Visual control system for grip of glasses oriented to assistance robotics

Abstract

Assistance robotics is presented as a means of improving the quality of life of people with disabilities, an application case is presented in assisted feeding. This paper presents the development of a system based on artificial intelligence techniques, for the grip of a glass, so that it does not slip during its manipulation by means of a robotic arm, as the liquid level varies. A faster R-CNN is used for the detection of the glass and the arm's gripper, and from the data obtained by the network, the mass of the beverage is estimated, and a delta of distance between the gripper and the liquid. These estimated values are used as inputs for a fuzzy system which has as output the torque that the motor that drives the gripper must exert. It was possible to obtain a 97.3% accuracy in the detection of the elements of interest in the environment with the faster R-CNN, and a 76% performance in the grips of the glass through the fuzzy algorithm.