UNIVERSITY OF HERTFORDSHIRE
             COMPUTER SCIENCE RESEARCH COLLOQUIUM

                            presents

         "Cognitive Humanoid Robots Interacting with Humans:
    Adaptive Real-time Synchronization and Communicative Imitation"


                    Konstantinos Theofilis
          (Adaptive Systems Research Group, School of
          Computer Science, University of Hertfordshire)

                     11 March 2015 (Wednesday)
                          11 am -12 noon

                  Hatfield, College Lane Campus
                       Seminar Room D102

                  Everyone is Welcome to Attend

                  Refreshments will be available


Abstract:

In Human-Human Interaction (HHI), imitation, apart from its learning
function, has a communicative function that facilitates the exchange
of information and the grounding of a common context. The prominent
characteristic of communicative imitation is synchronization, in the
sense of coordinated imitation. Studies in HHI have shown that humans
tend to synchronize their behaviours on many channels, both consciously
and unconsciously.

This synchronization is also a central theme for the joint cognitive
offloading paradigm, where the two actors act as a joint cognitive
system. That cognitive offloading is considered the answer to the
"dangerous degrees of freedom" problem, i.e., how to reduce detailed
planning for each behavioural channel during the interaction.
This paradigm is especially suited to robotics, where action planning
can be severely limited by the limited computational resources available.
In this talk, an adaptive real-time system for synchronization and
communicative imitation, implemented on the iCub humanoid robot, is
presented. The system is capable of adaptive multimodal behaviour and
is using an internal scheduler that is affected by the presence or
not of synchronization from the human actor, in a simple interacion
rhythmic game.

In addition to measuring the system's performance, the human
participants' behaviour is evaluated and compared to results from HHI,
using an analysis technique that is novel for Human-Robot Interaction
(HRI) but that steadily gains ground in HHI studies, as it avoid the
disadvantages of traditional statistics in non-linear, time-dependent
systems.
---------------------------------------------------
Hertfordshire Computer Science Research Colloquium
http://cs-colloq.stca.herts.ac.uk