Information Technology Reference
In-Depth Information
Fig. 7.
(left)
Experimental setup (right) Experimental scene
5.4
Participant and Ex
perimental Flow
Thirty-six participants part
females. We assigned 18 p
and the remaining 18 parti
each group were paired (a m
The experiment was divi
the experiment, we instruc
need to create general com
task. Do not use any kind
instruction served the purpo
At first, each manipulat
body. Then, the pairs starte
gave instructions for any k
each pair made trial-and-err
When the pair determine
ymore, the experiment mo
one of the three examples i
assemble the building with
answered the questionnaire
ticipated in the experiment. There were 34 males an
participants (including one female) to the humanlike gro
icipants to the head-fixed group. Eighteen participants
manipulator and a player). Each group had nine pairs.
ided into the testing phase and the recording phase. Bef
cted the participants as follows: "In this experiment, y
mmunication strategies for the robot with the assembl
d of code that is incomprehensible to other person." T
ose to keep the designed communication strategies gener
tor calibrated the robot parameters to the scale of his/
ed the testing phase. During this phase, each manipula
kind of buildings she/he could imagine. The members
rors efforts and improved their communication strategie
ed that they could not improve their manipulation time
ved to the recording phase. We assigned the manipula
in Fig. 6 and recorded the interaction. The pair required
hin 300 s. When the recording finished, each particip
and the experiment was terminated.
nd 2
oup
s on
fore
you
ling
This
ral.
/her
ator
s in
es.
an-
ator
d to
pant
5.5
Prediction: Overlap
pped Time Ratio and Confirmation Ratio
Pre-evaluation confirmed t
raction with instructions on
that head movement played
itself is difficult to evaluate
lacks verbal cues.
We used the overlap tim
each manipulator and playe
increase in overlapped verb
We extended this idea to n
problems, the behavior of th
that turn-taking behavior was used in human-human in
n how to assemble the blocks. The evaluation also revea
d a key role on regulating turn taking. However, turn tak
e by video recording data, especially when this evaluat
nte-
aled
king
tion
me ratio as an indicator of turn-taking behavior betw
er. A previous HRI study using humanoids showed that
bal cues of both persons suggests failure of turn taking [2
nonverbal situations. If turn taking took place without
he robot and the human did not overlap. In contrast, if t
ween
the
25].
any
turn
