Scholarlink

<p>This article proposes a model free sliding mode predictive control (MFSMPC) method based on long short-term memory and Actor-Critic Neural Networks (LSTM-ACNN) for trajectory tracking of hovercraft with external disturbances. Firstly, a position tracking controller is designed based on the fixed time theory of the Lambert W-function and the construction of auxiliary variables. Then, based on LSTM, a Critic Neural Network (CNN) is constructed to predict the status of the hovercraft online using historical input–output data. Based on Actor NN (ANN) for disturbance estimation, combined with model free sliding mode prediction method, a finite time tracking controller is solved based on input–output data. The proposed method is a purely data-driven control approach that has better robustness and smaller steady-state tracking errors compared to existing data-driven methods. Finally, the effectiveness of the proposed scheme and the convergence of tracking error were verified through two simulation examples.</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

Data-driven sliding mode predictive control for hovercraft trajectory tracking with LSTM-Based actor–critic

  • 期刊论文

  • 发表日期：

    2026年2月10日

查看全文
  • Weiqiu Zhang,
  • Yujie Xu,
  • Mingyu Fu,
  • Zhipeng Fan,
  • Guorong Zhang,
  • Lijing Dong

摘要

This article proposes a model free sliding mode predictive control (MFSMPC) method based on long short-term memory and Actor-Critic Neural Networks (LSTM-ACNN) for trajectory tracking of hovercraft with external disturbances. Firstly, a position tracking controller is designed based on the fixed time theory of the Lambert W-function and the construction of auxiliary variables. Then, based on LSTM, a Critic Neural Network (CNN) is constructed to predict the status of the hovercraft online using historical input–output data. Based on Actor NN (ANN) for disturbance estimation, combined with model free sliding mode prediction method, a finite time tracking controller is solved based on input–output data. The proposed method is a purely data-driven control approach that has better robustness and smaller steady-state tracking errors compared to existing data-driven methods. Finally, the effectiveness of the proposed scheme and the convergence of tracking error were verified through two simulation examples.