DAGHAN 699 week2 Learning about natural human-robot ...

bouncerarcheryAI and Robotics

Nov 14, 2013 (3 years and 10 months ago)

63 views

Learning about natural

human

robot interaction styles

By

Tamie Salter,

Kerstin Dautenhahn,

Ren
´
e te Boekhorst

AuRoRA project


Attempting to develop robots for use
with children with
autism

in a
threapeutic and educational context


This study is a part of the AuRoRA project

Beside Vision and Sound


Touch Sensors can be used for HRI
(Human Robot Interaction):


IR sensors:


Cheap


Shorter than 2cm: touch


Kid Classification


Type A (Boisterous/Active):


He was considered naughty at home and school


He seemed confident, unafraid and active


Type C (Passive/Shy):


They usually did as they were told both at school
and at home


they seemed to require the security of being with
a familiar adult


they did not readily explore on their own


Type B :


Rest

AIM

CAN WE CLASSIFY KIDS VIA KID
-
ROBOT INTERACTION USING
INFRARED SENSORS AS TOUCH
SENSORS?

Experiment Setup


Experiment Setup


Pekee robot:


Dimensions:


400mm long, 255mm wide, 210mm heigh


Weight:


2.9 kg


Speed:


Max 6km/h, used: 3km/h


Sensors:


15 Infrared sensors (shorter than 2cm: touch)


6 front, 3 left, 3 right, 3 back



Experiment Setup


Area:


2m x 2m


Wooden borders


Time:


1 minute (longer times can cause boredom/fright)


Robot Strategy


Movement with obstacle avoidance (same in all
experiments)

Experiment Setup


6 kids with ages 5 to 7


Victor (type A)


Thomas (type A)


Bill (type B)


Chris (type B)


Len (type C)


Colin (type C)

Results


Sensor Readings


Results


Sensor Readings: Dendrogram
Clustering


Results


Behavioral Activity Score

Results


Behavioral Activity Score: Dendrogram
Clustering


Future Work


After classifiying the kid, adaptively
change the behaviour according to kid


A more dynamic/active robot for type A
kids


Turn around, beep etc.


A calmer robot for type C kids


Move slowly, sometimes stop etc.