implementing gibsonian virtual environments

16
This article was downloaded by: [Ryerson University] On: 10 October 2014, At: 00:59 Publisher: Taylor & Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK Cybernetics and Systems: An International Journal Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/ loi/ucbs20 IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS DIETER SCHMALSTIEG AND MICHAEL GERVAUTZ Published online: 29 Oct 2010. To cite this article: DIETER SCHMALSTIEG AND MICHAEL GERVAUTZ (1996) IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS, Cybernetics and Systems: An International Journal, 27:6, 527-540 To link to this article: http:// dx.doi.org/10.1080/019697296126318 PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis,

Upload: dinhdieu

Post on 12-Feb-2017

213 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

This article was downloaded by: [Ryerson University]On: 10 October 2014, At: 00:59Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales RegisteredNumber: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Cybernetics andSystems: AnInternationalJournalPublication details, includinginstructions for authors andsubscription information:http://www.tandfonline.com/loi/ucbs20

IMPLEMENTINGGIBSONIAN VIRTUALENVIRONMENTSDIETER SCHMALSTIEG ANDMICHAEL GERVAUTZPublished online: 29 Oct 2010.

To cite this article: DIETER SCHMALSTIEG AND MICHAEL GERVAUTZ(1996) IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS,Cybernetics and Systems: An International Journal, 27:6, 527-540

To link to this article: http://dx.doi.org/10.1080/019697296126318

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracyof all the information (the “Content”) contained in thepublications on our platform. However, Taylor & Francis,

Page 2: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

our agents, and our licensors make no representations orwarranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinionsand views of the authors, and are not the views of orendorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independentlyverified with primary sources of information. Taylor andFrancis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, andother liabilities whatsoever or howsoever caused arisingdirectly or indirectly in connection with, in relation to orarising out of the use of the Content.

This article may be used for research, teaching, and privatestudy purposes. Any substantial or systematic reproduction,redistribution, reselling, loan, sub-licensing, systematicsupply, or distribution in any form to anyone is expresslyforbidden. Terms & Conditions of access and use can befound at http://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 3: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUALENVIR ONMENTS

D IETER SCHMALSTIEG AND MICHAEL GERVAUTZ

Institute of Computer Graphics, Vienna University ofTechnology, Vienna, Austria

We speculate on the possibilities for a re alization of the idea of a three-di-

mensional cyberspace , as sketched in the nove ls of William Gibson, using

today’s technology. V irtual re ality and global networks such as the Inte rnet

togethe r can provide the fundamentals for such an effort. From an examina-

tion of the te chnological principle s of a Gibsonian virtual environment, we

conclude what properties such an implementation must have. An implemen-

tation of a system with these properties faces a numbe r of problems in the

large and in the small; in particular, the semantics of the virtual environ-

ment and its inhabitants must be de fined. We also report on our ongoing

implementation of a prototype software system that was designed to evalu-

ate the feasibility of these ideas.

s .In his famous nove l Neurom ancer Gibson, 1984 , William Gibson

presents the idea of the cybe rspace , a sensual representation of a

global, networked database that is pre sented to the user directly by

brain stimulation. Although such technology remains fiction, recent

trends indicate that the fundamental idea of cyberspace is not only a

vision for technologists, but there may be ways to implement it. A

Address correspond ence to Diete r Schmalstieg, Institute of Computer Graphics,

V ienna University of Technology, Karlsplatz 13r 186r 2, A-1040, Vienna, Austria. E-mail:

<schmalstieg ge rvautz@ cg.tuwien.ac.at http:r r www.cg.tuwien.ac.at

Cybernetics and Systems: An International Journal, 27:527 ] 540, 1996Copyright Q 1996 Taylor & Francis

0196-9722 ¤96 $12.00 + .00 527

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 4: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ52 8

possible implementation will depend on technology from two major

areas:

V irtual reality research has provided means of conveying multisensory

information to human participants in real time , with the goal of

creating the illusion of presence in a simulated environment.

Global networks are expanding at a dramatic pace . Without exagge ra-

tion it may be stated that the way we communicate is being

revolutionized.

It is important to realize that although Gibsonian cyberspace is a

fictional concept, it se rves as a source of inspiration for the deve lop-

ment of virtual reality technology. If we take Gibson lite rally, we must

separate those concepts from his fictitious world that can be realized

s .from those that cannot yet? . O nce this is done , strategie s for imple -

s .menting the possible can be deve loped. Gelernter 1992 has shown that

a structured approach to such a ve nture is possible .

