On-line identification of nonlinear systems using Volterra polynomial basis function neural networks

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Nonlinear control systems present challenges in accurate modeling due to their complex dynamics.
• Traditional identification methods may struggle to adapt to real-time variations in system parameters and operating points.

Purpose of the Study:

• To develop an efficient on-line identification scheme for nonlinear control systems.
• To enable adaptive modeling using Volterra polynomial basis function (VPBF) neural networks.
• To ensure the stability and convergence of the identification process.

Main Methods:

• Utilized Volterra polynomial basis function (VPBF) neural networks for system identification.
• Implemented an orthogonal least-squares algorithm for off-line structure selection.
• Employed a growing network technique for on-line structure selection.
• Developed a recursive weight learning algorithm for adaptive parameter tuning.
• Applied Lyapunov techniques to establish convergence of weights and estimation errors.

Main Results:

• Successfully demonstrated an on-line identification scheme for nonlinear systems.
• The proposed method effectively adapts to variations in system characteristics and operating points.
• Convergence of network weights and estimation errors was theoretically established.
• Simulated examples validated the efficacy of the identification procedure.

Conclusions:

• The developed on-line identification scheme using VPBF neural networks provides an adaptive and robust solution for nonlinear control systems.
• The combination of off-line and on-line structure selection with recursive weight learning ensures accurate and stable system identification.
• This approach holds potential for real-time applications requiring dynamic model adaptation.