Vision-based kinematic calibration of a 4-DOF pick-and-place robot

Pick-and-place robot is usually a 4 degrees-of-freedom (DOFs) apparatus and indispensable in flip chip bonder. Placement accuracy is the most important index to evaluate it. To improve placement accuracy, machine vision is generally utilized to provide the exact pose parameters of chip and pad, which two must be aligned with each other before placement. This paper proposes an effective vision-based kinematic calibration method which calibrates 3 translational DOFs and 1 rotational DOF of pick-and-place robot, respectively. For the 3 translational DOFs, a down-look vision system takes an image and obtains position of a fixed reference cross mark when it moves with pick-and-place robot and the relationship between displacement of pick-and-place robot and position parameters of the cross mark in down-look vision system is the key factor to calibrate it. For the rotational DOF, error parameters of it can be derived after the projective curve of its trajectory in the image plane of up-look vision system is fitted with an ellipse. Placement accuracy is improved to better than ±30μm in the experiment after kinematic calibration. Result shows this vision-based kinematic calibration method is valid and very suitable for 4-DOF pick-and-place robot in flip chip bonder.