logo SBA

ETD

Archivio digitale delle tesi discusse presso l’Università di Pisa

Tesi etd-02032016-184609


Tipo di tesi
Tesi di laurea magistrale
Autore
PARDI, TOMMASO
URN
etd-02032016-184609
Titolo
Design, development and control of a soft robot for object manipulation in Amazon factory-like environment.
Dipartimento
INGEGNERIA DELL'INFORMAZIONE
Corso di studi
INGEGNERIA ROBOTICA E DELL'AUTOMAZIONE
Relatori
relatore Prof. Bicchi, Antonio
Parole chiave
  • environmental constraints
  • grasp strategies
  • motion planner
  • pick and place
  • softhand
  • strategie di approccio
  • vincoli ambientali
  • VSA
Data inizio appello
25/02/2016
Consultabilità
Completa
Riassunto
In last years robotics took advantage of a strong improvement in hardware design which produced a new technology, called soft robotics, based on new actuators that can modulate their own compliance and have a safe interaction with unstructured environments.

Soft robots compete in DARPA Robotic challenge where humanoid robots attempted to execute operations, in a simulated and real scenarios, WALK-man by IIT and University of Pisa, but are also used in industry, ABB company introduced a new soft gripper to handle fragile things.

The Soft Robotics certainly has potential but also posed new challenges at planning and control level, researchers propose different approaches to solve problem but, at moment, it represent an open issue.

In this thesis we investigated one of most interesting rising approach take advantage of softness which regard the exploitation of environment constrain (EC) like an help on performing grasp and/or manipulation tasks.
We designed a \textit{Pick 'n Place} manipulator composed of Variable Stiffness Actuators (VSA) and an end-effector call Pisa/IIT SoftHand, an underactuated anthropomorphic hand with 19 DOF but only one motor. Moreover we provide a three-dimensional depth sensor mounted on top of mechanical structure which gives us geometry of actual scene.

In the first part of the thesis, we analyzed how Pisa/IIT SoftHand works in specific situations through empirical experiments, we used Handle device to approach objects in a human-like way and gathered informations in a grasp database, in next step we elaborated manipulation strategies based on previous result and a set of objects with various dimensions. After a study on observed strategies we summarize them in subgroup which depends on objects properties.
In last stage, we tested new approached in a constrained environment which is represented by the Amazon shelf and present result.
As a case of study we participate at \textit{Amazon Picking Challenge}, in this competition the robot challenged other teams in a \textit{Pick 'n Place} task.


File