Learning-based cable coupling effect modeling for robotic manipulation of heavy industrial cables

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Robotic manipulation of heavy industrial cables presents significant challenges due to complex dynamics and high degrees of freedom.
• Existing methods struggle with accurate modeling and control of rigid-flexible coupling inherent in cable systems.

Purpose of the Study:

• To develop an accurate and efficient model for cable effects in robotic manipulation.
• To create a robust control methodology for high-quality robotic cable handling.
• To address practical limitations like measurement constraints and time efficiency for real-world applications.

Main Methods:

• A 2D convolutional neural network (CNN) was employed for an effective cable representation, enabling estimation of cable coupling forces and torques.
• A control strategy integrating active disturbance rejection control (ADRC) with sliding mode control (SMC) was developed.
• The cable effect model was integrated into the control scheme for enhanced robotic manipulation.

Main Results:

• The developed model achieved over 85% accuracy in estimating cable coupling effects.
• Robotic manipulation tasks were accomplished with a positioning error of less than 0.01 mm.
• The method successfully addressed the dynamic payload effect of heavy industrial cables in experimental settings.

Conclusions:

• The proposed deep learning-based cable effect model and integrated control methodology offer a promising solution for complex robotic cable manipulation.
• The system demonstrates high-quality performance in simulated and real-world robotic systems, including challenging cable insertion tasks.
• This research overcomes previous limitations in modeling and controlling heavy industrial cables, paving the way for advanced robotic applications.