Abstract: The pose adjustment of manipulator led and controlled by micro-vision is deduced for the problem how to make three points on its end-effector move for three set points in three-dimensional space respectively. A move planning is designed to avoid the collision and intervention of parts and to hold the end-effector in the visual field of the microscope, which stipulates one point moving along a line to its goal position and the other two points moving to their respective goal positions by rotating on the first point. According to this planning, a method based on the distance and angle between lines is proposed to estimate the image feature deviation, and the image Jacobian and kinematic model of manipulator based on the image feature are defined and built. An optimal control algorithm for the system with image delay is derived by the minimization of a cost function that allows weights to be placed on the error of the image features and the control signal. A micro-visual servoing control structure is designed to improve the positional accuracy, which includes a proportional controller and an optimal controller. The switch between the two controllers is selected by an error switch. Using a six-degree micromanipulator, experiments for adjusting line pose and tracking a moving target are performed to validate these researches. The results demonstrate that these researches are feasible and correct, and the location error is less than 2 pixels, which can satisfy the needs of the micro assembly.

Key words: Image Jacobian, Micro-assembly, Micro-visual servoing, Optimal control, modifier, compressive strength, surface tension, animal glue