A good starting point for such a plan is to analyze the most

important network infrastructure of these days } the World Wide Web.

It has proved to be operational on a truly global scale , and its popularity

is ove rwhelming. We believe that its success can be attributed to a

unique combination of factors, among them the simplicity of the hyper-

text paradigm, platform independence via the use of standardized

protocols, and cost-e ffe ctivene ss by running on garden-variety PCs.

Howe ve r, we should realize that the WWW ingeniously circumvents

most of the hard problems of large distributed databases: The file -ori-

ented serve r archite cture makes persistence issues trivial. Generally

unrestricted access and connectionle ss client-se rver protocols large ly

avoid the need for synchronization. A re latively strict separation of

information producers and information consumers effe ctive ly makes

information flow one way only, the reby working around consistency

issue s. The insignificance of synchronization persistence , in combination

s .with the we ak ` glue ’ ’ of hype rlinks and unified resource locators URLs

that connects local information spaces, allows simple and unbound

scalability } up to a literally global size . The nature of inte raction

s .document re trieval does not impose any hard limits on responsivene ss

s} people are willing to live with bad performance ``World Wide

.Wait’ ’ .

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 5: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 529

These issue s that the WWW so cleverly avoids must be addressed

when attempting to move toward Gibsonian virtual environments.

There fore we cannot simply extend the W WW or any other current

ne twork infrastructure to three dimensions. New ideas are needed.

The goal of this pape r is to gain insight into the feasibility of a

realization of cybe rspace . We the refore first examine the technical side

of Gibson’ s fiction and analyze the state of current technology. We then

discuss important semantic questions and design conside rations for a

possible implementation of a Gibsonian virtual environment. An analy-

sis of this kind has to operate on multiple laye rs of detail. Some

s .conside rations apply to the environment as a whole macroscopic leve l ,

s .othe rs to the elements that form the environment microscopic leve l .

O n each leve l, issue s can be arrange d according to the underlying

re lation: que stions concerning semantics and design of single compo-

s . snents intrare lation and questions concerning the inte raction inte rre -

.lation between multiple components. Table 1 shows some of the

questions and also se rves as a map to the re st of this pape r. Finally, we

report on our ongoing prototype implementation of a Gibsonian virtual

environment.

C ONCEPTS OF GIBSONIAN CYBERSPAC E

In the real world, we perform interaction in three-dimensional space .

The large st amount of information is input to the human via the visual

system, and our manual capabilities are ve ry sophisticated. Conse -

quently, three -dimensional inte raction is the most natural form of

information exchange for human be ings. There fore an artificial system

that successfully mimics the way interaction takes place in real life will

Table 1. Road map to semantics of virtual environments

Que stions Macroscopic Microscop ic

Intrarelation How does the environment re late What is the structure of the

sto itse lf in the pre se nce of elements that the virtual

.multiple use rs ? What are the environment is composed of?

se mantics of shared content?

Interrelation How do multiple environments How do the e lements of the

re late to each other? How are virtual environment interact? How

environments connected? can objects live in different

virtual environments?

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 6: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ53 0

be ve ry e ffe ctive. This does not mean that the real world should be

duplicated to the most minute detail at all cost. Large amounts of

information have to be represented by abstractions to be usable, but

clearly three -dimensional abstractions are more powerful than conve n-

tional ones. The problem so far with this kind of paradigm is insupe rior

quality: Crude and slow three -dimensional applications are not ac-

cepted, but technological advances will address this problem. A work-

able implementation of cybe rspace implie s that the performance prob-

slem for typical inte ractions is gene rally solved be it by powerful

.hardware or by sophisticated software , so that smooth interaction in

near real time is available .

A global information space should be continuous. By this we mean

the user’ s view of the information should not be restricted by arbitrary

bounds in the domain of the metaphor. For hypertext systems such as

the WWW this means that links work the same way no matter whe the r

s .the target document is local or remote. For three-dimensional 3-D

virtual environments, this means that information is arranged spatially

and that the world does not ` end’ ’ arbitrarily. With continuity comes

the need for scalability: A global information space is workable only if

the architecture is easily scalable .

