Application of a Precision-Enhancing Measure in 3D Rigid-Body Positioning Using Camera-Space Manipulation

This article presents an iterative procedure for improving the precision obtainable with camera-space manipulation. It also reports on extensive experimental results using this procedure with a task of some historical note in the area of vision-guided robotics, a task that involves close-tolerance, 3D, rigid-body assembly. The procedure was devised to achieve higher pre cision by introducing to parameter estimates the accuracy of the "perspective" camera model, while retaining the numerical advantage of the orthographic model. The camera-space coor dinates of each of several visually detected cues on the grasped object are multiplied by an iteratively improved correction fac tor, so that the modified image-plane appearances of the cues more closely approach the ideal of an orthographic projection. The correction factor is normalized near the point of maneuver termination so that camera-space maneuver objectives may stay unchanged. A similar strategy may be used to determine the camera-space objectives themselves. Results of experimental and simulation studies are presented and discussed.