| 一种基于学习的高超声速飞行器智能控制方法 |
| |
| 引用本文： | 王冠,夏红伟.一种基于学习的高超声速飞行器智能控制方法[J].宇航学报,2023,44(2):233-242. |
| |
| 作者姓名： | 王冠 夏红伟 |
| |
| 作者单位： | 哈尔滨工业大学航天学院，哈尔滨 150001 |
| |
| 基金项目： | 国家自然科学基金（61304108）；国家重点研发计划（2020YFC2200600） |
| |
| 摘 要： | 针对吸气式高超声速飞行器的飞行控制问题，提出一种基于学习的智能控制方法。为便于控制器设计，将飞行器动力学模型划分为速度子系统和高度子系统：为解决速度子系统控制输入受限的问题，提出一种基于强化学习的智能控制方案；对于考虑有限通信资源的高度子系统跟踪控制问题，提出一种基于事件触发的确定学习控制方案。该方案包含离线学习训练和在线触发控制两个阶段。首先在本地离线学习训练阶段获取并存储系统的未知动态知识，随后利用所获取的经验知识设计基于事件触发机制的在线触发控制器。本文所提方案基于学习的思想将离线学习训练获取的智能体和经验知识应用于在线控制，使得所提方案能够快速计算控制指令且通信资源占用少。仿真结果说明了所提出方法的有效性。
|
| 关 键 词： | 高超声速飞行器 强化学习 确定学习 飞行控制 事件触发 |
| 收稿时间： | 2022-06-26 |
A Learning based Intelligent Control Method for Hypersonic Flight Vehicle |
| |
| WANG Guan,XIA Hongwei.A Learning based Intelligent Control Method for Hypersonic Flight Vehicle[J].Journal of Astronautics,2023,44(2):233-242. |
| |
| Authors: | WANG Guan XIA Hongwei |
| |
| Affiliation: | School of Astronautics, Harbin Institute of Technology, Harbin 150001, China |
| |
| Abstract: | Aiming at the flight control problem of air breathing hypersonic vehicle, an intelligent control method based on learning is proposed. In order to facilitate the controller design, the vehicle dynamics model is divided into velocity subsystem and altitude subsystem. For the constrained input problem in the velocity subsystem, an intelligent control scheme based on reinforcement learning is proposed. For the altitude control problem with limited communication resources, an event triggered deterministic learning control scheme is proposed. The method includes two stages: offline learning and online triggered control. Firstly, the unknown dynamic knowledge of the system is acquired and stored in the local offline learning phase, and then an online event triggered controller is designed using the acquired empirical knowledge. Based on the idea of learning, the agent and experience knowledge acquired by offline learning are applied to online control, so that the proposed method can quickly calculate control commands and occupy less communication resources. The effectiveness of the proposed method is verified by simulation results. |
| |
| Keywords: | Hypersonic vehicle Reinforcement learning Deterministic learning Flight control Event trigger |
|
| 点击此处可从《宇航学报》浏览原始摘要信息 |
|
点击此处可从《宇航学报》下载全文 |
