Grasping and Sorting Cutlery in an Unconstrained Environment with a 6 DoF Robotic Arm and an RGB+D Camera

Abstract

Object manipulation is the ability to perform tasks of grabbing, pushing or moving objects. Although this is something humans do naturally and without realizing it, for robots it is still hard to do. Despite that, robotic manipulation has seen an exponential evolution over the years, although there are still many open challenges in the field. One of these challenges is cutlery manipulation since this kind of object has features that make them hard to manipulate, like the long and thin shape, the metallic specular reflections and the fact that cutlery is usually mixed in a container. In this paper, we propose a method that sorts cutlery, which includes three main steps: the detection of the particular object in the point cloud, the subsequent grasping that can be helped by pushing motions powered by an algorithm that was previously developed, and finally the placing of the object in a particular container, according to its type. We show that we are able to perform this sorting task in simulation using a UR3 robotic arm and a RGB+D camera. We explore different settings and improve a previous pushing and grasping method.