Design and development of universal soft robotic end effector through machine learning on the IRB 360 robot

Abstract

The end effector (gripper) is an important part of a robotic system that is used for industrial and domestic tasks like grasping, carrying, manipulating, assembling, painting, and so on. For handling different types of objects hard as well as soft, require different types of the gripper. The employment of compliant soft-robotic grasping systems, which are characterized by high flexibility in terms of workpiece shape, dimension, and anatomy, is a good method to incorporate greater flexibility into production. The study's major goal is to build and analyses the soft-robotic grippers in terms of repeatability with large payload capacities. End effector (soft gripper) control is crucial for precision work by applying different gripping forces according to the object going to pick. The selection of suitable gripping force for a particular object is done by the process of machine learning (ML). The soft gripper is designed, fabricated, and tested using Industrial Robot (IRB 360) flex picker robot. The virtual environment is created to move the linear path using Robot studio software with rapid programming language. The accuracy, precision, recall, and receiver operating characteristic curve (ROC) curve are analyzed and predict the gripper force accurately with 94% when compared with experimental value. The gripper is working effectively from 1.4 to 2.8 bars with a maximum payload of 500 g. The soft flexible gripper angle is measured based on the pressure using an image processing edge detection technique. The optimized best possible gripping force is predicted using different objects and control action is done to supply exact force to the gripper.