Automatic 2D grasp planning for 2-finger grippers

Abstract

In this thesis, a method is proposed which automatically determines grasp vectors of an object for the robot with a parallel-jaw gripper. The grasp vector is the combination of grasp position and orientation. This method is quite general and versatile: the geometry of the object is arbitrary, no a prior information is needed except the parameters of the gripper, and so the method can be applied to all the automatic grasping problems when the object and its orientation are unknown. This method extracts information about an object by using robot vision techniques. This method segments and parameterizes the object boundary pertinently, and uses a robust and efficient filtering process. Just it yields an explicit solution and very fast and simple algorithm. Practically, our method is executable in real time. Also, by considering various grasp types and using good stability criterion, reliable results are produced.