Neuro-Adaptive-Based Fixed-Time Composite Learning Control for Manipulators With Given Transient Performance.

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Manipulators face dynamic uncertainties and position errors.
• Existing control methods may struggle with unknown system dynamics and initial condition constraints.

Purpose of the Study:

• To develop an adaptive neural network (NN) control technique for manipulators.
• To achieve fixed-time tracking performance under constrained position error.
• To address dynamic uncertainties using composite learning.

Main Methods:

• Integrating composite learning into neural networks (NNs) for adaptive weight updating.
• Utilizing relaxed interval excitation (IE) conditions for NN weight adaptation.
• Employing error transfer functions and asymmetric barrier functions to manage position errors.
• Applying fixed-time control methodology and Lyapunov stability criteria for guaranteed convergence.

Main Results:

• The proposed NN composite learning technique effectively addresses dynamic uncertainties in manipulators.
• Fixed-time convergence of system signals to a compact neighborhood around the origin is guaranteed.
• The control strategy demonstrates robust performance in both steady and transient states for position error.
• Numerical simulations and experiments with the Baxter robot validate the control approach.

Conclusions:

• The adaptive NN control with composite learning provides a viable solution for fixed-time tracking in manipulators.
• The method successfully handles position errors and dynamic uncertainties.
• The study confirms the efficacy of the proposed technique through simulations and real-world experiments.