Automated tool selection and optimal grasping point generation for robotics

Abstract

This thesis addresses the challenge of automating optimal grasping point generation for robotic manipulation in unstructured environments. Traditional industrial robots rely on manually defined grasping points, limiting their flexibility and efficiency when handling diverse components with varying geometries and positions. To overcome this, an algorithm is developed to analyze CAD files of objects and generate optimal grasping points tailored to different robotic end-effectors, including suction cups, sponges, and grippers. The method involves creating an initial dataset of potential grasping points by analyzing the geometry extracted from CAD models. These grasping points are then filtered based on tool-specific constraints such as surface flatness, accepted torque, and accessibility to ensure successful and stable grasping. The algorithm’s effectiveness is validated through visualizations (and to a lesser degree simple simulations) of generated grasping points on various objects of differing shapes and sizes. The developed algorithm reduces the need for human intervention, improving the automated production line, and lays the foundation for increased automation in the future.