Learning to detect slip for stable grasping

Abstract

As an important basis of stable grasping, slip detection plays a critical role on improving the operation level of robots. In this paper, a novel slip detection method that combines unsupervised learning and supervised learning is proposed. The window matching pursuit is used to extract features and then the SVM is applied to classify the slip and stable events. Superior to other methods, the proposed method has no restriction of grasped object and can be easily applied to other robot hands. In addition, a novel slip-tagging method based on infrared sensor that measures relative distance of object and robot hand is proposed. The platform consisting of Universal Robot, Barrett hand and the infrared sensor is setup. And experiments are implemented to prove effectiveness of the proposed methods.