Neural networks adaptive predefined-time control for pure-feedback nonlinear systems: a case study on robotic...

Area of Science:

• Robotics
• Control Theory
• Artificial Intelligence

Background:

• Non-affine pure-feedback nonlinear systems present significant control challenges.
• Existing control methods often lack guaranteed convergence times or struggle with state constraints.
• Robotic exoskeletons require precise and rapid control for effective human-assistance applications.

Purpose of the Study:

• To develop a novel predefined-time (PT) tracking adaptive control method for non-affine pure-feedback nonlinear systems.
• To apply this control method to robotic exoskeleton technology, addressing state constraints.
• To enhance system convergence speed and performance over conventional fixed-time control.

Main Methods:

• Implementation of a PT neural networks control algorithm.
• Utilizing neural network approximation for unknown nonlinearities.
• Employing backstepping technique, barrier functions, and Mean Value Theorem.
• Designing an adaptive law based on the PT Lyapunov stability criterion.

Main Results:

• The proposed methodology guarantees system convergence within a pre-established time.
• Neural networks effectively approximate system nonlinearities.
• The control approach demonstrates enhanced performance over fixed-time methods.
• Simulation results validate the control efficacy for robotic exoskeletons under state constraints.

Conclusions:

• The developed PT neural networks control algorithm offers a robust solution for controlling nonlinear systems, particularly robotic exoskeletons.
• The method provides guaranteed convergence within a predefined time, outperforming traditional control strategies.
• The approach shows significant potential for real-world applications requiring fast and constrained control.