This paper presents a novel vision-based method for accurate Tool Center Point (TCP) calibration that eliminates the need for manual contact-based procedures and expensive measurement systems. This method relies on a 3D vision sensor to perform calibration during the robot movements. By observing multiple tool poses, a system of linear equations from coordinate transformations is solved to estimate the TCP position, while an orthogonalization step constructs the TCP orientation. Real-world experiments with an industrial robot and a stereo vision system validate the approach. The calibrated TCP is compared to reference measurements from a coordinate measuring machine, achieving a position error of 1. 1 3 4 mm and angular deviation of 0.406°.
Haas et al. (Sun,) studied this question.