航空学报 > 2026, Vol. 47 Issue (15): 333091-333091   doi: 10.7527/S1000-6893.2025.33091

基于分层自适应的无人机多模式容错控制方法

魏鹏轩1, 盛汉霖1(), 李枫韵1, 王一杰1, 史昊蓝1, 李嘉诚2, 陈芊1   

  1. 1.南京航空航天大学 能源与动力学院,南京 210016
    2.南京航空航天大学 通用航空与飞行学院,南京 210016
  • 收稿日期:2025-11-17 修回日期:2025-12-03 接受日期:2025-12-09 出版日期:2025-12-17 发布日期:2025-12-15
  • 通讯作者: 盛汉霖 E-mail:dreamshl@nuaa.edu.cn
  • 基金资助:
    国家自然科学基金(52502476)

Hierarchical adaptive multi-mode fault-tolerant control method for UAVs

Pengxuan WEI1, Hanlin SHENG1(), Fengyun LI1, Yijie WANG1, Haolan SHI1, Jiacheng LI2, Qian CHEN1   

  1. 1.College of Energy and Power Engineering,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China
    2.College of General Aviation and Flight,Nanjing University of Aeronautics and Astronautics,Nanjing 210016,China
  • Received:2025-11-17 Revised:2025-12-03 Accepted:2025-12-09 Online:2025-12-17 Published:2025-12-15
  • Contact: Hanlin SHENG E-mail:dreamshl@nuaa.edu.cn
  • Supported by:
    National Natural Science Foundation of China(52502476)

摘要:

旋翼失效使四旋翼无人机呈现出强非线性、快时变与严重欠驱动等特性,传统控制方法难以维持其飞行稳定,极易导致无人机失稳坠毁。因此,提出了一种基于分层自适应思想的多模式容错控制方法。该方法创新性地构建了一个由动态权重非线性模型预测控制(DWNMPC)和自适应增量非线性动态逆控制(AINDI)组成的分层控制框架。上层DWNMPC控制器通过设计一种状态依赖的权重自适应机制,根据无人机姿态误差在线动态调整代价函数中各状态的权重,实现了在故障瞬间优先保障姿态稳定以抑制翻滚,待系统稳定后则平滑过渡至精确轨迹跟踪任务。为应对严重故障下的模型不确定性与强气动干扰,下层设计了AINDI控制器对DWNMPC指令进行在线鲁棒自适应修正,该控制器利用传感器测量实时补偿未建模力矩,并采用带遗忘因子的递推最小二乘法(RLS)在线辨识转动惯量等关键参数,显著增强了系统的鲁棒性。实验结果表明,所提出的分层自适应容错控制方法在无故障、单旋翼部分失效和完全失效工况下均展现出良好的轨迹跟踪能力,且控制过程仅依赖无人机自身机载传感器进行状态估计,体现了其在实际物理环境下的高普适性。在单旋翼完全失效并导致机体以约-10.5 rad/s高速自旋的极端情况下,轨迹跟踪均方根误差相较于无故障时在xyz轴上仅分别增加了0.047 6 m、0.054 5 m和0.083 m,显著提升了无人机的可靠性与安全性。

关键词: 四旋翼无人机, 执行器故障, 在线优化, 跟踪控制, 容错控制

Abstract:

Rotor failure induces characteristics in quadrotor UAVs such as strong nonlinearity, rapid time-variance, and severe under-actuation. Consequently, traditional control methods often struggle to maintain flight stability, frequently resulting in loss of control and crashes. To address these challenges, this paper proposes a multi-mode fault-tolerant control method based on a hierarchical adaptive framework. This method innovatively constructs a control architecture comprising Dynamic Weighting Nonlinear Model Predictive Control (DWNMPC) and Adaptive Incremental Nonlinear Dynamic Inversion (AINDI). The upper-level DWNMPC controller employs a state-dependent weight adaptive mechanism. By dynamically adjusting the weights of various states in the cost function based on attitude errors, it prioritizes attitude stability to suppress tumbling at the instant of failure, subsequently transitioning smoothly to precise trajectory tracking once the system stabilizes. To cope with model uncertainties and strong aerodynamic disturbances under severe failure conditions, the lower-level AINDI controller is designed to provide online robust adaptive correction to DWNMPC commands. This controller utilizes sensor measurements to compensate for unmodeled moments in real-time and adopts the Recursive Least Squares (RLS) method with a forgetting factor to identify key parameters, such as the moment of inertia, thereby significantly enhancing system robustness. Experimental results demonstrate that the proposed method exhibits excellent trajectory tracking capabilities under fault-free, partial rotor failure, and complete rotor failure conditions. Furthermore, the control process relies solely on onboard sensors for state estimation, reflecting its high applicability in actual physical environments. In the extreme scenario of complete single-rotor failure resulting in a high-speed spin of approximately -10.5 rad/s, the Root Mean Square Error (RMSE) of trajectory tracking increased by only 0.047 6 m, 0.054 5 m, and 0.083 m on the xy, and z axes, respectively, compared to the fault-free condition, significantly improving the reliability and safety of the UAV.

Key words: quadrotor UAV, actuator fault, online optimization, tracking control, fault-tolerant control

中图分类号: