Scholarlink

<p>For wing-rock dynamic nonlinear systems with time-varying parameters and external disturbances, the uncertainties of time-varying parameters, unmodeled dynamics, and external disturbances are treated as lumped uncertainty. To address the limitations of existing single-estimator approaches, this paper proposes a novel multilateral learning adaptive mechanism and two corresponding multilateral learning state observers. A multilateral learning adaptive mechanism and multilateral learning state observers are designed. The multilateral learning adaptive mechanism includes multiple estimators for the lumped uncertainty, which are fused through linear weighting parameters to form the adaptive compensation term in the controller. This multi-branch architecture enables the system to simultaneously explore multiple potential uncertainty hypotheses, significantly improving both transient and steady-state estimation accuracy. Considering that the initial value of the lumped uncertainty follows a uniform distribution within a possible range, an initial value allocation strategy for the multilateral uncertainty estimators is designed. Adaptive update laws for all uncertainty estimators and multilateral weighting parameters are developed. To prevent excessive fluctuations in the weighting parameters, a saturation transfer function is introduced to constrain their rate of change. The estimated lumped uncertainty output by multilateral learning is used as a compensation term, replacing the extended state of the linear extended state observer (LESO) to form the first type of multilateral learning state observer, termed the imitate linear extended state observer (ILESO). Considering the sensitivity of the multilateral learning mechanism to high gains, a low-gain observer (LGO) is designed as the second type of multilateral learning state observer, which achieves superior tracking accuracy with substantially reduced gain requirements. The proposed control method is compared with existing approaches. Simulation results demonstrate that the MLOBC-LGO controller achieves the lowest cumulative tracking errors. The exponential convergence of the closed-loop system is rigorously proved through Lyapunov stability analysis.</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

A novel multilateral learning state observation controller for wing-rock nonlinear dynamic systems

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月24日

查看全文
  • Guangwei Chen,
  • Zengliang Zhang,
  • Xiaohong Yang

摘要

For wing-rock dynamic nonlinear systems with time-varying parameters and external disturbances, the uncertainties of time-varying parameters, unmodeled dynamics, and external disturbances are treated as lumped uncertainty. To address the limitations of existing single-estimator approaches, this paper proposes a novel multilateral learning adaptive mechanism and two corresponding multilateral learning state observers. A multilateral learning adaptive mechanism and multilateral learning state observers are designed. The multilateral learning adaptive mechanism includes multiple estimators for the lumped uncertainty, which are fused through linear weighting parameters to form the adaptive compensation term in the controller. This multi-branch architecture enables the system to simultaneously explore multiple potential uncertainty hypotheses, significantly improving both transient and steady-state estimation accuracy. Considering that the initial value of the lumped uncertainty follows a uniform distribution within a possible range, an initial value allocation strategy for the multilateral uncertainty estimators is designed. Adaptive update laws for all uncertainty estimators and multilateral weighting parameters are developed. To prevent excessive fluctuations in the weighting parameters, a saturation transfer function is introduced to constrain their rate of change. The estimated lumped uncertainty output by multilateral learning is used as a compensation term, replacing the extended state of the linear extended state observer (LESO) to form the first type of multilateral learning state observer, termed the imitate linear extended state observer (ILESO). Considering the sensitivity of the multilateral learning mechanism to high gains, a low-gain observer (LGO) is designed as the second type of multilateral learning state observer, which achieves superior tracking accuracy with substantially reduced gain requirements. The proposed control method is compared with existing approaches. Simulation results demonstrate that the MLOBC-LGO controller achieves the lowest cumulative tracking errors. The exponential convergence of the closed-loop system is rigorously proved through Lyapunov stability analysis.