Date of Award

2023

Degree Type

Thesis

Degree Name

Master of Science in Electrical Engineering (MSEE)

Department

Electrical, Computer, and Biomedical Engineering

First Advisor

Reza Abiri

Abstract

Humans have a remarkable ability to grasp and manipulate various objects in their everyday life. However, individuals who suffer from paralysis that result in limited or no voluntary limb muscle control, require assistance from caregivers for activities of daily living, such as eating and drinking tasks. Hereby, a recent study predicts a global shortage of millions of healthcare workers in 2030. The study concludes that this shortage may not occur when labor productivity increases through the better utilization of technology. Assistive robots have the potential to significantly reduce this shortage and assist healthcare professionals in doing so. However, most of the current robotic grasping research focuses on top-down grasps with two-finger or parallel-jaw grippers for robotic arms in industry rather than for the healthcare sector. These traditional grasps are not always helpful in the context of providing assistance to these individuals. In activities of daily living, where the arm needs to grasp objects upright and without obstructing their openings, as in the case of drinking from a mug, these approaches are not useful. This study aims to perform reaching and grasping tasks in a human-inspired manner using only image data. Utilizing closed-loop deep reinforcement learning algorithms and a human-inspired formulation regarding camera position, alignment of the robotic arm and type of gripper used, the goal of this work is to perform a reach and grasp task with the most natural execution possible. The results of this study shows, that the proposed approach is able to reach and grasp one object following the human-inspired formulation with a success rate of 90%.

Download

Available for download on Friday, September 05, 2025

Share

COinS