[Comp-neuro] Call for a PhD student in Humanoid Robotics and Cognition, DIST University of Genova and RBCS Italian Institute of Technology

Ryo SAEGUSA ryo.saegusa at iit.it
Mon Sep 13 22:37:45 CEST 2010


We announce a call for a PhD student starting from January 2011.

The topic is "Active perception for learning the self and others" within 
the PhD program of "Life and Humanoid Technologies" in University of 
Genova and Italian Institute of Technology, Genoa, Italy.

For further details concerning the call, please contact: 
ryo.saegusa at iit.it, lorenzo.natale at iit.it, and giorgio.metta at iit.it. The 
official deadline for the application is the 24th September, 2010.

======================================================
Theme 1.8: Active perception for learning the self and others
Tutor: Dr. Ryo Saegusa
N. of available positions: 1

How can a robot identify the self, and associate it with others?

This is a fundamental question for the early life of primates and also 
embodied intelligence (Metta et al.2006, Stoytchev 2007). Monkeys are 
able to recognize their own body under various conditions, and extend 
their body schema while using a tool (Iriki et al.1996,2001). Also, they 
associate others' behaviors with their own (Gallese et al.1996). This 
kind of cognitive functions may have potential to break a limit in 
existing hand-coded intelligence of humanoids, and bring more 
considerate interaction ability with humans.

Our goal is to realize a cognitive system which actively develops its 
perception of itself and others through sensorimotor interaction. The 
humanoid robot iCub (http://www.icub.org) will be the reference platform 
for this theme. Example of cognitive functions to be considered are (1) 
merging vision and proprioception for the acquisition of a body 
representation - distinguish the robot's body from others and the 
environment -, (2) learn hand-eye coordination exploiting knowledge of 
the robot own body, (3) reach and touch an object, then observe the 
effect, and (4) build models of objects. An object in sight is just a 
texture before interaction, but afterwards, the robot can recognize 
independence of the object as well as the executed action itself.

The scope of the theme, moreover, encompasses modeling of humans and 
their intentions. How do primates imitate or mirror others' behaviors, 
and why do they need it? One of interpretations, here we are drawing, is 
that primates might transfer motor intelligence by mirroring actions and 
sharing the intention. Here, we introduce a simple task which uses 
knowledge about objects and their behavior to "exchange" information 
between a human experimenter and the robot. How can we configure the 
interaction for them to share the intention behind the pointing action?

Examples of the solutions are (5) identify the robot hand, the human 
hand, and the object by the acquired knowledge, (6) mirror hand motions 
between the human and the robot by observation and performance, (7) 
mirror object motions (the effect of pushing) as well, (8) associate 
one's pointing action with the other's pushing action based on 
reinforcement feedback signals.

Requirements: engineering or psychology background, (optional) machine 
learning, visual/haptic processing

For further details concerning the research project, please contact: 
ryo.saegusa at iit.it, lorenzo.natale at iit.it, and giorgio.metta at iit.it
======================================================



More information about the Comp-neuro mailing list