Complete calibration of industrial robot with limited parameters and neural network

Abstract

Poor accuracy of industrial robots cannot meet the requirements of some high precision assignments, especially in heavy load condition. Geometric errors and joint compliances are mainly responsible for this poor accuracy. This paper presents a complete calibration method considering geometric errors, joint compliances and exterior load. An integrated inverse kinematics algorithm combining neural network and analytical method is proposed to calculate controller angle inputs in calibration. For the convenience of calculation, limited parameters calibration is used and then simulated in Matlab. As experimental validation on ABB IRB 6640 shows, the proposed complete calibration model based on limited parameters and neural network can dramatically improve the positioning accuracy of industrial robot.