Title: Task-centric Optimization for Assistive Mobile Manipulators

Ariel Kapusta

Robotics Ph.D. Candidate

School of Mechanical Engineering

Georgia Institute of Technology

Date: Thursday, June 7th, 2017

Time: 11:00am to 1:00pm (EST)

Location: BME Whitaker building - Room 1232

Committee Members:

---------------

Dr. Charles C. Kemp (Advisor), Dept. of Biomedical Engineering, Georgia Institute of Technology & Emory University

Dr. Lena Ting, Dept. of Biomedical Engineering and School of Mechanical Engineering, Georgia Institute of Technology 

Dr. C. Karen Liu, School of Interactive Computing, Georgia Institute of Technology
Dr. Greg Turk, School of Interactive Computing, Georgia Institute of Technology 

Dr Frank L. Hammond, Dept. of Biomedical Engineering and School of Mechanical Engineering, Georgia Institute of Technology 

Abstract:

-----------

Assistive mobile manipulators could enable people with disabilities to perform tasks for themselves which would otherwise be difficult or impossible. The robot's assistance has the potential to increase independence and quality of life. Through this dissertation we have explored methods to realize that assistance.

Although many groups have looked at how a robot could execute some specific tasks, few have considered where to place the robot to better provide assistance. We have observed that this problem arises frequently in real-world settings and solving the problem can be challenging, even for an expert user. We first answer present an answer to the question, "How should a robot choose a configuration of its base to be better able to provide assistance?" In answering this question we expand the problem to better match common scenarios in assistive robotics, where the task may be complicated and may take place in a bed or wheelchair.  First we present task-centric optimization of robot configurations (TOC), a method for addressing this question. We demonstrate how TOC can select one or more robot configurations for many assistive tasks that involve the robot moving a tool around a person’s body. We additionally provide evidence that TOC outperforms baseline methods from literature. We present an assistive robotic system with a robotic bed and a mobile manipulator that used TOC to allow the two robot to autonomously collaborate to better provide assistance. We tested this system with a person with severe quadriplegia in his home, providing evidence of the feasibility of TOC and the robotic system for providing assistance to real people. 

Through our work on assistive robotics, we recognized that an important activity of daily living (ADL), dressing, contains special challenges not fully addressed by TOC. We present task optimization of robot assisted dressing (TOORAD), a method for selecting actions for both the robot and a person that will likely result in successful dressing. We demonstrate the efficacy of TOORAD in a study with participants with disabilities receiving dressing assistance from a mobile manipulator. In that study, we also administered surveys on habits, needs, capabilities, and views on robot-assisted dressing that we expect will provide guidance to future research.