Optimal Learning Control Scheme for Discrete-Time Systems With Nonuniform Trials.

Area of Science:

• Robotics and Control Systems
• Biomedical Engineering
• Machine Learning

Background:

• Repetitive control is crucial in rehabilitation training but faces challenges due to varying operational lengths.
• Existing control schemes may not effectively handle partially available information during learning processes.

Purpose of the Study:

• To investigate an intermittent optimal learning control scheme for applications requiring repetitive control with adaptable operational lengths.
• To address the challenge of varying iteration lengths in control systems by utilizing partially available information.

Main Methods:

• Developed an intermittent optimal learning control scheme.
• Defined a performance index to minimize tracking error and input drift.
• Utilized iteration-time-varying learning gain adjusted by historical data.
• Employed theoretical analysis to demonstrate convergence properties.

Main Results:

• The proposed scheme achieves asymptotic convergence of timewise tracking error along the iteration direction.
• Tracking error can vanish at a finite iteration under specific conditions (null amplifier).
• Numerical simulations on a permanent magnet linear motor model validated the scheme's effectiveness.

Conclusions:

• The intermittent optimal learning control scheme is effective for systems with iteration-varying operational lengths.
• The method demonstrates robust performance and convergence in practical scenarios like rehabilitation training.
• The approach offers a viable solution for adaptive and precise control in dynamic environments.