Title: Data-driven Mechanical Design and Control Method of Dexterous Upper-Limb Prosthesis

Date: July 27th, 2022 (Wednesday)

Time: 2 pm - 4 pm EDT

Location: 3115 (McIntire Conference Room), BME Whitaker Building

Joshua Lee

Robotics PhD Student

Woodruff School of Mechanical Engineering

Georgia Institute of Technology

Committee

Dr. Frank Hammond III (Advisor) - Woodruff School of Mechanical Engineering, Georgia Institute of Technology

Dr. Jaydev Desai - Coulter Department of Biomedical Engineering, Georgia Institute of Technology

Dr. Jun Ueda - Woodruff School of Mechanical Engineering, Georgia Institute of Technology

Dr. Lewis Wheaton - School of Biological Sciences, Georgia Institute of Technology

Dr. Marco Santello - School of Biological and Health Systems Engineering, Arizona State University
 

Summary

With an increasing number of people suffering from impaired upper limb function due to various medical conditions like stroke and blunt trauma, the demand for highly functional upper limb prostheses is increasing; however, the rates of rejection of prostheses are still high due to factors such as lack of functionality, high cost, weight, and lack of sensory feedback. Modern robotics has led to the development of more affordable and dexterous upper limb prostheses with mostly anthropomorphic designs. However, most are still economically prohibitive and suffer from control complexity due to increased cognitive load on the user.

Thus, this thesis work aims to explore the designs and control methods of prostheses that relies on the emulation of the kinematics and contact forces involved in grasping tasks with healthy human hands rather than on biomimicry. This is accomplished by 1) experimentally characterizing human grasp kinematics and kinetics as a basis for data-driven prosthesis design. Using the grasp data, steps are taken to 2) develop a data-driven design and control method of an upper limb prosthesis that shares the kinematics and kinetics required for healthy human grasps without taking the anthropomorphic design.

This thesis demonstrates an approach to decrease the gap between the functionality of the human hand and robotic upper limb prostheses by introducing a method to optimize the design and control method of upper limb prosthesis by collecting grasp data from human subjects with a motion and force capture glove and minimizing control complexity by reducing the dimensionality of the device while fulfilling the kinematic and kinetic requirements of daily grasping tasks. Using these techniques, a task-oriented upper limb prosthetic is synthesized and tested in simulation and physical environment.