Robot's Voice Helps It Find Its Way
Posted 10 Dec 2003 at 18:48 UTC by Rog-a-matic
Researchers at the University of Toronto have are using a Trilobot from Arrick Robotics to navigate usin using its own voice. A paper will be available online in a future issue of the journal Information Fusion. The system could be used as a tour guide within two years according to its inventors.
More info..., posted 10 Dec 2003 at 19:49 UTC by steve »
Not entirely appropriate, but when I read the linked piece about this
robot in Scientific American, the old Victor record logo came to
mind. See: http://www.hauntedink.com/ghost/victor-label.JPG
To me a robot talking to itself all the time would start to get
aggravating after a while.
What happened to Ultrasound? It doesn't bother anyone except bats, cats
and dogs maybe.
using Ultrasound would be better I think.
Trilobot does have ultrasound.
This demo was just to show that the robot could be located by using
sound output and microphones around the room. Sort of like sound
GPS. They really couldn't have done the same thing with ultrasound
because ultrasound is directional so it would be hard to pick up on
more than one microphone around the room to triangulate where the
robot was. Perhaps the robot could shoot some ultrasonic sound and
then turn it's head and shoot again at another microphone and turn
it's head the other way again to shoot some ultrasound, but that would
be time consuming.
The question I'm wondering is if it was used as a tourguide, would the
people in the room who were speaking ruin this technique. IE:
wouldn't this robot be the only one allowed to speak so the
microphones around the room could sense where it is. If others are
speaking, how could it tell the difference and filter out everyone
Therefore, I think this is a cool technique for something like a
vacuum while you're away bot where you don't expect anyone else to
speak or a security guard robot, but for a tour guide, I don't see it.
I love the Trilobot! It's a very impressive robot. It has just about
everything you could wish for in a feature on a robot.
I received this email from Parham Aarabi today:
Actually, we have been working on this problem for a long time. The
key is that the Robot's phrases are pre-recorded (i.e. we can extract a
speech signature from these phrases beforehand). Afterwards, we look for
those specific speech signatures (the ones that only correspond to the
robot's speech, and not the other conversations in the room), and
localize based on that. Hence, localization in noisy environments, with
many people and numerous conversations, is possible. Of course, it is
to a certain extent, but if you have several people shouting as loud as
they can, our localization system WILL fail! But, in most typical
situations it does fine. Thanks for your feedback and your comments ...