Adaptive Quantized Iterative Learning Control Using Encoding-Decoding Strategy.

Area of Science:

• Control Engineering
• Signal Processing
• Machine Learning

Background:

• Iterative learning control (ILC) is effective for repetitive tasks.
• Quantization in ILC can lead to performance degradation and saturation.
• Existing adaptive ILC methods often have restrictive initial conditions.

Purpose of the Study:

• To develop a dynamic encoding-decoding mechanism for adaptive quantized iterative learning control.
• To enable adaptive adjustment of quantization parameters for improved ILC performance.
• To relax constraints on initial input signals and reduce quantizer saturation.

Main Methods:

• Designed dynamic encoding-decoding pairs for error and output signals.
• Employed a uniform quantizer with finite levels and specified distinct lower bounds.
• Incorporated zoom-out and zoom-in strategies in the encoder and decoder for quantizer adaptation.

Main Results:

• The proposed adaptive quantization mechanisms ensure system output convergence to the reference.
• The scheme prevents quantizer saturation under any initial input condition.
• Constraints on initial inputs are relaxed, and the quantizer saturation bound is simplified and reduced.

Conclusions:

• The novel adaptive quantized iterative learning control scheme effectively addresses performance limitations.
• The dynamic encoding-decoding approach enhances robustness and simplifies implementation.
• Validated through numerical simulations and experimental results, demonstrating practical applicability.