Nearly 800,000 people in the USA have strokes each year, with 50% of survivors suffering loss in neuromotor control which leads to a severely diminished quality of life. Movement therapy is an essential component of post-stroke rehabilitation, and is traditionally based on the physical interaction between a therapist and the patient. Soft wearable robots have the potential of supporting post-stroke rehabilitation, due to their lightweight nature and individual patient conformability. Our lab is focused on developing soft wearable robots based on cable-conduit transmissions. In these systems, the heavy actuators are relocated away from the joints, thus reducing dynamic loading. While a flexible and lightweight solution, cable-conduit systems introduce large frictional forces into the system that may adversely affect user motion.

For my thesis I worked on two different approaches to achieving low-impedance cable-conduit actuators for upper extremity rehabilitation. The first solution was the development of a novel dynamic non-linear model for cable-conduit transmission capable of interaction with non-passive environments (such as a patient during rehabilitation), featuring bi-directional motion propagation. This work expanded previous work in our lab on this topic by creating a state estimator for use in real-time control with an emphasis on code optimization for implementation ease. The developed model incorporated a novel linear solver, which outperformed traditional nonlinear solvers by over two orders of magnitude in terms of computation speed. The estimator underwent validation testing to determine the possibility of using the developed model to estimate cable tension based on measurements distributed along the cable (sensor mean estimation error of 0.3N with presence of 10% Gaussian white noise). Tests were also done on the estimator to validate optimal estimation performance based on the number of sensors available along the cable, their noise, and their location (mean estimation profile error of 0.38N when averaging only 2-5 sensors and 0.33N when averaging 6-11 sensors located at the cable’s distal end while under known direction of mass motion).

The second approach to low-impedance cable-conduit actuation was based on a modified mechanical design of the actuator. This option was explored due to the requirement of pretension which would apply a constant load to the user’s joint, which is undesirable. With a human-in-the-loop, the need for the system to start from slack is paramount as it allows zero resistance (“complete transparency”) between the human and the actuator when not in use. A slack enabling feeder system designed to eliminate slack around the motor and keep the cable from unwrapping was built in order to test if this component would reduce the frictional effects along the cable. Preliminary results using our first-generation prototype were promising. The system was able to keep the motor spool from unwrapping at low to midrange speeds, however, only visual confirmation was possible. High speeds produced unanticipated effects due to design flaws, tolerancing, and unanticipated material interactions. The completion of a second-generation prototype designed to remove observed design flaws is suggested, as well as further testing to validate the lower friction clams.

As part of a collaboration with the UD’s Department of Fashion & Apparel Studies, I contributed to the design of a wearable exosuit prototype, with the intention of creating a concealable, wearable rehabilitation device for upper extremity rehabilitation, integrating the developed cable-conduit system developed as part of this thesis. Through this process I researched optimal cable load path design and limitations to force transmission due to fabric deformation at the elbow. I designed costume components such as cable anchors and a 3D printed cable-conduit anchor designed as a low-profile connection for ergonomics and concealability.