Full 3-D tracking using the controlled active vision paradigm

This paper presents algorithms for 3-D robotic visual tracking of moving targets whose motion is 3-D and consists of translational and rotational components. The objective is to track selected fmura of the moving wget IDd ID position their projections on the image plane at deoired positions. The most important cliaracleriJtia of the proposed algorithm are the use or a single cmncra mounted on the end-effector of a robotic device (cye-in-hand configuration) and the fact that these algorithms do not require accunue knowledge of the relative distance of the target from the camera frame. This fact allows for the potential use of these algorithms in poorly calibrated environments. The entire problem is fonnulated as an application of the controlled active vision framework which states that controlled, rather than accidental. motion of the vision sensor can mu.imize the performance of any active vision algorithm. The camera model introduces a number of parameters lhllt must be estimated on-line. An adaptive control algorithm compensates for the modeling crron, the tracking erron, and the computational delays which are introduced by the time-consuming vision algorithms. Experimental re6ults arc presented to verify the potential of the proposed algorithms. The experiments were performed on the TROJKABOT robotic system which consists or three PUMA560's.