Improve 3D Modeless Robots Calibration Accuracy Using A Shallow Neural Network Technique

Abstract

This paper is an extended research for a novel technique used in the pose error compensations of the robot and manipulator calibration process based on 3D Shallow Neural Network Technique (SNNT). Robot calibrations can be classified into model-based and modeless methods. A model-based calibration method normally requires that the practitioners understand the kinematics of the robot therefore may pose a challenge for field engineers. An alternative yet effective means for robot calibration is to use a modeless method; however with such a method there is a conflict between the calibration accuracy of the robot and the number of grid points used in the calibration task. In this paper, an interpolation method developed based on a SNNT is applied to improve the compensation accuracy of the robot in its 3D workspace. The simulated results given in this paper show that not only robot compensation accuracy can be greatly improved with this method, but also the calibration process can be significantly simplified, and it is more suitable for practical applications.