A continuous information space raise s the expectation that the

information is pe rsistent. Things in the real world do not vanish arbi-

trarily or change in an unpredictable manner. There fore such a behav-

ior is undesirable , because it is confusing for the user. This requires that

information entities have some form of stable state that can be accessed

repeatedly and deterministically. Such a constraint is de finite ly hard to

fulfill, especially if the state is subject to change s due to ongoing

simulation, and requires safe mechanisms similar to database transac-

tions, which conflict with real-time responsiveness. If the underlying

software architecture operated with some form of data replication,

consistency and synchronization issue s must also be solved.

Pure information acce ss is a form of interaction with software

s .agents be they inte lligent or not . However, a virtual environment also

raise s the de sire for inte raction between multiple human use rs. If the

virtual environment can become the medium for communication be -

tween users, ve ry promising applications can be realized. Beyond mul-

tiuse r interaction, the content of the virtual environment should be

dive rse. In principle , eve ryone should be able to contribute to the

senvironment with reasonable effort i.e ., populate it with data, service s,

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 7: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 531

.agents, e tc. . People will be willing to contribute only to a sufficiently

democratic information infrastructure . This require s, on the one hand,

that standards are used so that diffe rent components can talk to each

othe r and, on the other hand, that the system is open so that nove l

feature s or implementation of accepted feature s can be made opera-

tional.

C URRENT STATE OF NETWORKED INFOR MATION SPAC ES

In orde r to de fine a status quo, we will brie fly e laborate on today’s

virtual environments, which are still largely research prototypes. V irtual

s .environments such as DIVE Carlson & Hagsand, 1993 , which allow

multiple use rs to inte ract with each othe r and with autonomous agents

in a simulated 3-D space , are usually limited to a small number of users

and run on local ne tworks for performance reasons. Typically applica-

stions are some form of computer-supported cooperative work design,

.information sharing, training, playing .

Example s of systems that scale beyond local workgroup usage are

srare. The most prominent representative is the NPSNET Macedonia et

.al., 1994 family: It has been demonstrated to work for a large number

sof concurrent users, but it is limited to a specific domain military

.simulation and training , and the underlying networking protocol

severe ly limits the varie ty of possible interactions.

These re search systems provide a framework from which arbitrary

applications can be built, but they ge nerally fail to provide a solution for

data persistence . Typically, application-specific data is loaded at system

start and the internal state of the system changes as the simulation

evolves. Howeve r, dynamic reconfiguration of applications is limited,

and often the internal state of the system has to be abandoned for such

a task.

MACR OSCOPIC CONSIDERATIONS

Semantics for Shared Virtual Environments

V irtual environments, as has been indicated, should support multiple

concurrent users. This produces some subtle semantic problems. If

multiple use rs share one virtual environment, it is not necessary that the

same content is presented to eve ry user. Furthermore , if some or all of

the content is pre sented to multiple use rs, any change s one user makes

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 8: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ53 2

can or cannot affect the view others have of the manipulated environ-

ment. Semantic rule s are required that de fine the way the data are

managed by the system. In the following, we present four different

semantic models, from the simple st to the most complicated, which is

true to Gibson’ s spirit but also most difficult to implement.

1. Sharing is re lative ly simple if no modifications to the content of the

virtual environment are allowe d. This is frequently used for so-called

swalk-through applications with the aim of data visualization e.g.,

.archite ctural mode ls . The user cannot manipulate anything, so the re

is no need to communicate anything about the user to other partici-

pants in the same environment. This scheme is used by the current

s .virtual reality modeling language Pesce , 1995 extension to the

s .WWW. A scene environment? is downloaded upon request from a

WWW serve r and then displayed locally. Multiple use rs can view the

same scene concurrently, but no real sharing occurs.

2. As an extension of the pure walk-through mode l, obje cts in the

s .environment can be attributed with behavior Z yda e t al., 1993 . The

user can inte ract with a local instance of the environment and make

modifications. Howe ve r, the se change s are not propagated to othe r

users pre sent in the same environment, and change s are not recorded

anywhere .

3. The previous model can be extended if change s can be propagated

between users that share one virtual environment. This require s a

communication channel between the users } eithe r directly in a

pee r-to-peer style or mediated by a serve r. Which set of use rs are

ssharing the same virtual environment has to be defined compare

.Singh et al., 1994 . A user can join the environment by logging into a

session. Multiple instances of the same environment can exist con-

currently, with different se ts of users and different internal state s. A

new session can be started at any time , with de fault value s for all

changeable attributes. Thus the re is no persistence . Such a mode l is

useful for multiuse r games that require sharing of the gaming envi-

ronment and state but also the possibility to re set the environment

for a new game.

