Work

Excitation‐Oriented Recursive Learning Control

Year: 2025

Type: article

Abstract: ABSTRACT To address the severe dependence on the high‐level persistent excitation (PE) condition and the performance deficiencies in traditional stochastic gradient descent‐based neural network learning control (SGD‐NNLC), which is grounded in deterministic learning theory, we proposed a novel online learning framework for neural networks named excitation‐oriented recursive learning control (EORLC). EORLC employs excitation‐oriented forgetting recursive least squares (EOFRLS) to guide the weight update laws of radial basis function neural networks (RBFNNs). The forgetting factor is allocated based on the PE level and error at each time point along the training trajectory, thereby enabling the RBFNN to achieve superior learning performance. Finally, this paper theoretically proves the exponential stability of the closed‐loop system of EORLC under the PE condition. Experimental validation conducted on a computer numerical control machine tool confirms the superiority of this learning control algorithm over SGD‐NNLC in terms of learning performance.

Source: International Journal of Robust and Nonlinear Control

DOI: 10.1002/rnc.70344

Authors: Juncheng Xu, Lukai Bin, Zhiyu Hu, Jiangang Li, Yiming Fei, Yanan Li

Institutions: Harbin Institute of Technology, Zhejiang University, University of Sussex

Language: English

Topic: Neural Networks and Applications

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

Funders: National Natural Science Foundation of China, Shenzhen Science and Technology Innovation Program

Awards: G1542800931, G8189970752