现代防御技术 ›› 2026, Vol. 54 ›› Issue (1): 156-164.DOI: 10.3969/j.issn.1009-086x.2026.01.016

• 论文 • 上一篇    

基于RBF神经网络的二阶不确定系统自适应滑模控制

马强1, 张杨2, 杨珂3   

  1. 1.平顶山工业职业技术学院 机电工程学院,河南 平顶山 467000
    2.广东理工学院 信息技术学院,广东 肇庆 526100
    3.河南平高通用电气公司生产物资部,河南 平顶山 467000
  • 收稿日期:2024-12-23 修回日期:2025-06-09 出版日期:2026-01-28 发布日期:2026-02-11
  • 通讯作者: 张杨
  • 作者简介:马强(1982-),男,河南平顶山人。讲师,学士,研究方向为机电一体化技术。

Adaptive Sliding-mode Control for Second-Order Uncertain Systems Based on an RBF Neural Network

Qiang MA1, Yang ZHANG2, ke YANG3   

  1. 1.School of Mechanical and Electrical Engineering,Pingdingshan Polytechnic College,Pingdingshan 467000,China
    2.School of Information Technology,Guangdong Technology College,Zhaoqing 526100,China
    3.Production Materials Department,Henan Pinggao General Electric Company,Pingdingshan 467000,China
  • Received:2024-12-23 Revised:2025-06-09 Online:2026-01-28 Published:2026-02-11
  • Contact: Yang ZHANG

摘要:

针对二阶不确定系统,特别是模型未知且伴随动力学扰动的复杂情况,以提升控制效能为目标展开研究。以板球系统为实验对象,提出了一种新颖的控制策略。采用RBF神经网络(RBF1)预测系统关键参数,并通过自适应算法动态调整其内部参数以确保预测精度;基于预测模型设计了一种基于积分滑模面的滑模控制器,利用积分滑模面的特性使系统状态直接进入滑动模态,提高了系统的鲁棒性和响应速度。为进一步优化控制性能,创新性地引入第2个RBF神经网络(RBF2)来动态调整滑模控制器参数,通过梯度下降法实现参数的整定,增强了控制策略的灵活性和适应性。仿真实验表明,该控制策略在板球系统轨迹跟踪中表现优异,能够有效应对系统不确定性和扰动,展现了良好的控制性能和实际应用前景。

关键词: 二阶系统, 滑模控制, RBF神经网络, 梯度下降法, 板球控制系统

Abstract:

This study aims to enhance the control performance of second-order uncertain systems, especially those characterized by unknown models and accompanied by dynamic disturbances. A novel control strategy was proposed using the ball-and-plate system as the experimental object. An RBF neural network (RBF1) was used to predict the key parameters of the system, and its internal parameters were dynamically adjusted through adaptive algorithms to ensure prediction accuracy. Based on this predictive model, a sliding mode controller with an integral sliding surface was designed, enabling the system states to reach the sliding manifold directly and thereby improving the robustness and response speed. To further optimize control performance, this paper innovatively introduced a second RBF neural network (RBF2) to dynamically adjust the sliding mode controller parameters, which were tuned via the gradient descent method, enhancing the flexibility and adaptability of the control strategy. Simulation experiments show that this control strategy performs well in trajectory tracking of ball-and-plate systems, effectively dealing with system uncertainties and disturbances, which confirms its favorable control performance and practical application prospects.

Key words: second-order system, sliding mode control, RBF neural network, gradient descent method, board and ball control system

中图分类号: