Scholarlink

<p>This paper proposes an Adaptive Practical Prescribed-Time Control (APPTC) method integrated with a Prescribed-Time Disturbance Observer (PTDO) to address the trajectory tracking problem for Unmanned Surface Vehicles (USVs) under strict time constraints and environmental disturbances. A novel Saturated Prescribed-Time Adjustment (SPTA) function is introduced to design both the PTDO and the controller, ensuring bounded control gains and continuous transitions. The PTDO accurately estimates lumped disturbances within a prescribed time. The APPTC, developed within an actor-critic framework using backstepping and reinforcement learning (RL), guarantees convergence of tracking errors and neural network parameters within a user-specified time, independent of initial conditions. RL is employed to optimize the long-term cost function, balancing tracking performance and energy efficiency by approximating the Hamilton-Jacobi-Bellman equation, thus enhancing adaptability to nonlinear dynamics and disturbances. An auxiliary system mitigates input saturation, improving practical applicability. Simulation experiments validate the effectiveness of the proposed method.</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

Adaptive prescribed-time trajectory tracking control for USVs with disturbance and input saturation

  • 期刊论文

  • 发表日期：

    2025年9月26日

查看全文
  • Zhenyao Fan,
  • Lipeng Wang,
  • Hao Meng

摘要

This paper proposes an Adaptive Practical Prescribed-Time Control (APPTC) method integrated with a Prescribed-Time Disturbance Observer (PTDO) to address the trajectory tracking problem for Unmanned Surface Vehicles (USVs) under strict time constraints and environmental disturbances. A novel Saturated Prescribed-Time Adjustment (SPTA) function is introduced to design both the PTDO and the controller, ensuring bounded control gains and continuous transitions. The PTDO accurately estimates lumped disturbances within a prescribed time. The APPTC, developed within an actor-critic framework using backstepping and reinforcement learning (RL), guarantees convergence of tracking errors and neural network parameters within a user-specified time, independent of initial conditions. RL is employed to optimize the long-term cost function, balancing tracking performance and energy efficiency by approximating the Hamilton-Jacobi-Bellman equation, thus enhancing adaptability to nonlinear dynamics and disturbances. An auxiliary system mitigates input saturation, improving practical applicability. Simulation experiments validate the effectiveness of the proposed method.