Adaptive dynamic surface control of flexible-joint robots using self-recurrent wavelet neural networks

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Flexible-joint (FJ) robots present significant control challenges due to inherent model uncertainties in both robot and actuator dynamics.
• Traditional control methods often struggle with the 'explosion of complexity' inherent in backstepping approaches for such systems.

Purpose of the Study:

• To propose a new robust control strategy for FJ robots that effectively handles model uncertainties.
• To improve the position tracking performance and robustness of FJ robots against payload variations and external disturbances.

Main Methods:

• A hybrid control system integrating adaptive dynamic surface control (DSC) with self-recurrent wavelet neural networks (SRWNN).
• SRWNNs are employed for online observation and compensation of arbitrary model uncertainties in FJ robots.
• Lyapunov stability analysis is used to derive adaptation laws and prove the uniform ultimate boundedness of the closed-loop system.

Main Results:

• The proposed adaptive DSC-SRWNN control system demonstrates effective handling of model uncertainties.
• Simulation results for a three-link FJ robot validate superior position tracking performance.
• The control system exhibits significant robustness against payload uncertainties and external disturbances.

Conclusions:

• The combined adaptive DSC and SRWNN approach offers a robust and effective solution for controlling FJ robots with model uncertainties.
• The method successfully addresses the complexity issues of backstepping controllers while ensuring system stability.
• The validated performance highlights the potential of this control strategy for real-world robotic applications.