An occlusion prediction algorithm for visual servoing tasks in a multi-arm robotic cell

This paper presents a new approach to the prediction of visual occlusions for a robotic system performing visual servoing tasks. The case of a multi-arm robotic cell equipped with a multi-camera eye-to-hand visual system is considered. The proposed approach is able to predict the self occlusions as well as the mutual occlusions caused by the robot links and tools. The algorithm is based on the geometric representation of the cell in terms of a BSP tree, assuming that the CAD geometry of the objects and the kinematic model of the multi-arm robotic system are known. Experimental results are presented for the case of two industrial robots performing a vision-guided grasping task.