Grasp Registration and Learning in Virtual Reality Environments

The objective of this paper is to present a grasping registration technique by using Virtual Reality technologies. In this context, an individual interacts with a virtual world by means of several devices which are integrated into a common architecture. A 6 DOF tracking system gives the position of the user in the scene and a VR glove provides the flexion angles of the fingers of a virtual hand. Through a Head Mounted Display the user achieves a total immersive visualization of the scene. This technique is being developed in order to provide a large dataset to be processed by grasp learning algorithms in robotic environments. In this document, we show the design and implementation stages of this technique as well as the tests executed in both simple and complex scenes. The results obtained let us assure the applicability of our system for learning-based grasp solution in more complex and general scenarios. Key-Words: Virtual Data Registration, Virtual Grasping, 3D Pose and tracking, Manipulation Learning, Virtual Reality.