Graduation Year
2023
Document Type
Dissertation
Degree
Ph.D.
Degree Name
Doctor of Philosophy (Ph.D.)
Degree Granting Department
Mechanical Engineering
Major Professor
Rajiv Dubey, Ph.D.
Co-Major Professor
Redwan Alqasemi, Ph.D.
Committee Member
Sudeep Sarkar, Ph.D.
Committee Member
Kyle Reed, Ph.D.
Committee Member
Kandethody Ramachandran, Ph.D.
Keywords
Assistive Robots, Computer Vision, Grasp Planning, Grasp Pose Refinement, Grasp Quality Measure
Abstract
The goal of this dissertation is to develop a grasping system for assistive robots that can help people with disabilities and the elderly to perform tasks of daily living. In developing this robot grasping system, we maximize its reliability, accuracy, and autonomy. High reliability and accuracy are required for robots to perform tasks around human users and to safely interact with objects that might be fragile or have contents that could spill. High autonomy is desired as users with disabilities are usually not dexterous enough to directly operate the robot. In this dissertation, a human-in-the-loop (HitL) robot grasping system is developed using computer vision and deep learning-based grasp detection. A grasp pose refinement module was created to finetune the grasp poses before robot execution to increase the grasping system's accuracy. A novel object movement-based grasp quality measure was developed to evaluate the grasp poses and provide indications for the grasp pose refinement. The grasping system includes the human user to supervise and provide corrections to increase reliability. While the default mode of the grasping system is highly automated, the user can directly control the robot in an assisted velocity control mode when the high autonomy mode fails. The scope of this work is object-oriented robot grasp planning, where we mainly use the object geometry features and the gripper-object contact properties to plan the object’s grasp pose. The main contributions of this dissertation are the grasp quality measure and the grasp pose refinement method. The secondary contributions include the development of a vision module, an initial grasp detection module, an assisted velocity control module, and a graphical user interface module. We tested the grasp quality measure with a real robot grasping experiment and a method comparison experiment. The results of the robot grasping experiment proved that the grasp quality scores calculated by the developed grasp quality measure closely match the real robot grasping results. The quality measure comparison experiment revealed that our grasp quality measure outperforms the state-of-the-art grasp quality measures in evaluating parallel-jaw gripper grasp poses when considering object movement. The HitL robot grasping system was tested by comparing its high autonomy mode with a state-of-the-art grasp detection network-based robot grasping system in real robot grasping. The results showed that our robot grasping system with the grasp refinement module works on a wider range of objects and could significantly improve the grasp success rate and reduce object movement compared to the grasp detection network-based robot grasping system.
Scholar Commons Citation
Tan, Tian, "A Human-in-the-Loop Robot Grasping System with Grasp Quality Refinement" (2023). USF Tampa Graduate Theses and Dissertations.
https://digitalcommons.usf.edu/etd/9935
