Bat optimization of hybrid neural network-FOPID controllers for robust robot manipulator control

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Robot manipulator control faces challenges with accuracy and stability due to unknown loads and disturbances.
• Existing control methods struggle with signal chattering in multi-input, multi-output systems.

Purpose of the Study:

• To propose and evaluate three hybrid control structures for a three-link rigid robot manipulator (3-LRRM).
• To address issues of accuracy, stability, and signal chattering in robot manipulator control.
• To compare the performance of novel NN-FOPID hybrid controllers.

Main Methods:

• Development of three hybrid control schemes: NN-FOPIPD, NN + FOPID, and ELNN-FOPID.
• Application of the Bat Optimization Algorithm (BOA) to tune controller parameters by minimizing Integral Time Square Error (ITSE).
• Simulation analysis using MATLAB to evaluate controller performance, robustness against uncertainties, and disturbances.

Main Results:

• The NN-FOPIPD controller exhibited the best performance among the proposed hybrid control schemes.
• All proposed controllers showed robustness against system parameter uncertainties, external disturbances, and initial position variations.
• The study successfully mitigated signal chattering issues in the 3-LRRM system.

Conclusions:

• Hybrid control structures combining neural networks and fractional order PID controllers offer significant improvements in robot manipulator control.
• The NN-FOPIPD structure is highly effective for precise and stable trajectory tracking of rigid-link robot manipulators.
• The proposed methods provide a robust solution for real-world robotic applications facing unpredictable conditions.