Using locally adjustable hand-eye calibrations to reduce robot localization error

Abstract

Accurate pose of a robot end effector is required in many applications. Typically, this is achieved by robot calibration and then, registration of the robot frame to the world frame. In this paper, the registration of a poorly calibrated robot to a world frame was performed using many local hand-eye registrations and using one global registration. Both approaches were evaluated using a set of target poses. The use of the properly chosen local transformation applied to each target led to a tenfold reduction in the orientation and position error in comparison with a use of one global transformation. The median orientation error was reduced to 0.029° and the median position error was reduced to 0.221 mm which is approximately four times larger than robot specified position repeatability.