Fixed-Time Recurrent NN Learning Control of Uncertain Robotic Manipulators with Time-Varying Constraints:...

Area of Science:

• Robotics and Control Systems
• Machine Learning for Control
• Artificial Intelligence in Automation

Background:

• Robotic manipulators require precise dynamic tracking for complex tasks.
• Traditional control methods struggle with unknown dynamics and external disturbances.
• Existing approaches often lack guarantees for fixed-time convergence and output constraints.

Purpose of the Study:

• To develop a learning control framework for robotic manipulators.
• To achieve fixed-time convergence of tracking errors under output constraints.
• To address unknown manipulator dynamics and external disturbances using online learning.

Main Methods:

• Proposed a dynamic surface control (DSC) framework integrating a barrier Lyapunov function (BLF) and a recurrent neural network (RNN).
• Introduced a time-varying tangent-type BLF for fixed-time virtual controller design.
• Employed an RNN-based online approximator to compensate for unknown system dynamics and external disturbances.

Main Results:

• Guaranteed fixed-time convergence of tracking errors to small neighborhoods around the origin.
• Ensured that the manipulator's actual trajectories remain within prescribed output constraints.
• Demonstrated improved tracking accuracy and effective online estimation of unknown dynamics via experimental results.

Conclusions:

• The proposed fixed-time, output-constrained neural learning controller effectively enhances robotic manipulator tracking performance.
• The integration of BLF and RNN provides a robust solution for unknown dynamics and disturbances.
• The framework offers a promising approach for high-precision robotic control applications.