University of Twente Student Theses


Data-Driven Gravity and Stiffness Compensation for a 4-DoF Upper Extremity Robotic Exoskeleton

Kuenen, Thom (2024) Data-Driven Gravity and Stiffness Compensation for a 4-DoF Upper Extremity Robotic Exoskeleton.

[img] PDF
Abstract:Duchenne Muscular Dystrophy (DMD) is a progressive degenerative neuromuscular disorder, causing 1:5000 boys and men to lose functionality in their muscles from a young age. From age 10-13 the upper extremity functionality is affected and patients are unable to function individually. The DAROR-01 is a robotic exoskeleton for the right upper arm, designed to aid these patients during Activities of Daily Living. For this, a gravity and stiffness compensation model is required. This work compares an a priori modelling approach with data driven models. For the data driven models, a white box approach using the kinematic structure and a grey/black box approach using Neural Networks (NN), are designed and validated. The data for the data-driven model is acquired with an identification protocol of 15 minutes in which the system moves through the workspace. The white box model that used the kinematic structure with an extension to also contain joint stiffness proved to be able to learn the gravitational and stiffness compensation model from very few data. This model achieved a Root Mean Square Error (RMSE) of 0.7 Nm, whereas the A Priori model had a RMSE of 1.2 Nm. The grey box model performed better than the black box model, with an RMSE of 2.4 Nm and 2.6 Nm respectively. From this offline analysis can be concluded that a data-driven white box modelling approach thus yields a better compensation model than a an A Priori model.
Item Type:Essay (Master)
Faculty:ET: Engineering Technology
Subject:33 physics, 42 biology, 44 medicine, 50 technical science in general, 52 mechanical engineering
Programme:Biomedical Engineering MSc (66226)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page