Robotic Rehabilitation After Stroke - A Modular System for Training Distal Upper Limb Functions

Patrick Weiss


Stroke is the leading cause of acquired disabilities, while new ways of treatment are needed to provide effective rehabilitation at acceptable costs. This thesis deals with the development of a novel modular system for inexpensive, portable, and patient- specific robotic rehabilitation, augmenting traditional therapy with a focus on home rehabilitation. The developed system consists of two main modules allowing for train- ing several distal upper limb functions that are commonly impaired after stroke.
The first module called m·ReSR focuses on training supination/pronation, dorsiflex- ion, and finger-focused rotational movements. The first design iteration was applied in a preliminary study to determine important parameters for a rehabilitation paradigm, improved backdrivability with a developed friction and inertia compensation, and pro- vided feedback from patients and therapists. The gained experience was incorporated in the further development. Custom-made electronics in the second iteration improve the torque measurement and control, making it feasible to omit expensive force/torque sensors. The high flexibility and modularity of m·ReSR’s hard- and software allowed it to be applied for the enhancement of a robot for reaching training, developed at the Toronto Rehabilitation Institute, with an additional degree of freedom.
The second module m·ReSX is an exoskeleton for grasping and pinching exercises, incorporating an innovative parametrization method for size customization. The intro- duction of parameters in the CAD model and the use of 3D printing allows to precisely fit individualized exoskeletons to different hand sizes, thereby reducing misalignment with the patients’ unique anatomical features and avoiding tedious mechanical ad- justments. The fitting quality could already be estimated before 3D printing with a novel pre-print evaluation. Apart from analyzing the measurement data, projected 2D representations of the models overlaid onto images of the hands already indicated the good fitting. This was confirmed with the manufactured exoskeletons fitted to both a stroke patient and a healthy subject. Furthermore, different sensor types were implemented and evaluated. With a maximum root-mean square error of 1.89°, the evaluation showed that the sensors are very accurate and even perform better than a specialized goniometer for joint angle measurements of the finger. Tendon-based transmission is used to transmit forces from remotely placed motors. The force on the cables can be accurately controlled to adapt the assistance or resistance to the pa- tient’s abilities. A model was developed to exploit the motor encoder measurements for an additional joint angle estimation. A special gearbox to distribute the torque of one motor to four fingers, reducing the number of expensive actuators while providing independent movement training, was developed and filed for patenting.
Original languageEnglish
QualificationDoctorate / Phd
Awarding Institution
  • Maehle, Erik, Supervisor
  • Schlaefer, Alexander, Supervisor
Publication statusPublished - 30.04.2015


Dive into the research topics of 'Robotic Rehabilitation After Stroke - A Modular System for Training Distal Upper Limb Functions'. Together they form a unique fingerprint.

Cite this