Study on ANN based Upper Limb Exoskeleton

Abstract

This paper analyzes the case of designing an exoskeleton for people with right arm mobility impairments. The type of exoskeleton considered consists of three joints, one of translation along the spine and two of rotation for the arm of the exoskeleton. The first approach presented is for the people who have the intention to move at the level of the muscles, which can be detected with the help of non-invasive EMG sensors, but cannot sustain the movement for a long time. In this approach within the control system there are two components. The first is computational only, which includes the Denavit-Hartenberg models: direct for the human arm and inverted for the exoskeleton. The second component is an ANN-based estimator that estimates the arm’s movement angle based on the acquired EMG signals. The second approach presented in the paper is for people who cannot voluntarily perform the arm movement. For this case, an exoskeleton control system was designed based on the training of an ANN to associate the EMG signals collected from muscle groups controllable by the subject with exoskeleton movements.