Experimental Analysis of Robot Hybrid Calibration Based on Geometrical Identification and Artificial Neural Network

Abstract

Industrial robots are known to have good repeatability and poor accuracy. However, accuracy can be improved through calibration process. Different methods of calibration can be found in the literature. In this paper, a hybrid calibration approach was applied to improve the accuracy of a lightweight collaborative robot. The approach is based on an analytical model to compensate geometric errors and on an artificial neural network to compensate residual errors (stiffness, gear errors,…. etc). The suggested approach is analysed and optimised in the work. The approach can reduce the positioning error from 3.10mm to 0.13mm on a lightweight collaborative robot in a specific sub-workspace.