Model-Reconstructed RBFNN-DOB for FJR Trajectory Control with External Disturbances.

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Flexible Joint Robots (FJRs) face challenges in precise control due to parameter uncertainties and external disturbances.
• Traditional Disturbance Observers (DOBs) require accurate dynamic models, which are difficult to achieve for FJRs, limiting their effectiveness.
• Unmodeled dynamics and external disturbances hinder robust and efficient control in industrial robotic applications.

Purpose of the Study:

• To develop a robust control strategy for FJRs that addresses both parameter uncertainties and external time-varying disturbances.
• To enhance the precision and reliability of FJR control in industrial settings.
• To improve the computational efficiency of disturbance compensation in robotic systems.

Main Methods:

• A hybrid controller combining a nominal model-based Disturbance Observer (DOB) with a Radial Basis Function Neural Network (RBFNN) was designed.
• An adaptive RBFNN weight update law was formulated using Lyapunov stability theory for selective estimation of unmodeled dynamics.
• The controller was experimentally validated on an FJR system.

Main Results:

• The proposed RBFNN-based DOB effectively compensated for unmodeled dynamics and external disturbances in FJRs.
• The adaptive law enabled selective estimation, minimizing computational load while ensuring robust compensation.
• Experimental results confirmed the efficacy of the hybrid controller in enhancing control precision and handling rapidly changing disturbances.

Conclusions:

• The hybrid RBFNNDOB state feedback controller offers a robust and computationally efficient solution for controlling FJRs with uncertainties and disturbances.
• The Lyapunov-based adaptive approach ensures stability and accurate tracking error boundedness.
• This method significantly improves the applicability and performance of DOBs in complex robotic systems.