I. Introduction
In image-based visual servoing (IBVS), the choice of the set of visual features to be used in the control scheme is still an open question. Image features that can be, and have been, used include the coordinates of interest points, the parameters of lines or conics, the moments of planar patches, etc. We wish to choose features that lead to control behavior that is optimal with respect to the image, e.g., keeping points within the field of view, as well as with respect to 3-D camera motion, e.g., minimal distance moved and avoidance of robot singularities.