Error modeling, parameter identification and non-parametric calibration of a 3PUS/S spherical parallel manipulator

Abstract

In comparison to serial manipulators, parallel robots have many applications in industries due to their unique characteristics such as larger stiffness-to-mass and payload-to-weight ratios. Since, the accuracy and error reduction of the manipulators are very prominent, various methods have been proposed to calibrate parallel robots with different structures so far. This article presents the kinematic parameters estimation and non-parametric calibration of a 3PUS/S spherical parallel manipulator. In this regard, the kinematic equations of the manipulator are derived to identify the error sources, then an error model is developed to correlate the errors of the robot's outputs to the inaccuracies of the manipulator's structural parameters and joint variables. In addition, a prototype of the manipulator is designed and made to collect suitable data sets for calibration, then many experiments in the robot's workspace are performed, and the input and output values are recorded by encoder and AHRS sensors with high accuracy. Next, the estimation of geometric parameters is fulfilled by an optimization method, then the artificial neural network is employed for non-parametric calibration. The results show that the geometric calibration based on parameter estimation and non-parametric calibration could improve the robot accuracy about 73 % and 96 %, respectively.