Neuro-adaptive command-filtered backstepping control of uncertain flexible-joint manipulators with input saturation...

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Flexible-joint manipulators (FJMs) present control challenges due to inherent uncertainties and input saturation.
• Existing output-feedback control methods struggle to address both system uncertainties and input saturation simultaneously.

Purpose of the Study:

• To develop a novel neuro-adaptive command-filtered output-feedback backstepping control scheme for FJMs.
• To address precise trajectory tracking under uncertainties, input saturation, and limited state feedback.

Main Methods:

• Utilized a smooth hyperbolic function and mean-value theorem to manage input saturation.
• Developed an adaptive radial basis function (RBF) neural network-based state observer for state estimation.
• Implemented a second-order command-filtered backstepping framework with error compensation to manage complexity and filtering errors.

Main Results:

• The proposed controller effectively mitigates input saturation and estimates unmeasurable states.
• Achieved uniform ultimate boundedness of all closed-loop signals, verified by Lyapunov stability analysis.
• Demonstrated superior transient and steady-state tracking accuracy compared to existing backstepping methods in simulations.

Conclusions:

• The novel neuro-adaptive control scheme successfully addresses limitations of prior methods for FJMs.
• The approach does not require precise manipulator models and ensures robust stability.
• Offers enhanced trajectory tracking performance for flexible-joint manipulators in complex operating conditions.