航空学报 > 2024, Vol. 45 Issue (15): 329652-329652   doi: 10.7527/S1000-6893.2023.29652

基于马尔可夫决策的四轴飞行器自动着陆方法

顾兆军1, 赵欢1,2, 王家亮2(), 聂留阳1,2   

  1. 1.中国民航大学 信息安全测评中心,天津 300300
    2.中国民航大学 计算机科学与技术学院,天津 300300
  • 收稿日期:2023-09-26 修回日期:2023-11-05 接受日期:2023-12-27 出版日期:2024-01-25 发布日期:2024-01-11
  • 通讯作者: 王家亮 E-mail:jl-wang@cauc.edu.cn
  • 基金资助:
    民航安全能力建设资金(PESA2022093);中国民航大学信息安全测评中心开放基金(ISECCA-202007);中国民航大学研究生科技创新基金(2022YJS064)

Automatic landing method for quad-rotor helicopter based on Markov decision process

Zhaojun GU1, Huan ZHAO1,2, Jialiang WANG2(), Liuyang NIE1,2   

  1. 1.Information Security Evaluation Center,Civil Aviation University of China,Tianjin 300300,China
    2.College of Computer Science and Technology,Civil Aviation University of China,Tianjin 300300,China
  • Received:2023-09-26 Revised:2023-11-05 Accepted:2023-12-27 Online:2024-01-25 Published:2024-01-11
  • Contact: Jialiang WANG E-mail:jl-wang@cauc.edu.cn
  • Supported by:
    Civil Aviation Safety Capacity Building Funding(PESA2022093);Open Fund of Information Security Evaluation Center of Civil Aviation University of China(ISECCA-202007);Graduate Science and Technology Innovation Fund of Civil Aviation University of China(2022YJS064)

摘要:

针对仅使用民用GPS引导四轴飞行器自动着陆时偏差较大的问题,基于马尔可夫决策设计了一种能够有效提升自动着陆精度的引导方法。首先,设计一种基于改进人工势场算法的对准方法,使用AprilTag3识别算法快速识别着陆标签,对标签中心划定势场,使飞行器可在力学牵引下进行偏航对准和云台俯仰对准;其次,将飞行器的着陆过程建立为马尔可夫决策模型,设置动作集、状态集与奖励集之间的状态转移关系,防止飞行器偏离预期状态;最后,结合飞行器自动着陆的问题环境,在传统马尔可夫决策模型的基础上增加信息集,利用飞行信息辅助转移关系的判定,提高状态与动作的决策准确性。通过安娜飞(ANAFI)四轴飞行器平台进行实际飞行验证,实验结果表明,基于马尔可夫决策模型的自动着陆引导方法能够有效抑制飞行器在着陆过程中的过度动作和误判断,着陆误差由米级缩小至约10 cm,满足四轴飞行器自动着陆的精度要求。

关键词: 马尔可夫决策过程, 自动着陆, 数据融合, 人工势场, 四轴飞行器

Abstract:

When using only civilian GPS guidance for automatic landing of quad-rotor helicopters, there exists the problem of significant deviation. A guidance method based on Markov decision is designed to effectively improve the accuracy of automatic landing. Firstly, an alignment method based on the improved artificial potential field algorithm is designed. The AprilTag3 recognition algorithm is used to quickly identify the landing tag and build the potential field at its center. The yaw alignment of aircraft and pitch alignment of gimbal are performed under the gravitational pull. Secondly, the landing process is established as a Markov Decision Process (MDP) model. The state transition relationship among the action set, state set, and reward set is designed to prevent the aircraft from deviating from the expected state. Finally, considering the environment of automatic landing, an information set is added to the MDP model. The information during flight is used to assist in the determination of the transition relationship, so as to improve the decision-making accuracy of states and actions. Actual flight verification is carried out on the ANAFI quad-rotor helicopter platform. The experimental results show that the automatic landing guidance method based on the MDP model can effectively suppress excessive movements and misjudgments of the aircraft during the landing process. The landing error has been reduced from the meter level to about 10 centimeters, meeting the accuracy requirements for automatic landing of quad-rotor helicopters.

Key words: Markov Decision Process (MDP), automatic landing, data fusion, artificial potential field, quad-rotor helicopter

中图分类号: