Chapter 1 - Introduction
The goal of this research was to determine an intuitive and efficient
approach for facilitating scalable interpersonal communication within the
GreenSpace system. To accomplish this, we evaluated the existing methods for
communication within large-scale virtual environments and then explored the
adaptation of the most promising of those methods to GreenSpace. This thesis
document presents the results of that exploration, including a discussion of
existing multi-user systems and an evaluation of the prototype communication
control system developed through this research.
The rest of this chapter includes background on the GreenSpace project
(which this thesis is a small part of), motivation for this topic of research, and a
description of the thesis document's organization.
The GreenSpace project was born in April of 1993, through discussions
between Human Interface Technology Laboratory (HITLab) director Dr. Thomas
Furness and Fujitsu Research Institute director Dr. Masahiro Kawahata. They
planned for the development and demonstration of "an immersive communication
medium" to allow people on both sides of the Pacific ocean, eventually people all
over the world, to share experiences in a "virtual common" [Dans94][Jone95].
An immersive computer system is one in which the user's senses are
surrounded by computer-generated stimuli that enable a feeling of presence in the
simulated environment. Presence is the subjective feeling of actually being in a
place, whether it is real, augmented, or virtual [Pulk94]. Users are able to look,
listen, navigate, and manipulate objects within the computer generated
environment as if they were physically there. Immersive systems are typically
called virtual reality (VR) or virtual environment systems, although not all systems
that claim to be VR are immersive. Immersion can be accomplished in a number
of ways, including combining head-mounted displays (HMDs), position tracking,
and high-speed interactive 3D computer graphics.
Rather than have the attendees of a meeting physically travel to a common
location, they can use computer networks to meet in a common computer
generated immersive environment, or "virtual common".
In November of 1994, the GreenSpace team, led by Paul Danset,
simutaneously demonstrated a prelimary test-bed at the NICOGRAPH
Convention in Tokyo and the HITLab in Seattle. On four consecutive days,
attendees of NICOGRAPH and visitors tothe HITLab participated in trans-Pacific
virtual teleconferences (see Figure 1.1).
Figure 1.1: Danset, Kawahata, and Furness in GreenSpace
The software system was developed at the HITLab and was duplicated for
the NICOGRAPH site. The hardware setup at each side of the demonstration was
nearly homogeneous and included the following:
- Silicon Graphics (SGI) Onyx for graphics rendering,
- SGI Indy for audio rendering and image capture,
- Polhemus Fasttrack, a magnetic position and orientation tracking system,
for head and hand tracking,
- Virtual Research HMDs, for visual and audio display, and
- various network routing components.
For various reasons, both in hardware and software, the demonstration
was limited to a total of four users at any given time (two at each site). In the
future, GreenSpace plans to support over one hundred users at a time, using
heterogeneous hardware systems and broadband networks. In order to achieve
this, the system software must be scalable to an arbitrary number of users.
The demonstration application was a cooperative game in which the users
attempted to herd flying creatures into goals at the corners of their meeting table.
Of course, this was a toy application intended to show how people in physically
different parts of the world can virtually meet and work together in real-time.
Actual applications of the GreenSpace system include collaborative design, virtual
space teleconferencing, situational training, and entertainment.
The preliminary testbed used for the demonstration has given the
GreenSpace group an opportunity to begin to identify the issues involved in
creating a "immersive communication medium." Work is well underway to create a
more general GreenSpace software infrastructure to achieve this goal. This thesis
will focus on one aspect of that research, interpersonal communication within
large-scale mutli-user virtual environments.
During one of the HITLab Consortium meeting demonstrations of VEOS
[Bric93], we observed that people in a two-user virtual environment spent more
time interacting with each other, even when they had never met before, than they
did with the rest of their environment. While there could be many reasons for this,
adding additional users to a virtual environment certainly had a profound impact.
The infrastructure required for multi-user virtual reality systems is a superset of
that required for single-user systems, so it quickly became clear that more
research had to be done into the unique aspects of multi-user systems. Tools
which enable real-time arbitration, communication, or cooperation are unique to
The Naval Postgraduate School (NPS), which is home to a premiere multi-
user virtual environment research facility led by Michael Zyda, described what
they believed to be the important virtual environment research topics to the
Computer Science and Telecommunications Board of the National Research
Council [Brut95]. Among their conclusions in that report, they claimed that "any
workable solution must address scaling up to arbitrary sizes." Indeed, as we move
away from single-user environments to multi-user environments, it is not enough
to simply add more users. The naive approaches of centralized and distributed
communication, do not scale well to an arbitrary number of users [Goss94]. Not
only do the servers or networks become overloaded with data being passed
between users, but the user's machines can become overwhelmed by a deluge of
data from the activities of users that they may not even be interested in. Various
techniques have been developed to overcome these limititations, upon which this
research will build.
Furthermore, many of the existing tools for interpersonal communication
are confined to narrow modes of communication [Suga94]. For instance, the
telephone only allows for planned verbal communication. The user must select a
person to call and usually the topic to discuss, before establishing communication.
Many of the communication modes we take for granted in face-to-face
communication, such as body language, are lost with the telephone and other
interpersonal communication tools. These narrow bands of communication in
traditional teleconferencing tools can present barriers to communicating that
make many people fearful or hostile towards the tools themselves, thereby
making communication even more difficult [Mill76][Kell85]. According to
researchers at the NTT Human Interface Laboratories in Japan, what is needed
are tools that provide seamless support for real-time communication and stored
data and planned and casual communication, as well as multi-media
communication modes [Suga94]. The prototype developed through this research
supports these goals.
The next chapter describes various Driving Applications, which could
benefit from this research and that of the more general GreenSpace project.
The Multi-User Systems chapter examines many existing multi-user
communication systems, while searching for useful techniques to apply to
The Spatial Culling Prototype chapter describes the prototype that was
developed through this research and how it relates to existing systems. The
design, implementation, and evaluation will be discussed in detail.
Finally we will present some Conclusions about what this research has told
us about interpersonal communication within GreenSpace and where future
research might lead.
There are two appendices: the GreenSpace Prototype User's Guide, which
was made available as an HTML (Hypertext Markup Language) document to
users of the system, and the GreenSpace Prototype Survey, which was an online
HTML form that users of the system were asked to fill-out.