Planning of contact points and grasping based on contour attributes using image processin

(Robot Vision)
Future robotic systems will operate in complex and constantly changing environments, such as private homes and offices. This requires knowledge about the robot’s surroundings, including persons, objects and obstacles, which is typically obtained by cameras and computer vision systems. Subsequently, autonomous operations, such as grasping or moving, are planned based on this knowledge.

In this thesis, you’ll focus on the planning of contact points and grasping of small elongated objects. Imagine a bunch of different screws spread around a table. The task for the robot is to sort the screws according to their size (and type). Once a screw has been detected, the robot needs to determine suitable contact points and plan how to grasp the screw with a two-finger gripper. The fact that the elongated objects rest on a plane (the table) simplifies the problem at hand and allows to estimate contact points from the object’s geometry.

This work is conducted within the RoVi project, which develops low-cost robotic systems with vision-based sensors. RoVi aims to transfer research results into products and a startup company.
Keywords: robotic, grasp planning, computer vision