Learning Deformable Linear Object Dynamics From a Single Trajectory

The dynamic manipulation of deformable objects poses a significant challenge in robotics. While model-based approaches for controlling such objects hold significant potential, their effectiveness hinges on the availability of an accurate and computationally efficient dynamics model. This work focuse...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:IEEE robotics and automation letters Jg. 10; H. 7; S. 7635 - 7642
Hauptverfasser: Mamedov, Shamil, Geist, A. Rene, Viljoen, Ruan, Trimpe, Sebastian, Swevers, Jan
Format: Journal Article
Sprache:Englisch
Veröffentlicht: Piscataway IEEE 01.07.2025
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Schlagworte:
and learning for soft robots
Computational modeling
control
Deformable models
Deformation
Discretization
Dynamics
flexible robotics
Formability
Heuristic algorithms
Kinematics
Learning
Mathematical models
Model learning for control
modeling
Neural networks
Numerical models
Physical properties
Predictive models
Rigid structures
Robotics
Robots
ISSN:2377-3766, 2377-3766
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:The dynamic manipulation of deformable objects poses a significant challenge in robotics. While model-based approaches for controlling such objects hold significant potential, their effectiveness hinges on the availability of an accurate and computationally efficient dynamics model. This work focuses on sample-efficient learning of models to capture the dynamic behavior of deformable linear objects (DLOs). Inspired by the pseudo-rigid body method, we present a physics-informed neural ODE that approximates a DLO as a serial chain of rigid bodies interconnected by passive elastic joints. However, unlike traditional uniform spatial discretization and linear spring-damper joints, our approach involves learning-based discretization and nonlinear elastic joints that characterize interaction forces via a neural network. Through real-world and simulation experiments involving DLOs with markedly different physical properties, we demonstrate the model's ability to accurately predict DLO motion.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:2377-3766
2377-3766
DOI:10.1109/LRA.2025.3577421