Image-based visual servoing for linear path control

Abstract

In this dissertation, we explore the curved trajectory problem in image-based visual servoing and propose a new image-based visual servoing method that can control the path linearly using the defined position and angle errors on image plane of stereo cameras. We analyzed that the cause of curved trajectory is the inadequate error definition on image plane and the control method to decrease the errors at the same rate. The defined errors on image plane cannot exactly correspond to the position and orientation errors in three-dimensional space. In case of general visual servoing, the image errors are normally defined by the difference of coordinates between the image point of features on the end-effector and the image set points. When the target pose is much different from the current pose of the end-effector, an attempt to decrease errors on image plane of stereo cameras causes one or more of the point features on the end-effector to move back and forth along z-axis (optical axis) of camera during visual servoing. This unnecessary movement along the z-axis direction could lead the robot to collide with other objects or to the limit of robot workspace. In this dissertation, we propose a new image-based visual servoing method that improves the linearity of path using the position and angle errors under large pose error. To implement the proposed method, first, we select one point on the end-effector to be moved linearly toward the set point, then we define the distance error between the selected point and set point. This selected point is a pivot point to rotate the end-effector. Next, we define the angle errors between the feature lines including the selected point on the end-effector and the corresponding target lines. When we make such defined errors to be reduced, the selected one point can be linearly moved toward the target point and the end-effector can be rotated around the selected point and finally can be located on the set points of the target position....