Neuro-adaptive force/position control with prescribed performance and guaranteed contact maintenance

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Robot force/position tracking faces challenges with contact maintenance, overshoot, response speed, and accuracy.
• Uncertainties in force deformation models and joint disturbances complicate control objectives.

Purpose of the Study:

• To develop a robust controller for robot force/position tracking that concurrently addresses contact maintenance, overshoot, response speed, and accuracy.
• To manage uncertainties in force deformation models and external disturbances.

Main Methods:

• Utilized a neural network linear in the weights to approximate unknown nonlinearities arising from force deformation model uncertainties.
• Developed a control strategy that guarantees the neural network approximation holds for all time.

Main Results:

• The proposed controller ensures concurrent satisfaction of contact maintenance, lack of overshoot, desired response speed, and accuracy.
• The controller demonstrates robustness against modeling errors, disturbances, and controller gains.
• Simulation results on a 6-DOF robot validate the theoretical findings.

Conclusions:

• The neural network-based approach effectively handles uncertainties and disturbances in robot control.
• The controller design simplifies gain selection and allows for tolerance of significant approximation errors and disturbances.
• This method enhances the reliability and performance of robotic systems in complex tasks.