Kinematic Calibration of a Spherical Parallel Robot

Abstract

Kinematic calibration is essential for improving the accuracy of parallel robots by compensating for geometric errors. This paper presents a calibration method for spherical parallel robots using relative pose measurements. By comparing relative end-effector poses from the robot’s forward kinematics to those from a ground truth measurement system, the method identifies kinematic parameters without relying on the robot base frame. An error model based on relative poses in the SE(3) manifold is formulated. Parameters are optimized using the Levenberg-Marquardt algorithm to minimize the error between measured and estimated relative poses. The methodology is demonstrated by calibrating the ARAS-DIAMOND spherical parallel robot. Results show improved repeatability and trajectory tracking after calibration, confirming its efficacy.