Real-Time Slip Detection and Control Using Machine Learning

The handling and gripping of objects by a prosthesis depend on the precise applied force control and the slip detection of the grasped object. These two features combined allow for the adjustment of the minimum grip force required to prevent slipping. Based on this statement, a system was developed to control the slip of objects, composed of a grip controller, for which the objective was to hold the object, and through the signal of a tactile sensor, slip is detected. An artificial neural network was used to identify the slip event for different types of objects. If the response from the classifier is positive, indicating slip, the system sends a signal to the grip controller, so that it increases the grip force performed on the object, aiming at minimizing slippage. In the end, the performance of the system for different objects was analyzed; the result encountered was that the system efficiency is proportional to the mass and the rigidity of the grasped object.