Scholarlink

<p>In this paper, a reinforcement learning (RL)-based prescribed performance fault-tolerant control (FTC) method is proposed for nonlinear servo systems with actuator faults and external disturbances. Firstly, the equivalent unconstrained error signal derived from the prescribed performance transformation is incorporated into the backstepping control design, and an RL-based approximately optimal controller is developed as an intermediate control signal under the identifier-critic-actor framework. On this basis, to cope with the partial loss of control effectiveness and bias actuator faults, an actual FTC protocol is devised with the help of adaptive techniques. This protocol can effectively compensate for the impact of actuator faults and external disturbances on system performance, ensuring that the tracking error converges to the preassigned boundaries. Finally, comparative simulations verify the feasibility and superiority of the proposed scheme.</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

Reinforcement learning-based fault-tolerant control of nonlinear servo systems with performance guarantees

  • 期刊论文

  • 发表日期：

    2026年1月20日

查看全文
  • Ning Xie,
  • Shaoyu Lv,
  • Zhengguo Huang,
  • Hao Shen

摘要

In this paper, a reinforcement learning (RL)-based prescribed performance fault-tolerant control (FTC) method is proposed for nonlinear servo systems with actuator faults and external disturbances. Firstly, the equivalent unconstrained error signal derived from the prescribed performance transformation is incorporated into the backstepping control design, and an RL-based approximately optimal controller is developed as an intermediate control signal under the identifier-critic-actor framework. On this basis, to cope with the partial loss of control effectiveness and bias actuator faults, an actual FTC protocol is devised with the help of adaptive techniques. This protocol can effectively compensate for the impact of actuator faults and external disturbances on system performance, ensuring that the tracking error converges to the preassigned boundaries. Finally, comparative simulations verify the feasibility and superiority of the proposed scheme.