Scholarlink

<p>This paper proposes a novel neural network-based multilateral learning adaptive control scheme within a model reference adaptive framework aimed at improving control accuracy and transient response for affine nonlinear systems with uncertainties. Recognizing the significant impact of initial uncertainty estimates, neural network bias nodes are initialized based on either offline data distribution characteristics or a defined boundary uniform distribution strategy. The proposed controller employs a combined feedforward-feedback structure enhanced with an adaptive term addressing system uncertainties, constituting a multilateral learning adaptive controller. This multilateral learning mechanism integrates outputs from multiple neural networks through linear combinations to effectively approximate and compensate for uncertainties. Additionally, a second-order linear filter provides differential signals of tracking errors, indirectly estimating model uncertainties. To prevent parameter oscillations due to rapid variations, a saturation transfer function constrains the multilateral learning weight update rate within predefined boundaries. Stability of the closed-loop system is rigorously established via Lyapunov analysis, and the effectiveness of the controller design is demonstrated through validation on an inverted pendulum system.</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

Model Reference Neural Adaptive Controller Design With Multilateral Learning Mechanism for Uncertain Nonlinear Systems

  • 期刊论文

  • 发表日期：

    2025年4月12日

查看全文
  • Qunpo Liu,
  • Jiakun Li,
  • Xuhui Bu,
  • Jianjun Zhang,
  • Naohiko Hanajima

摘要

This paper proposes a novel neural network-based multilateral learning adaptive control scheme within a model reference adaptive framework aimed at improving control accuracy and transient response for affine nonlinear systems with uncertainties. Recognizing the significant impact of initial uncertainty estimates, neural network bias nodes are initialized based on either offline data distribution characteristics or a defined boundary uniform distribution strategy. The proposed controller employs a combined feedforward-feedback structure enhanced with an adaptive term addressing system uncertainties, constituting a multilateral learning adaptive controller. This multilateral learning mechanism integrates outputs from multiple neural networks through linear combinations to effectively approximate and compensate for uncertainties. Additionally, a second-order linear filter provides differential signals of tracking errors, indirectly estimating model uncertainties. To prevent parameter oscillations due to rapid variations, a saturation transfer function constrains the multilateral learning weight update rate within predefined boundaries. Stability of the closed-loop system is rigorously established via Lyapunov analysis, and the effectiveness of the controller design is demonstrated through validation on an inverted pendulum system.