Tesi etd-03282025-144252
Link copiato negli appunti
Tipo di tesi
Dottorato
Autore
AGABITI, CAMILLA
URN
etd-03282025-144252
Titolo
Elephant trunk-Inspired Reaching and Grasping Strategies for Soft Robots Control
Settore scientifico disciplinare
ING-IND/34
Corso di studi
Istituto di Biorobotica - PHD IN BIOROBOTICA
Relatori
relatore Dott. FALOTICO, EGIDIO
Parole chiave
- Soft Robotics
- Bioinspired Control
- Elephant Trunk Modeling
- Rod Theory
- Whole-arm Grasping
- Reinforcement Learning
- Space Applications
Data inizio appello
27/04/2026;
Disponibilità
parziale
Riassunto analitico
This Ph.D. thesis investigates biologically inspired modeling and control
of soft continuum robots, informed by the functional principles of reaching
movements and grasping behaviors observed in the elephant trunk. The
research is structured around two main contributions. The first focuses
on the development of a three-dimensional dynamic model of the elephant
trunk using rod theory, capturing its structure as a segmented system with
muscle-like actuation. Using experimental motion data, the model enables
accurate simulation of planar reaching tasks and allows the derivation of
linear force–shape mappings, termed stereotypical laws, which relate desired
trunk configurations to internal actuation forces. The extracted laws reveal
that trunk postures are associated with repeatable internal force patterns,
offering a framework to determine actuation strategies for generating specific
shapes during planar reaching. By distilling the trunk’s complex dynamics
into simple, linear relationships, this work takes a first step toward bridging
the gap between desired configurations and the forces required to realize
them.
The second contribution concerns the design of whole-arm grasping
strategies for soft robotic arms, inspired by the grasping behavior of the
elephant trunk. The strategy is implemented in a simulation framework
combining Finite Element Modeling (FEM) and Reinforcement Learning
(RL), and tested in the context of space debris capture. By introducing
space dynamics concepts, such as free-floating behavior and relative motion
between robot and target, the simulations demonstrate the effectiveness
of the strategy and its robustness to perturbations. Curriculum learning
techniques further improve policy convergence and generalization.
Overall, the thesis establishes a methodology for deeper exploration of
elephant trunk motion strategies and their translation into advanced robotic
control frameworks, while also demonstrating the possibility of developing
simplified grasping strategies for soft robotic arms operating in complex and
unstructured environments, such as space debris capture.
of soft continuum robots, informed by the functional principles of reaching
movements and grasping behaviors observed in the elephant trunk. The
research is structured around two main contributions. The first focuses
on the development of a three-dimensional dynamic model of the elephant
trunk using rod theory, capturing its structure as a segmented system with
muscle-like actuation. Using experimental motion data, the model enables
accurate simulation of planar reaching tasks and allows the derivation of
linear force–shape mappings, termed stereotypical laws, which relate desired
trunk configurations to internal actuation forces. The extracted laws reveal
that trunk postures are associated with repeatable internal force patterns,
offering a framework to determine actuation strategies for generating specific
shapes during planar reaching. By distilling the trunk’s complex dynamics
into simple, linear relationships, this work takes a first step toward bridging
the gap between desired configurations and the forces required to realize
them.
The second contribution concerns the design of whole-arm grasping
strategies for soft robotic arms, inspired by the grasping behavior of the
elephant trunk. The strategy is implemented in a simulation framework
combining Finite Element Modeling (FEM) and Reinforcement Learning
(RL), and tested in the context of space debris capture. By introducing
space dynamics concepts, such as free-floating behavior and relative motion
between robot and target, the simulations demonstrate the effectiveness
of the strategy and its robustness to perturbations. Curriculum learning
techniques further improve policy convergence and generalization.
Overall, the thesis establishes a methodology for deeper exploration of
elephant trunk motion strategies and their translation into advanced robotic
control frameworks, while also demonstrating the possibility of developing
simplified grasping strategies for soft robotic arms operating in complex and
unstructured environments, such as space debris capture.
File
| Nome file | Dimensione |
|---|---|
Ci sono 1 file riservati su richiesta dell'autore. |
|