A dual neural network for constrained joint torque optimization of kinematically redundant manipulators

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Kinematically redundant manipulators require advanced control strategies for efficient operation.
• Minimizing global kinetic energy is crucial for optimizing robot mechanism performance.
• Existing inverse kinematics methods face challenges with redundancy resolution.

Purpose of the Study:

• To introduce a dual neural network for real-time joint torque optimization.
• To address the challenge of resolving kinematic redundancy in limited-joint-range manipulators.
• To achieve global kinetic energy minimization in robot mechanisms.

Main Methods:

• Development of a dual neural network architecture at the acceleration level.
• Utilizing a single-layer neural network for computational efficiency.
• Proving global exponential convergence to optimal solutions.

Main Results:

• The dual neural network effectively resolves redundancy in kinematically redundant manipulators.
• Demonstrated real-time joint torque optimization capabilities.
• Simulation results with the PUMA 560 robot arm confirmed the network's effectiveness.

Conclusions:

• The proposed dual neural network offers an efficient solution for real-time torque optimization.
• The method successfully addresses kinematic redundancy, improving robotic control.
• The simple yet effective architecture ensures convergence to optimal solutions.