Solving Assembly Tasks via Deep Learning


61e81ed2 7536 4d1f 9720 c2f69cee753d.png 500 500

Starting Date: earliest start: (2019-02-01) latest end: (2019-09-30)

Organization: Advanced Interactive Technologies

Involved Host(s): Stevsic Stefan

Abstract: The goal of the project is to enable precise positioning of the objects in 6D space using RGB camera as a main source of feedback.

Description: Robotic arms can precisely position objects in space. This enables robots to do complex assembly tasks in factory settings where position of all objects is precisely known. However, in home environments, where the configuration of objects is constantly changing, this is still an open problem. Recent methods from computer vision enable to determine position and orientation of objects in the scene. This opens up possibilities to do assembly tasks in home environments and in collaboration with humans. Current methods rely on the facts that objects are fully visible and that the scene is static. However, when working with big objects or in collaboration with humans these assumptions are not valid anymore. The goal of this project is to investigate possibilities to overcome these issues. State-of-the-art methods need to be extended to work in a closed loop setting and to cases when objects are only partially visible.

Contact Details: stefan.stevsic@inf.ethz.ch

Robotics Deep Learning Robot Arms


Labels: IDEA League Student Grant (IDL) Master Thesis CLS Student Project (MPG ETH CLS) ETH Organization's Labels (ETHZ)
Topics: Information, Computing and Communication Sciences Engineering and Technology
Applicant Organizations: ETH Zurich EPFL - Ecole Polytechnique Fédérale de Lausanne