Peaking-Free Output-Feedback Adaptive Neural Control Under a Nonseparation Principle

Area of Science:

• Control Systems Engineering
• Nonlinear System Analysis
• Artificial Intelligence in Control

Background:

• High-gain observers are common in adaptive neural control (ANC) for uncertain nonlinear systems but suffer from peaking and noise sensitivity.
• Existing adaptive neural network (NN) observers struggle to overcome these high-gain observer limitations.
• A nonlinear separation principle is often used, but this paper explores a nonseparation principle.

Purpose of the Study:

• To develop an output-feedback indirect ANC strategy that overcomes the limitations of high-gain observers.
• To propose a hybrid estimation scheme integrating an adaptive NN observer with state variable filters.
• To ensure practical asymptotic stability without peaking responses or control saturation.

Main Methods:

• A hybrid estimation scheme combining an adaptive NN observer and state variable filters to estimate system states.
• An indirect adaptive neural control (ANC) strategy implemented under a nonseparation principle.
• Analysis using a single Lyapunov function candidate to prove system stability.

Main Results:

• The proposed approach achieves practical asymptotic stability for the closed-loop system.
• The system demonstrates avoidance of peaking responses and control saturation.
• The method exhibits favorable noise rejection capabilities, outperforming existing techniques.

Conclusions:

• The novel hybrid estimation scheme effectively addresses limitations of high-gain observers in ANC.
• The developed indirect ANC strategy offers improved stability, noise rejection, and robustness.
• This approach provides a superior alternative for controlling uncertain nonlinear systems.