Scholarlink

<p>Achieving both rapid transient response and high steady-state accuracy in disturbance estimation for nonlinear systems remains challenging: model-based observers offer fast transient response but limited steady-state precision, whereas data-driven estimators achieve high steady-state accuracy but suffer from the cold-start issue. Motivated by this challenge, this paper develops a hybrid nonlinear disturbance estimation and rejection framework based on an interval type-2 self-organizing fuzzy neural network (IT2SOFNN). Our key insight is that the IT2SOFNN learns the mismatched time-varying components of disturbances that cannot be effectively captured by the nonlinear disturbance observer (NDOB). Specifically, the NDOB rapidly estimates the dominant part of disturbances during the cold-start phase, while the IT2SOFNN models the remaining residuals through self-organizing fuzzy structures, thereby improving the framework’s capability to handle mismatched time-varying disturbances. Simulations on a chaotic system and experiments on a 6-DOF robotic manipulator demonstrate significant improvements over representative comparison methods in both transient response and steady-state accuracy.</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

Hybrid disturbance estimation and rejection for nonlinear systems based on a self-organizing fuzzy neural network

  • 期刊论文

  • 发表日期：

    2026年7月17日

查看全文
  • Jiayu Zhao,
  • Rui Zhou,
  • Tao Zhao

摘要

Achieving both rapid transient response and high steady-state accuracy in disturbance estimation for nonlinear systems remains challenging: model-based observers offer fast transient response but limited steady-state precision, whereas data-driven estimators achieve high steady-state accuracy but suffer from the cold-start issue. Motivated by this challenge, this paper develops a hybrid nonlinear disturbance estimation and rejection framework based on an interval type-2 self-organizing fuzzy neural network (IT2SOFNN). Our key insight is that the IT2SOFNN learns the mismatched time-varying components of disturbances that cannot be effectively captured by the nonlinear disturbance observer (NDOB). Specifically, the NDOB rapidly estimates the dominant part of disturbances during the cold-start phase, while the IT2SOFNN models the remaining residuals through self-organizing fuzzy structures, thereby improving the framework’s capability to handle mismatched time-varying disturbances. Simulations on a chaotic system and experiments on a 6-DOF robotic manipulator demonstrate significant improvements over representative comparison methods in both transient response and steady-state accuracy.