Join us > Job offer

PhD Student - Topological measures to plan flexible objects bimanual robotic manipulation (ERC CLOTHILDE)

Date of call

31/07/2018

Deadline

15/12/2018

Responsible

Júlia Borràs Sol

Type

Temporary

Warning Warning: this offer is out of date!

Description

The "Institut de Robòtica i Informàtica Industrial (IRI), CSIC-UPC", is seeking a PhD student to work under the framework of the European Project CLOTHILDE (ERC Advanced Grant 2016).

The PhD student will be supervised by Dr. Júlia Borràs Sol and Prof. Federico Thomas.

ABOUT IRI:
IRI is a Joint University Research Institute participated by the Spanish National Research Council (CSIC) and the Technical University of Catalonia (UPC) that conducts research in human-centered robotics and automatic control. The institute is a key player in the Spanish robotics and automatic control scenes, and a valued participant in a large number of international collaborations including more than ten ongoing H2020 projects and an ERC Advanced Grant. Recently, is has been recognized as a Maria de Maeztu Excellence Unit, the main accreditation given by the Spanish Government to research units that stand out for the impact and international relevance of their results.

ABOUT CLOTHILDE:
The European Project CLOTHILDE (Cloth manipulation learning from demonstration) is an ERC Advanced Grant, awarded to Dr. Carme Torras, that started in January 2018.

The project aim is to establish the foundations of versatile cloth manipulation by robots. The theoretical ground will be to link machine learning and computational topology methods, so as to come up with a theory of cloth deformation under manipulation leading to a general framework for robots to learn to manipulate garments from human demonstrations. Such framework will encompass: non-expert teaching of a task, robot perception and skill learning, task-oriented cloth representation, probabilistic planning, robot task execution in varying initial conditions, failure diagnosis and informed requests for human help.

ABOUT THE PREDOC POSITION
Textile manipulation is intrinsically bimanual and its main challenge is how to predict the behaviour of completely shapeless objects when they are manipulated, in order to be able to plan an action such as folding or covering an object. We want to exploit different gripper designs that allow us to conform desired shapes of the pieces of cloth that are being manipulated in a way that is meaningful for different geometric representations of the configuration state of a bimanual system of grippers+cloth. The main idea behind this project will be to define new representations in terms of points or oriented lines that allow us to extract topological and shape information from each configuration of our grippers+cloth system. In the long term, we want to plan motions between different configurations exploiting optimization methods based on topological and geometric indexes.

To start, we will study the Gauss Linking Integral [1], as an index that measures the writhe between two open polygonal curves. This index has been recently used to synthesize motions for humanoid robots [2], grasping [3] and manipulation through holes in the environment [4]. We will revisit the computation of the index, its geometric interpretation and possible improvements, and in parallel we will explore the 2D extension of this index, starting by studying the writhe matrix between different significant curves on a 2D surface.

[1] Ricca, Renzo L., and Bernardo Nipoti. "Gauss'Linking Number revisited." Journal of Knot Theory and Its Ramifications 20.10 (2011): 1325-1343.

[2] Ho, Edmond SL, and Taku Komura. "Indexing and retrieving motions of characters in close contact." IEEE Transactions on Visualization & Computer Graphics 3 (2008): 481-492.

[3] Zarubin, Dmitry, et al. "Topological synergies for grasp transfer." Hand Synergies—How to Tame the Complexity of Grapsing, Workshop, IEEE International Conference on Robotics and Automation (ICRA 2013), Karlsruhe, Germany. Citeseer. 2013.

[4] Ivan, Vladimir, et al. "Topology-based representations for motion planning and generalization in dynamic environments with interactions." The International Journal of Robotics Research 32.9-10 (2013): 1151-1163.

Requirements

Qualifications: The candidate is expected to have a Master in mathematics, computer science or a related field,
and have interest in Geometry and Computational Topology topics
Languages: Fluent in English

Contacts

Please, send your documentation before 15/12/2018 to:

Email: jobs (at) iri.upc.edu
Address: C/ Llorens i Artigas 4-6, 2a planta. 08028 Barcelona, Spain.
Fax: 93 401 57 50
Note: Please, remember to reference the offer (ID-65) and include the following documentation: CV and academic transcripts