University of Twente Student Theses
Model-Free Reinforcement Learning Control of a Pneumatic-Driven Soft Continuum Robot
Karytsas, Konstantinos (2023) Model-Free Reinforcement Learning Control of a Pneumatic-Driven Soft Continuum Robot.
|
PDF
5MB |
| Abstract: | Soft continuum robots constitute a category of inherently compliant robots which can interact safely with their surroundings. Therefore, they are considered a promising technology in the biomedical field with possible applications in laparoscopy and endoscopy, among others. Unlike rigid-link robots which have defnite kinematic mapping, the accurate analytical modeling of soft continuum robots is diffcult, especially when considering their non-linear deformation when actuated, the non-linear elasticity of their materials, and their susceptibility to interactions with the environment. Hence, soft continuum robots face modeling errors which lower the performance of model-based control. The scope of the present thesis is the development of a model-free reinforcement learning control scheme in order to control a pneumatic driven soft continuum robot in 3D space. A physical simulation environment which is based on Cosserat rod theory and discrete differential geometry is used for policy training data generation while Proximal Policy Optimization is utilized for policy optimization (i.e. controller optimization). |
| Item Type: | Essay (Master) |
| Faculty: | ET: Engineering Technology |
| Subject: | 31 mathematics, 33 physics, 52 mechanical engineering, 54 computer science |
| Programme: | Systems and Control MSc (60359) |
| Link to this item: | https://purl.utwente.nl/essays/97673 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page