Autonomous grasping with pre-taught grasps

  • Home
  • -
  • Projects
  • -
  • Autonomous grasping with pre-taught grasps
Saisie autonome d’objets à l’aide de prises pré-enseignées

Autonomous grasping with pre-taught grasps

DESCRIPTION

The objective of this research project is to develop a method for grasping objects in a heap with a serial six-degree-of-freedom robot, the wrist camera and one of the gripper hands produced by Robotiq.

In this work, a heap means a random distribution of objects which can be close to each other or overlap one another. We also assume that each heap is constituted of multiple instances of the same object. The object geometry is known beforehand, generally through a 3D scan of the object made with the robot wrist camera. Robust grasps are also known in advance, being defined by the robot operator through direct teaching. The objects used in this work originate from the industrial environment and may have reflective surfaces. The research is divided in two segments: first, a classical model-based method will be developed, and second, an alternative method based on machine learning will be developped and compared to the other.
 


Team

Jean-Christophe Ruel
Jean-Christophe Ruel

Université Laval

Details