Fast and Smooth Composite Local Learning-Based Adaptive Control

Tao Jiang

doi:10.1109/TNNLS.2021.3130812

2023-09-01

PRE

Tao Jiang

黄

黄江帅 (Jiangshuai Huang)

苏

苏晓杰 (Xiaojie Su) *

DOI：10.1109/TNNLS.2021.3130812

原文链接

原文求助

摘要

En 中文

Model structure representation and fast estimation of perturbations are two key research aspects in adaptive control. This work proposes a composite local learning adaptive control framework, which possesses fast and flexible approximation to system uncertainties and meanwhile smoothens control inputs. Local learning, which is a nonparametric regression approach, is able to automatically adjust the structure of approximator based on data distribution from the local region, but it is sensitive to the outliers and measurement noises. To tackle this problem, the regression filter technique is employed to attenuate the adverse effect of noises by smoothing the output response and state features. In addition, the stable integral adaptation is integrated into local learning framework to further enhance the system robustness and smoothness of the estimation. Through the online elimination of uncertainties, the nominal control performance is recovered when the plant encounters violent perturbations. Stability analysis and numerical simulations are performed to demonstrate the effectiveness and benefits of the proposed control method. The proposed approach exhibits a promising performance in terms of rapid perturbation elimination and accurate tracking control.

Keyword:

Perturbation methods

Uncertainty

Adaptive control

Estimation

Smoothing methods

Adaptation models

Robustness

Adaptive control

composite learning

local learning

nonparametric representation