Assist-As-Needed exoskeleton system to recovery the hand motion after stroke.

Abstract

Robotic-assisted systems that operate collaboratively with their human operators to provide assistance are becoming a reality. They have been gained significant interest to include in the motion recovery process because they could provide speed up rehabilitation progress by increasing accurate motion control, and monitoring of several patient data along with the therapies continually. Many different paradigms for driving assistance have been developed. One of them is known as "Assistance-As-Needed" from the active-assistance paradigm, which aims to provide physical assistance specific to the subject's individual requirements; the subject needs should be determined and included in robotic-assisted systems. Two challenges due to the former depend on the task being performed and the subject's capability, while the lastest should administrate the assistance according to the capability measurement.

This thesis presents the development of a new close-loop scheme to modulate the assistance given by an exoskeleton robot with a novel technique to the measurement of the subject capability. The close-loop scheme had been developed under the context of hand rehabilitation after stroke due to the hand functions are one of the most complicated to recover, and the stroke leads several motion impairments.

The current doctoral dissertation consists of four stages of development: i) study of the hand biomechanics and robotic system used in motion rehabilitation to define the mechanical structure and system criteria. ii) Robotic system modeling and datasets constructions. The former is the mathematical model used as the plant in the simulation scenarios, and the last is the myoelectrical motions signals used to compute the classification models and test the whole system. iii) Myoelectrical signals analysis to detect subject intention from stroke patients and muscular condition levels. Several pattern recognition techniques were implemented to analyze the feature influence and found the classification model with better performance. Both models are used in the closed-loop scheme. iv) The assistance closed-loop scheme consists of learning, assistance, and active-tracking modules. It is aimed to include the capability measurement to modulate the exoskeleton assistance and drive the exoskeleton position based on the trajectory reference and subject capability. v) Experiments are aimed at quantifying the classifier's performance using EMG signals from stroke patients, assistance modulation, and demonstrating the hypothesis of providing physical assistance specific to the individual requirements through proper modulation of the robot velocity.

Date of Award	19 Sep 2019
Original language	English
Awarding Institution	Pontificia Universidad Javeriana
Supervisor	Julian David Colorado Montaño (Director) & Iván Fernando Mondragón Bernal (Director)