Chargement...
 

Interaction

Domaine
Interaction
Domain - extra
Année
2013
Starting
01/09/2013
État
Open
Sujet
Transfert Dynamique et Adaptation des Compétences Sociales entre un Agent Virtuel et un Robot
Thesis advisor
MARTIN Jean-Claude
Co-advisors
Adriana TAPUS (ENSTA-ParisTech)
Laboratory
Collaborations
LIMSI-CNRS / ENSTA-ParisTECH
Abstract
Artificial systems are capable of limited social and spatial interaction with human
users. There is very little research that permits the collaboration, adaptation, and
transfer of social knowledge between a physically embodied robot and a virtual agent. We posit
that the usage of these two complementary types of embodiments of artificial agents (i.e.,
physical and virtual) will allow a more comprehensive social and spatial interaction with users.
The research that will be developed in this work will be structured as follows: (1) specifications
of the different scenarios of social interaction between a user, a robot and a virtual agent ; (2)
definition of an abstract model of a social interaction between artificial agents and a user; (3)
development of an architecture that allows the dynamic transfer and exchange of social
knowledge (previously predefined or learnt) between artificial agents ; (4) implementation on both MEKA robot (ENSTA-ParisTech) and MARC virtual agents (LIM
Context
Affective computing and social signal processing have become the focus of several
research teams. Some teams focus on increasing the social capabilities of virtual
agents. Other teams focus on increasing the social capabilities of robots. Recent projects suggest
a scientific convergence between researchers working on interaction with a virtual agent and
researchers working on interaction with a physical robot.
Some virtual agents and some robots share some common features (eg they can express
emotions using facial expressions), but main differences remain between these two types of
agents:
● the robot has a physical embodiment and can move and act in the real world ; but it
remains constrained in terms of expressions due to the physical limits of its motors
● the virtual agent is not physically limited in terms of expressions (for example it can
display exaggerated facial expressions with a very high dynamics) but it does not have a
direct physical access to the real physic
Objectives
The TACOS project does not aim at comparing separately a physical robot and a virtual agent. It
does not either tries to combine them as a single physical / virtual entity.
Instead, our project aims at exploring how to make the best use of a physical robot and a virtual
agent by enabling them to collaborate and to dynamically exchange knowledge or allocate tasks
depending on the social and physical context.
It will enable to improve the current models of interaction for social robotics.

The goal of the TACOS project is to enable a dynamic collaboration between a virtual agent and
a physical robot. This will include for example learning mechanisms for selecting appropriate
strategies and actions in a given social and spatial context.
In this section, we first describe the two platforms that will be used: MEKA and MARC. We
then explain how we propose to develop the exchange of social and spatial knowledge between
these two artificial agents.
Work program
● Task 1: specifications of the different scenarios of social interaction between a user, a
robot and a virtual agent ; (T0 - T0+6)
● Task 2: definition of an abstract model of a social interaction between artificial agents
and a user (T0+6 - T0+18)
● Task 3: development of an architecture that allows the dynamic transfer and exchange of
social knowledge (previously predefined or learnt) between artificial agents ; (T0+12 -
T0+24)
● Task 4: implementation of this architecture on both Meka robot (ENSTA-ParisTech) and
Marc virtual agent (LIMSI-CNRS) (T0+18 - T0+30)
● Task 5: evaluation of the impact of the proposed methodology on the interaction with
users. (T0+12 - T0+36)
Extra information
Prerequisite
Détails
Expected funding
Institutional funding
Status of funding
Expected
Candidates
Utilisateur
jean-claude.martin
Créé
Vendredi 14 juin 2013 21:12:45 CEST
dernière modif.
Lundi 08 juillet 2013 15:18:06 CEST

Fichiers joints

 filenamecrééhitsfilesize 
Aucun fichier joint à cette fiche


Ecole Doctorale Informatique Paris-Sud


Directrice
Nicole Bidoit
Assistante
Stéphanie Druetta
Conseiller aux thèses
Dominique Gouyou-Beauchamps

ED 427 - Université Paris-Sud
UFR Sciences Orsay
Bat 650 - aile nord - 417
Tel : 01 69 15 63 19
Fax : 01 69 15 63 87
courriel: ed-info à lri.fr