Deterministic learning-based neural control for output-constrained strict-feedback nonlinear systems.

Area of Science:

• Control Engineering
• Artificial Intelligence
• Nonlinear System Dynamics

Background:

• Output-constrained nonlinear systems present significant control challenges.
• Adaptive neural control offers a promising approach for handling system uncertainties.
• Existing methods often struggle with simultaneous constraint satisfaction and learning.

Purpose of the Study:

• To develop an adaptive neural control scheme for output-constrained strict-feedback uncertain nonlinear systems.
• To enable learning from the closed-loop control process for improved performance.
• To ensure system output tracks desired trajectories while respecting constraints.

Main Methods:

• State transformation to convert constrained output to unconstrained.
• System transformation to construct an equivalent affine nonlinear system.
• Dynamic Surface Control (DSC) with Radial Basis Function Neural Networks (RBF NNs).
• Development of a novel learning controller (LC) by reusing learned knowledge.

Main Results:

• Closed-loop signals are uniformly ultimately bounded.
• System output successfully tracks trajectories while satisfying constraints.
• Radial Basis Function Neural Networks precisely approximate uncertain dynamics.
• The learning control scheme enhances tracking accuracy and convergence speed.
• Online adaptation of neural weights is avoided, saving resources and improving transient performance.

Conclusions:

• The proposed adaptive neural learning control scheme effectively handles output constraints in uncertain nonlinear systems.
• The integration of DSC, RBF NNs, and a novel LC strategy ensures robust performance and efficient learning.
• The method offers significant advantages in terms of tracking accuracy, convergence speed, and computational efficiency.