PPGEE Doctoral Thesis Defense Session 46st - GRASPING AND IDENTIFYING OBJECTS WITH DEEP LEARNING METHODS

STUDENT: CAIO CRISTIANO BARROS VITURINO

DATE: 12/01/2023

TIME: 14:00

LOCATION: https://conferenciaweb.rnp.br/ufba/lar-laboratorio-de-robotica-ufba

TITLE:

 

GRASPING AND IDENTIFYING OBJECTS WITH DEEP LEARNING METHODS

 

 

KEYWORDS:

 

Robotic Grasping, Convolutional Neural Networks, Robotic Manipulators

 

 

PAGES: 144

SUMMARY:

 

In recent years, robotic grasping methods based on deep learning have surpassed analytical methods in performance. Despite the results obtained, most of these methods only use planar prehensions due to the high computational cost found in 6D prehensions. However, planar prehensions have spatial limitations that prevent their applicability in complex environments, such as the prehension of objects manufactured inside 3D printers. Furthermore, some robotic grasping techniques generate only one viable grasp per object. On the other hand, it is necessary to obtain multiple possible prehensions per object, because not every prehension generated is kinematically viable or does not collide with other nearby obstacles. To address these limitations, a robotic grasping method capable of generating several 6D selective grasps per object is proposed, avoiding collisions with adjacent obstacles. Grip tests were carried out in an Additive Manufacturing Unit, which presents a considerable level of complexity due to the high probability of collisions. Experimental results prove that it is possible to achieve a considerable success rate in grasping manufactured objects. The UR5 robotic arm, the Intel Realsense D435 camera and the Robotiq 2F-140 end effector are used to validate the proposed method in real experiments.

 

 

BOARD MEMBERS:

President - ANDRE GUSTAVO SCOLARI CONCEICAO - UFBA (Advisor)

Intern - TIAGO TRINDADE RIBEIRO - UFBA

Internal - EDUARDO FURTADO DE SIMAS FILHO _ UFBA

External to the Institution - VALDIR GRASSI JR - USP

External to the Institution - EDUARDO TELMO FONSECA SANTOS - IFBA

​

 

Published on 11/28/2023 

Em 03/12/2023