Intelligent Robotic Systems for Learning and Reproducing Therapeutic Interventions in Rehabilitation Medicine

  • Author / Creator
    Fong, Jason
  • As the world's population increases and ages, the demand for rehabilitation medicine services is on the rise. Recently, robot-assisted rehabilitation has become an appealing, powerful and economical means with which to address this demand while lowering the burden on practicing therapists and the healthcare system. However, current robotic rehabilitation systems take therapists out of the loop and replace them with an algorithm that determines what interactions to provide to a patient participating in therapy. The therapist can intervene but only through a computer interface and not on a hands-on basis. This is problematic because years of therapist's education and experience are not being used. Also, substantial changes to the way the robot interacts with the patient requires computer programming know-how that does not usually exist in clinics. This thesis focuses on the incorporation of Learning from Demonstration (LfD) algorithms into rehabilitation robotics for the purposes of efficiently time-sharing therapists across multiple patients, and to enable therapists with minimal computer programming knowledge to easily and intuitively reprogram the behaviors of rehabilitation robots on a patient-specific, task-specific, and session-specific basis. A secondary aim of introducing a generalized, manipulator-based robotic system to various areas of rehabilitation medicine is also explored. In current clinical practice, a multitude of equipment is often required to facilitate both rehabilitation and assessment, which is inefficient in both space and cost. Current robotic rehabilitation systems also come in many different designs that are specific to at most a few select therapy tasks, thereby failing to address either of these inefficiencies. The work in this thesis shows that one single system can be applied to both upper and lower limb post-stroke therapy, as well as occupational rehabilitation of injured workers.

    This thesis presents the use of LfD algorithms for learning position-based and impedance-based behaviors of a therapist when guiding a patient through a post-stroke robot-assisted therapy task. The proposed system is evaluated for upper limb tasks meant to resemble Activities of Daily Living (ADLs), and for lower limb therapy in the form of treadmill-based gait training. In each case, the therapeutic interventions associated with each exercise are learnt from demonstrations provided by a therapist and are then reproduced by the rehabilitation robotic system. The fidelity of these semi-autonomous reproductions is then evaluated. In addition, a comparison between the use of a telerobot (i.e., a pair of teleoperated robots) versus a single robot as the therapeutic medium is performed.

    Lastly, the intelligent robotics technologies developed throughout the thesis is applied to the field of occupational rehabilitation, which has seen relatively little robotic integration. The purpose of this application is to introduce the benefits of robotics specifically for Functional Capacity Evaluation (FCE) of workers injured in the course of their duties. The system is integrated with a virtual environment that simulates a workplace task, which is presented to users through both haptic feedback and an augmented-reality display for greater immersion. The realism of the system is evaluated by comparing user biomechanics when interacting with the robotic system, and when performing a real-world equivalent version of the therapy task.

    Evaluations for all of the included works are performed with able-bodied participants. The term "therapist" refers to participants with no formal training in rehabilitation medicine who are asked to train the robots. The term "patient" refers either to able-bodied participants (who may have a disability simulated through worn equipment), or to mechanical devices such as springs that are meant to simulate a patient with disability.

    Robotic rehabilitation systems such as the one presented in this thesis represent a large step forward in the rehabilitation medicine field. By enabling robots to learn from human experts and focusing on making designs generalizable, the potential to provide cost-efficient, intensive, and immersive therapy with minimal burden on healthcare providers while keeping specialists in the loop becomes closer to being realized.

  • Subjects / Keywords
  • Graduation date
    Fall 2019
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    Permission is hereby granted to the University of Alberta Libraries to reproduce single copies of this thesis and to lend or sell such copies for private, scholarly or scientific research purposes only. Where the thesis is converted to, or otherwise made available in digital form, the University of Alberta will advise potential users of the thesis of these terms. The author reserves all other publication and other rights in association with the copyright in the thesis and, except as herein before provided, neither the thesis nor any substantial portion thereof may be printed or otherwise reproduced in any material form whatsoever without the author's prior written permission.