Grasping novel objects

Abstract

The work explained in this technical report is about evaluating some recent algorithms to grasp unforeseen objects for table clearance tasks. A tabletop object segmentation algorithm is proposed, and two recently published methods to generate grasps are discussed. The report presents practical considerations on how to use the segmentation algorithm and how to perform the tests with the evaluated algorithms for the generation of grasping poses. Finally, the results of both approaches and their comparison are discussed.