Object Grasping and Manipulating According to User-Defined Method Using Key-Points

Abstract

This paper presents an approach for grasping objects by the robotic manipulator at predefined points by userdefined grasping methods for cases of arbitrary position and orientation of objects. The key idea is to implement a detection algorithm for user-defined key-points on the reference object image. As these points can be detected and localized on the object surface, it is possible to generalize the initial sequence of robotic manipulations, pre-recorded for the single scene, to the set of situations with arbitrary object orientation and location within the workspace. The robotic arm had a depth camera, mounted on the end effector for 3D key-points detection and anthropomorphic robotic hand.