问题,提出一种基于强化学习的滑模观测器和非线性干扰观测器的控制方法.首先介绍航天器的姿态动力学模型,其次采用强化学习来解决系统模型参数不确定问题,设计一种基于弹性能量函数的滑模趋近律来加快强化学习的收敛速度并结合非线性干扰观测器来估计外界未知干扰.最后仿真结果证明利用上述方法可以保持较高的姿态跟踪精度和强鲁棒性,以满足空间高动态目标跟踪控制的要求.
Abstract
Aiming at the problems of uncertain model parameters and unknown external interference in the space-craft attitude agile maneuver task,a control method of sliding mode observer based on reinforcement learning combi-ning nonlinear interference observer is proposed.Firstly,the attitude dynamics model of the spacecraft was introduced.Secondly,reinforcement learning was used to solve the uncertainty of the moment of inertia of the system,and a sliding mode reaching law based on elastic energy function was designed to speed up the convergence speed of reinforcement learning and combine the nonlinear disturbance observer to estimate the unknown external interference.Finally,the simulation results show that the method can maintain high attitude tracking accuracy and strong robustness,to meet the requirements of space high dynamic target tracking control.
关键词
高动态目标跟踪/强化学习/滑模观测器/弹性能量函数Key words
High dynamic attitude tracking/Reinforcement learning/Sliding mode observer/Elastic energy func-tion引用本文复制引用
基金项目
国家自然科学基金青蓝工程项目(61973153)
上海市优秀学术/技术带头人计划(20XD1430400)
出版年
2024
NETL
NSTL
万方数据