Design of 6-DoF Holonomic Drive-based Upper and Lower-Limb Stroke Rehabilitation System

Abstract

This paper discusses the design, construction, and characteristics of a holonomic drive-based multi-functional rehabilitation robot that caters to the therapeutic requirements of the upper and lower limb for stroke patients suffering from hemiparesis. The proposed device is aimed to substitute upper and lower limb exoskeletons or other end-effector robotic arm based rehabilitation systems, that are generally bulky, expensive due to customized design, and are applicable only in high-end clinical settings. The proposed device is capable of aiding patients to perform major rehabilitation exercises involving all anatomical movements at each of the joints except radial, ulnar deviation of the wrist and adduction, abduction of the ankle joint. The device utilizes holonomic drive implemented by the three omnidirectional wheels that are integrated to mobile base layout. The robotic device extends an end-effector plate covered with velostat-based pressure sensor matrix on the top. The end-effector plate allows patient to rest their hand or leg for rehabilitative movements. Additionally, the velostat pressure sensor matrix encased on top of end-effector plate enables to predict patients movement. The spatial data from the velostat-based pressure sensor matrix system is fed to a CNN model which runs on the highly efficient Sony Spresense ™ edge device to generate real-time prediction of patient intention of hand or leg movement. On prediction, the rehabilitation robot assists to complete the desired therapeutic movement. The developed CNN model exhibits an accuracy of 98.35% and outperforms other ML models that were realized. The proposed mobile base driven rehabilitation robot facilitates greater range of movements compared to other end-effector-based rehabilitation devices.