A new method for the control of discrete nonlinear dynamic systems using neural networks

Area of Science:

• Control Systems Engineering
• Nonlinear Dynamics
• Artificial Intelligence in Control

Background:

• Nonaffine nonlinear dynamic systems present significant control challenges.
• Existing methods often struggle with system complexity and performance limitations.

Purpose of the Study:

• To introduce a new, implementable controller design method for nonaffine nonlinear dynamic systems.
• To leverage identified neural network models for enhanced control performance.

Main Methods:

• Development of a novel control law applicable to discrete, deterministic, time-invariant nonlinear systems.
• Utilizing an identified neural network model of the nonlinear plant.
• Employing a real-time linearization of the plant's input-output model.

Main Results:

• The proposed control law is effective for systems within a specific subregion of stability.
• Control performance is robust and not solely dependent on proximity to the equilibrium state.
• The method successfully controls both minimum phase and nonminimum phase nonaffine plants.
• Empirical studies validate the controller's applicability to a general class of highly nonlinear multi-input-multi-output (MIMO) plants.

Conclusions:

• The presented controller design method offers a simplified and effective solution for complex nonaffine nonlinear systems.
• The use of neural network models and real-time linearization enhances control robustness and applicability.
• This approach demonstrates significant potential for controlling a wide range of challenging dynamic systems.