Object Classification with a Robot Gripper equipped with Force Sensitive Fingertips using Convolutional Neural Networks

Abstract

In this paper we present a solution for the identification and classification of grasped objects with an electrical robot gripper with force sensor arrays in its fingertips. The solution is based on a convolutional neural network (CNN). The CNN is trained with relatively few examples but gives already reasonable results. Objects to be detected are of geometrical shape like ring, pen, sphere. The challenges in such applications are the generation of random training data and interfaces between the different components such as gripper, sensor array fingertips and robot. The trained CNN is ported to a Raspberry Pi for real-time execution and communication between the gripper and the robot.