Scholarlink

<p>Aerial manipulator systems (AMSs) are characterized by severe internal dynamic coupling, external disturbances, and model mismatches, which make robust disturbance rejection essential for maintaining stable hovering and achieving high-precision trajectory tracking. This paper presents a predefined-time sliding-mode control framework integrated with a robust adaptive learning rate neural network state observer (ANN-PTSO) to enhance the active disturbance estimation and rejection capabilities of AMSs. To achieve high-fidelity state and disturbance estimation without prior knowledge of lumped disturbance upper bounds, the ANN-PTSO utilizes an error-driven adaptive learning rate mechanism to dynamically update radial basis function neural network (RBFNN) parameters. This approach enables the rapid reconstruction of unknown nonlinear disturbances while mitigating overly aggressive NN adaptation and compensation fluctuations associated with fixed high-gain observer or compensation settings, thereby improving transient performance. Furthermore, a predefined-time terminal sliding-mode controller (PTTSMC) is developed to ensure robust compensation. By employing a modified auxiliary function with a switching threshold, the proposed controller eliminates the singularity issues inherent in conventional terminal sliding modes and drives the tracking errors into an explicitly characterized practical residual neighborhood within a user-defined predefined time, independent of initial states. Simulations encompassing aggressive manipulator motion with payloads, abrupt external disturbance injection and removal, model mismatches, composite sensor-noise corruption, and heavy-object transportation with additional position feedback noise validate the framework’s superiority in disturbance estimation accuracy and robust anti-disturbance performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predefined-time sliding mode control for aerial manipulator based on the robust adaptive learning rate neural network state observer

  • 期刊论文

  • 发表日期：

    2026年8月3日

查看全文
  • Quman Xu,
  • Zhan Li,
  • Yipeng Yang,
  • Xinghu Yu

摘要

Aerial manipulator systems (AMSs) are characterized by severe internal dynamic coupling, external disturbances, and model mismatches, which make robust disturbance rejection essential for maintaining stable hovering and achieving high-precision trajectory tracking. This paper presents a predefined-time sliding-mode control framework integrated with a robust adaptive learning rate neural network state observer (ANN-PTSO) to enhance the active disturbance estimation and rejection capabilities of AMSs. To achieve high-fidelity state and disturbance estimation without prior knowledge of lumped disturbance upper bounds, the ANN-PTSO utilizes an error-driven adaptive learning rate mechanism to dynamically update radial basis function neural network (RBFNN) parameters. This approach enables the rapid reconstruction of unknown nonlinear disturbances while mitigating overly aggressive NN adaptation and compensation fluctuations associated with fixed high-gain observer or compensation settings, thereby improving transient performance. Furthermore, a predefined-time terminal sliding-mode controller (PTTSMC) is developed to ensure robust compensation. By employing a modified auxiliary function with a switching threshold, the proposed controller eliminates the singularity issues inherent in conventional terminal sliding modes and drives the tracking errors into an explicitly characterized practical residual neighborhood within a user-defined predefined time, independent of initial states. Simulations encompassing aggressive manipulator motion with payloads, abrupt external disturbance injection and removal, model mismatches, composite sensor-noise corruption, and heavy-object transportation with additional position feedback noise validate the framework’s superiority in disturbance estimation accuracy and robust anti-disturbance performance.