Pose Estimation of a Cobot Implemented on a Small AI-Powered Computing System and a Stereo Camera for Precision...

Area of Science:

• Robotics
• Computer Vision
• Machine Learning

Background:

• Robotic manipulators require high precision for industrial and medical applications, particularly in repetitive tasks.
• Geometric deformations commonly affect robotic manipulator accuracy.
• Existing 3D vision techniques help determine geometric distances but can be improved.

Purpose of the Study:

• To measure the angular position of a six-degrees-of-freedom robotic arm.
• To enhance the precision of robotic manipulators by compensating for geometric errors.
• To develop an intelligent visual measurement system as a complementary tool to encoders.

Main Methods:

• Utilized a stereo camera system for visual data acquisition.
• Implemented a convolutional neural network (CNN) algorithm for error compensation.
• Focused on reducing precision degradation caused by geometric inaccuracies.

Main Results:

• Achieved an accuracy of approximately one millimeter in camera testing.
• The proposed method demonstrated superior performance compared to traditional and basic neural network approaches.
• The intelligent visual system effectively compensated for geometric errors, enhancing overall accuracy.

Conclusions:

• The developed visual measurement system significantly improves robotic arm angular position accuracy.
• This intelligent system offers a viable method to enhance precision in robotic applications without replacing existing encoders.
• The integration of stereo cameras and CNNs presents a promising advancement in robotic vision for industrial and medical fields.