
An Internet Accessible Telepresence
A.E. Kaplan, S. Keshav, N.L. Schryer, J. H. Venutolo
AT&T Bell Laboratories, Murray Hill, N.J.
{aek keshav nls jhv}@research.att.com
ABSTRACT
The US Vice President, Al Gore, in a speech on the information superhighway,
suggested that it could be used to remotely control a nuclear reactor. We don’t have
enough confidence in computer software, hardware, or networks to try this experiment,
but have instead build an Internet-accessible, remote-controlled model car† that provides
a "race driver’s" view via a video camera mounted on the model car. The remote user can
see live video from the car, and using a mouse, control the speed and direction of the car.
The challenge was to build a car that could be controlled by novice users in narrow
corridors, and that would work not only with the full motion video that the car natively
provides, but also with the limited size and frame rate video available over the Internet
multicast backbone [1]. We have built a car that has been driven from a site 50 miles
away over a 56kbps IP link using nv format video at as little as one frame per second and
at as low as 100 by 100 pixel resolution. We also built hardware to control the car using
a slightly modified voice grade channel videophone. Our experience leads us to believe
that it is now possible to put together readily available hardware and software compo-
nents to build a cheap and effective telepresence.
1. Introduction
The original motivation for building a remote-controlled car was as an application for a locally devel-
oped infrared local area network, where the car would be controlled and transmit live compressed video
over the IR network. The toy car came together more quickly that the network and other necessary parts (a
digital video imager, in particular), and so we ended up building the Internet accessible telepresence that we
describe here.
One key goal was to allow novice users to be able to walk up to a workstation and control the car,
thus the user interface required careful thought. Second, since we wanted to control the car over the Inter-
net, we wanted it to deal with slow frame rates and variable delay. To do so, we needed to highly compress
the video in order to get even a few frames per second. Third, we wanted to multicast the car’s video over
the Internet Multicast Backbone (MBONE) [1] as well as over point to point logical links. Finally, to mini-
mize effort, we wanted to reuse standard software modules as much as possible.
All of these goals were met in whole or in part in our design. The car does run over the MBONE,
and since this video is compatible with the nv program [2, 3], this video can be received by a fairly large
number of receivers. We were able to use much of the nv software and retail hardware to build the car.
However, our desire to make the presence easy to pilot, even by a user unfamiliar with it or its environment
were not fully realized. We discuss why this is so in the Section 5.
__________________
†
Comentarios a estos manuales