Scholarlink

<p>In this paper, the backstepping approach is integrated with adaptive dynamic programming (ADP) to address the tracking control problem for continuous-time uncertain nonlinear systems with symmetric input constraints. The high-order nonlinear system is decomposed into multiple low-order subsystems by using the backstepping approach. In each step, the Lyapunov function is designed and the virtual control is derived. Then, the actual control is derived in the last step. By using the ADP approach, the Hamilton-Jacobi-Bellman equation is calculated to attain the optimal control. Under the identifier-critic-action architecture, unknown dynamics are estimated by utilizing neural networks. The ADP approach is implemented at each step to obtain the optimal virtual control law and the optimal actual control law while reducing the computational complexity and achieving the control objective. Meanwhile, the approximate saturation model is combined with optimal control to address the optimized backstepping control problem with symmetric saturation. Finally, a simulation case is used to confirm the efficacy of the current control technique.</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

Neural critic learning and optimal backstepping tracking control for unknown strict feedback systems with symmetric input constraints

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • Ding Wang,
  • Yingyun He,
  • Menghua Li,
  • Xinrui Ma,
  • Junfei Qiao

摘要

In this paper, the backstepping approach is integrated with adaptive dynamic programming (ADP) to address the tracking control problem for continuous-time uncertain nonlinear systems with symmetric input constraints. The high-order nonlinear system is decomposed into multiple low-order subsystems by using the backstepping approach. In each step, the Lyapunov function is designed and the virtual control is derived. Then, the actual control is derived in the last step. By using the ADP approach, the Hamilton-Jacobi-Bellman equation is calculated to attain the optimal control. Under the identifier-critic-action architecture, unknown dynamics are estimated by utilizing neural networks. The ADP approach is implemented at each step to obtain the optimal virtual control law and the optimal actual control law while reducing the computational complexity and achieving the control objective. Meanwhile, the approximate saturation model is combined with optimal control to address the optimized backstepping control problem with symmetric saturation. Finally, a simulation case is used to confirm the efficacy of the current control technique.