Full 3-D tracking using the controlled active vision paradigm

The authors present algorithms for 3D robotic visual tracking of moving targets whose motion is 3D and consists of translational and rotational components. The objective is to track selected features of the moving target and to position their projections on the image plane at desired positions. Only a single camera mounted on the end-effector of a robotic device is needed and the algorithms do not require accurate knowledge of the relative distance of the target from the camera frame. This fact makes these algorithms potentially useful in poorly calibrated environments. The problem is formulated as an application of the controlled active vision framework. The camera model introduces a number of parameters that must be estimated online. An adaptive control algorithm compensates for the modeling errors, the tracking errors, and the computational delays introduced by the vision algorithms. Experimental results for the TROIKABOT robotic system, which consists of three PUMA 560s, verify the potential of the algorithms