Camera pan/tilt to eliminate the workspace-size/pixel-resolution tradeoff with camera–space manipulation

Abstract

The successful implementation of close-tolerance, three-dimensional rigid body assembly has been robustly achieved using camera-space manipulation in a limited region of the manipulator's workspace. The extension of this capability to a broader region can in general be achieved by mounting the cameras on computer-controlled platforms or “pan/tilt” units. The use of this type of platform enables the encompassing of a large physical region within the fields of view of the cameras, while preserving an approximately constant image-plane resolution per unit physical space. The paper describes the derivations involved in the determination of view parameters when the information of the angles of pan/tilt rotation of the cameras is available. Such procedure enables adequate parameter observability with a greatly reduced sampling in terms of number and breadth. Practical considerations for the implementation of this capability for a high-precision, three-dimensional task across a large workspace region are also presented.