P1: IXL/GEG P2: IXL
WL040A-22 WL040/Bidgolio-Vol I WL040-Sample.cls June 20, 2003 17:34 Char Count= 0
594 VIRTUALREALITY ON THEINTERNET:COLLABORATIVEVIRTUALREALITYFigure 2: Remote participants in collaborative virtual reality sessions can be seen in the shared virtual space in several ways.
(Left) The user is interacting with the articulated computer-generated avatar of a remote participant in the CAVE. (Right) The user
is interacting with the live video avatar of a remote participant on the ImmersaDesk2.network bandwidth and careful setup to achieve high
quality. More details on video avatars can be found in
Rajan, et al. (2002; see Figure 2). Because the local com-
puter tracks the user’s position to correctly draw the vir-
tual world, the computer can send that position and ori-
entation data over the network to remote sites to posi-
tion an avatar at the same point in the remote virtual
environment as the real user occupies in the local vir-
tual environment. The local computer receives updated
position information from the trackers many times per
second. Because we want to get that information to the
other collaborators as quickly as possible, we use an un-
reliable user-datagram protocol (UDP) connection, rather
than a reliable transmission-control protocol (TCP) con-
nection. If one of the packets of position/orientation data
is lost, then another will be along shortly. We typically send
out avatar position and orientation information 15 times
per second. Because each tracker returns three positions
(X,Y, andZ) and three rotations (roll, pitch, and yaw),
each represented by a 32-bit floating point number, trans-
mitting these values 15 times per second requires only 3-K
bits per second per tracker. Typically, there will be at least
two trackers in use, one for the head and one for the hand,
bringing the bandwidth needs up to 6-K bits per second.
On the remote side, the position and orientation of the
avatar is usually interpolated from the network data to
smooth out the motions of the avatar.
Avatars are also useful in alerting other users to a per-
son’s next actions. For example, the declaration “I’m going
to move this chair” combined with the visual cue of your
avatar standing next to a chair and pointing at it alerts
other users to the fact that you are about to grab it. As in
real life, we know that if another collaborator tries to grab
the chair at the same time, it will be awkward.
Different avatar forms are useful in different virtual
worlds. In certain situations, such as pointing at an ob-
ject on the ImmersaDesk, pointers can be better than full
avatar bodies, because, just as in real life, a remote user’savatar body may block your view if you are standing close
together. For small objects, short pointers work well, but
in larger spaces it helps to have a long beam, allowing
users to point accurately at objects 100 feet away.
Avatars with highly stylized bodies are easier to differ-
entiate within the environment but they may not be appro-
priate for all types of users. First-time users tend to laugh
the first time they meet articulated avatars, who appear
to be living cartoon characters, but once the characters
begin interacting with them, they quickly adapt and have
no trouble treating the characters as living persons. Other
users desire a more “serious” representation of themselves
in the virtual world. Using photographs to generate avatar
heads that look like their actual users is a way to help
bridge recognition between the virtual world and the real
world. Name tags can also be useful in identifying avatars,
much as they are useful in real life (see Figure 3).
People feel very uncomfortable if they accidentally
walk through another person in the shared virtual world
and apologize profusely. In general, people tend to main-
tain an appropriate distance from other users and try to
avoid violating their personal space. Preventing collisions
by not allowing one user to move through another can
help maintain social comfort. However, this collision de-
tection can be a hindrance if several people are trying to
maneuver down a narrow hallway in the virtual world.
In certain situations, it is good to have real-world con-
straints, such as gravity and collision detection, and at
other times it is good to be able to turn reality off and be
able to do more than you could in real space.
Almost everyone’s initial steps into a collaborative vir-
tual world require a tour guide. Initially, this tour guide
was a person who would stand next to the new user,
show him the controls of the virtual reality hardware, and
point out the features of the virtual space. Now, we know
that it is useful to have this tour guide give the initial
tour remotely as an avatar. This draws the user into the
shared virtual space and immediately starts up a dialogue