- 6 -
relative to the car it is much more difficult to understand where the car is within the environment than it
otherwise would be. We find that, given this constraint, to drive well one needs to ‘internalize’ the environ-
ment in which one is going to drive before attempting to do it. This works well except when one finds him-
self facing a blank wall and able to see no other features. At this point being able to look around directly
would be a major help. The current way to accomplish this is to back up while turning the wheels sharply
until one again recognizes where the car is in the environment. (We have driven the car mostly in the corri-
dors of Bell Labs which are narrow and uniformly beige except for doors. This is as difficult a case as one
is likely to encounter.)
6. Other Presences
The setup we have allows operation with any presence using a two channel radio. We had so much
fun with the toy car we decided to fly a model airplane. The airplane requires a four channel radio and con-
troller. The ailerons and elevator are controlled analogously to the car’s steering and throttle respectively.
The throttle and the rudder are adjusted (trimmed really) using keys on the keyboard for increasing and
decreasing them. The challenge is to build a unit controllable over an Ethernet that can do software video
compression with small enough delay so that we can avoid crashing. (We have completed the airplane, but
haven’t flown it.)
We would like to try a model sailboat, because one of the authors found that sailing a radio controlled
model sailboat was much more difficult than sailing a real boat. This was due mostly to the difference in
visual perspective from the edge of the pond compared to the cockpit of a boat. (A lot of the fine control in
sailing depends on tactile feedback.) The ultimate would be to fly a toy helicopter, but this will involve a
different input scheme and more complex software and hardware, since model helicopters use five channel
radios.
We have contemplated building a telepresence for a long distance telecommuting application. Basi-
cally, one needs an enhanced version of the model car with two way audio. The unit needs to have more
battery power and some safety limit switches but is otherwise quite similar to the toy car. We decided
against doing this because of problems in the building environment. Such a remote presence in our envi-
ronment needs to range through a large complex of buildings and requires substantial battery power. It
needs to be strong enough to open doors and agile enough to do so as well as to call elevators. All of this
requires a large, heavy device that would be a danger if it were to, say, fall down a flight of stairs. Its grip
could likewise be a danger. If we had a door-less single floor in which to operate, such a presence could be
made both more simply and safely. It would still need limit switches, but no longer needs to really grip. It
could simply come in to chat and observe and would offer a long distance commuter a much better sense of
belonging to a group.
7. Potential Applications
Beyond the obvious application as a toy controller for the well- heeled hobbyist, what might be done
with this technology particularly when it is used over a network? We have hardware to allow the car to be
controlled from a slightly modified voice grade line videophone, in addition to the IP network control
already mentioned. In a network the car or something similar becomes useful for remote surveillance. One
can easily imagine a doctor wandering around a clinic looking a patients. One could also tour a house for
sale from a realtors office. A potential application is in the making of movies or videos. We have heard that
model helicopters, conventionally flown, are already occasionally used for this. Certainly somewhat simi-
lar of technology has been used to look around dangerous environments before, but networking would
make this sort of application even more appealing. Of course, the car might have to be modified consider-
ably for many dangerous environments for reasons of functionality or safety.
8. Conclusions
We have described a telepresence based on a modified model car, which is equipped with a video
camera and can be controlled over the Internet. The key contributions of our work are to recognize that
such presences can be built with off the shelf components. Second, we have given much thought to an easy
to use interface, which has proved to be successful. Third, our software is compatible with existing packet
video standards, and thus can be widely used. Finally, we have tuned the low bandwidth available from the
Comentarios a estos manuales