Adaptive Neural Network Method For An Exoskeleton Device Control

Abstract

Patients with brachial plexus injuries or those involving the spinal cord often experience a loss of hand function. They need a tool that can assist them in resuming normal life by regaining some use of their hands. Exoskeleton devices are becoming more popular as a treatment for this condition since they can actuate the fingers of patients, restoring their ability to grip items and carry out other more mundane tasks. In this dissertation, we propose the model of a revolutionary exoskeleton device controlled by an adaptive neural network-based controller. The network of neurons was motivated by the ease with which human hands grip a broad range of items. The gripping forces exerted by a human fingertip on an item in many distinct positions were measured. The neural network is used to estimate the unknown items, and adaptive control is utilized to realize the adaptive features in the unknown environment, in order to realize the stability and high precision control of the control system while facing human interferences. Adaptive control is used to carry out both of these tasks. The user initiates a grip, at which point the neural network uses information about the object's orientation, mass, and dimensions to calculate an estimate of the force needed in each of the five digits to hold it.