ResearchBib Share Your Research, Maximize Your Social Impacts
Sign for Notice Everyday Sign up >> Login

PHD Opening in "Learning visual cues of interactions for humanoid robots"

Country/Region : Italy

Website : http://phd.dibris.unige.it/biorob

Description

Applications to the PhD in Bioengineering and Robotics - Università degli Studi di Genova are open!
The Phd program in Biogengineering and Robotics is rooted in the multiple ongoing collaborations between Università degli Studi di Genova/University of Genoa (www.unige.it), Fondazione Istituto Italiano di Tecnologia/Italian Institute of Technology Foundation (www.iit.it) and the University of Cagliari (www.unica.it)
Deadline for application (strict): 22/08/2014 - 12:00 (noon).
More information on the PhD program, its curricula, research themes, and application modalities can be found at http://phd.dibris.unige.it/biorob/ and http://www.iit.it/en/openings/phd-calls/2015.html.
Within the program, I would like to draw your attention on the following research theme, highly interdisciplinary, with elements of computer science, cognitive science and robotics, and very ambitious in its scope:
Learning visual cues of interactions for humanoid robots
Theme n. 10 - curriculum on “Cognitive Sciences, Interactive and Rehabilitation Technologies”.
Tutors: Prof. Francesca Odone (DIBRIS Università di Genova) Prof. Giulio Sandini (RBCS Dept., IIT), Dr. Alessandra Sciutti (RBCS Dept., IIT)
Description: Since the earliest stage of their development, humans learn how to interact with other people. More specifically, babies begin to decode their own movements by acquiring visual evidence of the result of their motor commands. Then, as soon as they learn to perform a given action they also start to anticipate its outcome even when they see it executed by others. Hence, the knowledge accumulated by acting is exploited to understand and predict someone else’s intention.
The general goal of this project will be to provide iCub with a similar ability that will guarantee a more natural and proficient interaction with human users. The pipeline of the computational system includes an attentional module able to localize the portion of the scene containing biological motion. Then, iCub exploits this information to focus on the selected area and to describe it more precisely using its stereo system.
The visual properties of the observed movements will be used by the robot to decode the intentionality of the interacting partner by using machine learning techniques. View-invariance will be one of the crucial aspects of the system, together with the capability of inferring information about the 3D orientation of an agent.
Reference:
Sciutti A., Noceti N., Rea F., Odone F., Verri A. & Sandini G. 2014,'The informative content of optical flow features of biological motion', 37th European Conference on Visual Perception (ECVP 2014), Belgrade, Serbia, August, 24-28, 2014
See also:
Sciutti A., Patanè L., Nori F. & Sandini G., 2014, ‘Understanding object weight from human and humanoid lifting actions’, IEEE Transactions on Autonomous Mental Development, vol. 6, no. 2, pp. 80-92, doi: 10.1109/TAMD.2014.2312399
Requirements: degree in robotics, bioengineering, computer science, computer engineering, or related disciplines, attitude for problem solving, basic skills on c++ programming. A background on computer vision and machine learning is an asset.
Highly motivated students are welcome to apply!
Contacts Applicants are strongly encouraged to contact the perspective tutors before they submit their application:
francesca.odone-AT-unige.it, giulio.sandini-AT-iit.it, alessandra.sciutti-AT-iit.it

Last modified: 2014-07-12 12:23:58