| 引用本文: | 安航,鲜斌.无人直升机的姿态增强学习控制设计与验证[J].控制理论与应用,2019,36(4):516~524.[点击复制] |
| AN Hang,XIAN Bin.Attitude reinforcement learning control of an unmanned helicopter with verification[J].Control Theory and Technology,2019,36(4):516~524.[点击复制] |
|
| 无人直升机的姿态增强学习控制设计与验证 |
| Attitude reinforcement learning control of an unmanned helicopter with verification |
| 摘要点击 2362 全文点击 1255 投稿时间：2017-12-12 修订日期：2018-05-22 |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| DOI编号 10.7641/CTA.2018.70926 |
| 2019,36(4):516-524 |
| 中文关键词 无人直升机 增强学习 鲁棒控制 神经网络 实验验证 |
| 英文关键词 Unmanned helicopter Reinforcement learning Robust control Neural network Experimental verification |
| 基金项目 国家自然科学基金( 91748121, 60804004, 90916004), 天津市应用基础与前沿技术研究计划重点项目(14JCZDJC31900),天津市科技支撑计划重点项目(15ZCZDGX00810) |
|
| 中文摘要 |
| 本文针对小型无人直升机的姿态控制问题, 考虑到现有基于模型的控制方法对直升机动力学模型的先验信息依赖较大, 以及未建模动态系统的影响等问题, 设计了一种基于增强学习(Reinforcement Learning,RL)的飞行控制算法. 仅利用直升机的在线飞行数据, 补偿了未建模不确定性的影响. 同时为了抑制外界扰动, 提高系统的鲁棒性, 设计了一种基于误差符号函数积分的连续鲁棒(RISE)控制算法. 将两种算法结合, 并利用基于Lyapunov分析的方法, 证明了无人机姿态控制误差的半全局渐近收敛. 最后在无人直升机飞行控制实验平台上, 进行了姿态控制的实时实验验证. 实验结果表明, 本文提出的控制方法具有良好的控制效果, 对系统不确定性和外界风扰具有良好鲁棒性. |
| 英文摘要 |
| This paper presents a new control law based on reinforcement learning(RL) for the attitude control of a miniature unmanned helicopter. Most existing model based controllers depend on dynamic model knowledge. The proposed control law only uses input and output data of the helicopter to compensate for the unmodeling uncertainties of the helicopter. In order to reduce the effects associated with the unknown external disturbances and improve the robustness of the system, a continuous robust control strategy(RISE) is combined with the reinforcement based control methodology to formulate the attitude controller for the helicopter. Semi-global asymptotic convergence of the helicopter''s attitude control error and the stability of the closed loop system are proved via Lyapunov based stability analysis. Finally, real-time experiments are performed on a helicopter flight testbed. The experimental results show that the proposed control design can achieve a good control performance under the effects of system uncertainties and external wind disturbances |
|
|
|
|
|