The sharing in such an environment makes it nece ssary to solve

s .consistency problems Blau e t al., 1992 . Conside r two users that make

concurrent updates to the ir re spective local copie s of an object and then

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 9: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 533

propagate the changes. A priority mechanism is needed that re solves

s .such conflictual situations often based on time stamps .

Scaling of such a shared environment is compromised by a high

communication cost that comes from the need to propagate changes to

all participants. This problem can be partially solved by limiting propa-

gation of message s to the subse t of participants that really need to

srece ive the message , typically de fined by ge ometric proximity see

.Fahlen et al., 1993; Macedonia e t al., 1995 and by local simulation of

s .remote activity e.g., Singhal and Cheriton, 1995 .

4. Finally, a shared multiuse r environment can be equipped with sup-

port for persistence . If a user makes a change to the environment,

the change is not only propagated to othe r users that are in the

environment at the same time but also permanently recorded. A user

entering the environment later will see all changes made before his

or he r advent. Conceptually, there are no longer multiple local

copie s of one environment but only a single distributed environment.

This makes an implementation hard for pure pee r-to-peer network

archite ctures that are othe rwise often pre fe rred for performance

reasons: If the last user leave s the environment, knowledge about the

changes is gone.

s .Howe ve r, even if a central authority serve r is available , it is not

trivial to record the current state of a ``live ’ ’ virtual environment: A log

of change s can be built, but it may be difficult to reconstruct everything

from the log. The othe r option, a ` snapshot’ ’ of the dynamic state of the

environment at regular intervals, may be infeasible because of the size

of the underlying database. Although this is obviously a difficult prob-

lem, it is mandatory to provide a solution in orde r to create a persistent

global cyberspace: It cannot be expected that all parts of the system

operate without e rrors or downtimes. Therefore only fault-tolerant

soperation with degraded performance some region of the information

. sspace is unavailable combined with e rror recove ry that allows recon-

.struction of a recent dynamic state of the environment can provide a

workable solution.

Clearly, as shared persistent environment is the hardest to imple -

ment. Howeve r, it is also the only mode l whose semantics are clearly

Gibsonian. Although the othe r mode ls obviously have use ful applica-

tions, they do not embody the role of cybe rspace as the medium, rather

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 10: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ53 4

than the way of presentation of content. It is there fore important to aim

at implementing both sharing and persistence.

C onnecting Virtual Environments

O nce it is clear how a single virtual environment works, we can try to

answer the question of what the re lation of two or more virtual

environments is. The question is where the participant goe s when

leaving one environment: e ithe r back to the real world or to anothe r

virtual environment. The way in which virtual environments are inte r-

connected have a strong influence on how the virtual environment is

pe rceived and what human use rs expect from it. There are two mode ls

for how virtual environments can coexist:

1. The dimension of the virtual environment has been de fined to be

three-dimensional space. O ne can view the environment itse lf as of

the same dimension as the contained elements. In this case , the

environment occupies a 3-D region. Two neighboring environments

can share a common borde r; in other words, they have a spatial

relation.

2. The environment has a higher dimension than the contained infor-

mation. Two environments can there fore not share a common bor-

der, they can only coexist in paralle l dimensions.

The first metaphor re sembles our pe rception of places that can be

s .reached by spatial traversal e .g., walking, flying . Although this is

appealing because of its simplicity, it requires that the information

sspace is subdivided among environments read: sites, nodes, serve rs,

.e tc. . We can expect that similar problems will arise for cybe rspace real

e state that we know from real world.

The alte rnative } multiple codimensions } does not allow a partici-

pant to use the same metaphor for the traversal of multiple environ-

ments that is used for traversing within an environment. Instead, a

dimensional shift be tween environments has to be provided, the 3-D

s .equivalent of a hyperlink portals . The creator of an environment has

complete fre edom to use all of 3-D information space. Howe ver, it is

unlike ly that it will be filled with meaningful content. We can rather

expect that only a small isolated island will be pre sent, surrounded by a

vast void, which might be disturbing for users.

