Reference-shaping adaptive control by using gradient descent optimizers.

Area of Science:

• Control Engineering
• Adaptive Systems
• Optimization Techniques

Background:

• Model reference adaptive control (MRAC) is crucial for systems with uncertainties.
• Existing MRAC methods may face challenges with time-varying parameters and noise.
• Gradient descent optimization offers a robust framework for adaptive control.

Purpose of the Study:

• To propose a novel Reference-Shaping Adaptive Control (RSAC) scheme.
• To enhance control signal generation and adaptation gain estimation.
• To improve system performance and reduce noise sensitivity.

Main Methods:

• Implementing a model reference adaptive control scheme using a reference-shaping approach.
• Utilizing two gradient descent optimization processes: a control optimizer and an adaptation optimizer.
• Estimating time-varying adaptation gain and applying a dead zone rule to mitigate noise.

Main Results:

• The control optimizer generates effective control signals for output tracking.
• The adaptation optimizer improves control signal generation through gain estimation.
• Numerical simulations validated the RSAC method's performance across various scenarios.
• Experimental results confirmed the method's applicability to rotor control.

Conclusions:

• The proposed RSAC method effectively achieves system output tracking via gradient descent optimization.
• The integration of a dead zone rule enhances robustness against noise.
• The RSAC approach shows promise for real-world applications, as evidenced by rotor control experiments.