Related Experiment Videos
Direct adaptive iterative learning control of nonlinear systems using an output-recurrent fuzzy neural network
Ying-Chung Wang1, Chiang-Ju Chien, Ching-Cheng Teng
1Department of Electrical and Control Engineering, National Chiao-Tung University, Hsinchu, Taiwan, ROC.
Summary
This study introduces a direct adaptive iterative learning control (DAILC) using an output-recurrent fuzzy neural network (ORFNN) for nonlinear systems. The novel approach effectively handles unknown nonlinearities and initial errors, ensuring precise tracking performance.
Area of Science:
- Control Engineering
- Artificial Intelligence
- Nonlinear System Dynamics
Background:
- Iterative learning control (ILC) is effective for repeatable tasks but struggles with initial errors and unknown nonlinearities.
- Fuzzy neural networks offer powerful function approximation capabilities for complex systems.
Purpose of the Study:
- To develop a direct adaptive iterative learning control (DAILC) strategy for nonlinear systems with unknown dynamics and variable initial conditions.
- To enhance tracking accuracy and ensure convergence in the presence of system uncertainties.
Main Methods:
- A novel output-recurrent fuzzy neural network (ORFNN) is employed to approximate the optimal equivalent controller.
- A time-varying boundary layer concept is introduced to manage initial state errors.
- An adaptive algorithm with a projection mechanism is derived for updating ORFNN parameters (consequent, premise, recurrent).
Main Results:
- The proposed DAILC scheme effectively handles unknown nonlinearities and nonlinear input gain.
- Lyapunov-like analysis confirms that all parameters and signals remain bounded.
- The state tracking error norm converges asymptotically to a tunable residual set.
Conclusions:
- The ORFNN-based DAILC provides a robust and efficient control solution for repeatable nonlinear systems.
- The method demonstrates superior tracking performance on benchmark systems like the inverted pendulum and Chua's circuit.