Single Shot Learning of the Stiffness Distribution of a Soft Object

Master thesis (2021)

Authors

L.A. Besselaar Mechanical Engineering

Contributors

C. Della Santina Learning & Autonomous Control - Mechanical, Maritime and Materials Engineering (mentor)
J. Kober Learning & Autonomous Control - Mechanical, Maritime and Materials Engineering (graduation committee member)
M. Wiertlewski Human-Robot Interaction - Mechanical, Maritime and Materials Engineering (graduation committee member)
Faculty
Mechanical Engineering, Mechanical Engineering
Copyright: © 2021 Lars Besselaar
More Info
expand_more

Published Date

03-09-2021

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Mechanical Engineering

Abstract

With robotics rapidly expanding towards new user-oriented applications, such as agriculture, households, and classrooms, new tasks and new requirements arise. One of the ongoing and unsolved problems that comes with the unpredictable new environments robots find themselves in, is soft object manipulation. Without assuming the handled object to be rigid, its behavior under forces is no longer known: it could break, bend, plastically deform, or be crushed. In
this paper, we propose a method for the identification of the deformation model of a soft object. We consider a bimanual robotic system manipulating a deformable object with the intention of controlling the object’s shape. Since the object’s deformation parameters are not known in advance, we propose
a learning algorithm for their identification, and we specifically focus on learning the stiffness distribution along the object. By using a state space model based on modes of curvature, rather than a discrete Cartesian state space, we are able to learn the stiffness distribution of the deformable beam with only a single
experiment. Then, we prove that the proposed method provides full control over the shape of the beam by performing a control task in simulation. Given the promising accuracy of the method, this work provides a solid foundation for future work in the direction of the fast identification of deformation parameters.