A robust position/force learning controller of manipulators via nonlinear Hinfinity control and neural networks

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Robotic manipulators operating in uncertain environments require sophisticated control for simultaneous position and force regulation.
• Existing control schemes may be conservative or struggle with complex dynamics and environmental interactions.

Purpose of the Study:

• To develop a novel robust learning controller for uncertain constrained manipulators.
• To integrate adaptive neural network (NN) techniques with nonlinear H-infinity control.
• To improve tracking performance and reduce conservativeness in position and force control.

Main Methods:

• Constructing task-space reduced-order position dynamics and algebraic force descriptions.
• Employing adaptive neural networks (NNs) to learn uncertain manipulator dynamics and environmental interactions.
• Utilizing nonlinear H-infinity control to attenuate NN approximation errors and guarantee performance.

Main Results:

• The proposed controller adaptively learns system uncertainties, enhancing tracking accuracy.
• The integrated approach demonstrably outperforms H-infinity-only control schemes in simulations.
• Analytical proof confirms the reduced conservativeness of the adaptive H-infinity NN controller.

Conclusions:

• The adaptive H-infinity NN controller offers superior performance for simultaneous position and force control of uncertain constrained manipulators.
• This approach effectively handles structured/unstructured uncertainties and environmental modeling issues.
• The controller provides a less conservative and more robust solution compared to traditional methods.