A new partitioned approach to image-based visual servo control

In image-based visual servo control, where control is effected with respect to the image, there is no direct control over the Cartesian velocities of the robot end effector. As a result, the robot executes trajectories that are desirable in the image, but which can be indirect and seemingly contorted in Cartesian space. We introduce a partitioned approach to visual servo control that overcomes this problem. In particular, we decouple the x-axis rotational and translational components of the control from the remaining degrees of freedom. Then, to guarantee that all features remain in the image throughout the entire trajectory, we incorporate a potential function that repels feature points from the boundary of the image plane. We illustrate our control scheme with a variety of results.