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 11: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 535

The 3-D spatial arrangement appears natural and human-centric,

but that does not necessarily mean that real-world ge ometric re lations

have to be maintained at all costs. A virtual environment should be able

to represent situations that are not possible in a real environment. Here

we give some examples:

A continuous space can be equipped with virtual ` wormhole s’ ’ that

allow a use r to teleport to an arbitrary place. This does not break

the metaphor and adds additional value .

The Euclidean ge ometry that is fundamental to 3-D space can also be

extended in nonstandard ways, for example , using Moebius loops or

s .non-Euclidean ge ometry compare Groller, 1995 .È

sFrom the point of view of implementation, mode l 2 hyperlinked

.virtual environments is de finite ly easie r to handle . Eve ry environment

creator has all of 3-D space available and does not need to negotiate

with othe r creators what region to use . It would also avoid wrong

sexpectations about a corre spondence between real distances between

. s .se rvers and virtual distances in information space . However, as we

have seen the continuous 3-D space is the more natural and more

flexible metaphor. It is also the only option in the spirit of Gibsonian

cybe rspace . Implementational difficultie s should there fore be consid-

e red a challenge to be ove rcome .

MICROSCOPIC CONSID ERATIONS

Object Architecture for Virtual Environments

With the semantics of the environment itse lf having be ing defined, we

can risk a deeper look into the archite cture of the environment. From

san implementational point of view, the virtual environment is a distrib-

.uted database . The database is composed of items that go by different

names, such as obje ct, artifact, entity, actor. For this discussion, we will

adopt the gene ric term object. Note that although concre te virtual

environment implementations are mostly obje ct oriented, this is not a

requirement.

The software archite cture underlying the objects that form the

virtual environment determines the implementation of most othe r parts

of the software system and must there fore be care fully considered.

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 12: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ53 6

Passive Objects. The simple st archite cture de fine s complete ly passive

obje cts as simple data containe rs. O ften the object is simply defined by

s .its ge ometric representation e.g., V RML 1.0 . Eve ry modification to an

obje ct is be ing exe cuted by the environment in which the object lives.

Because it can respond only to an exte rnal stimulus, the object’ s

behavior is completely deterministic. The functions that can be invoked

on the obje ct are dete rmined by the system’ s set of features, which is

typically not easily extended. Although the implementation of such a

structure is re lative ly simple, the absence of a modular design makes it

hard to extend the system, which compromise s ge nerality.

Active Objects with Deterministic Behavior. O bject repre sentation

can be extended by simple built-in behavior that is composed from

standard mechanisms, such as kinematic animation and standard inte r-

s .action e.g., picking, dragging, scaling . A good example for such capa-

bilities is O penInve ntor. Note that while obje cts in this class are more

active than in the previous class, they are still de terministic; that is,

their behavior is predictable .

Active Objects with Nondeterministic Behavior. O bjects for which

behavior can be specified by the creator and that can hold inte rnal state

can be equipped with nonde terministic behavior. Typically, some form

of scripting language is used to formulate the obje ct’s behavior } if the

scripting language is powerful enough, inte re sting nondeterministic

sbehavior can be created formally, the language has to be Turing

complete; practically, it has provided enough feature s so that program-

.ming does not require too much effort .

Active Objects with External Stimulus. If the built-in capabilitie s of

the environment } eve n in combination with user programming } are

not powe rful enough to allow the construction of the desired applica-

s .tion e .g., for performance reasons , an inte rface to external applica-

tions is required. Any application can control obje cts in the environ-

sment, as long as it complies with the application inte rface Honda e t al.,

.1995 . The source of the exte rnal control should not be visible to the

user in the environment; communication with the exte rnal application

should be mediated via the object. From the environments point of

view, the user is an external application that is in control of the user’ s

s .representation in the environment or avatar .

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 13: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 537

O bviously, all of the se classe s of objects are needed to provide the

power for the environment to support any kind of application. Object-

oriented system design he lps to hide implementational diffe rences from

the users.

Object Semantics in Multiple Virtual Environments

Unfortunate ly, de fining the archite cture of obje cts in virtual environ-

ments is not sufficient to explain everything semantically important for

the objects. We still have to explain how objects re late to each othe r

s .and how objects re late to multiple , diverse environments.

The relationship between obje cts in a virtual environment is de -

fined by the obje cts’ possibilitie s to communicate . This is most easily

and flexibly achieve d by a gene ric message -passing mechanism that

allows arbitrary messages to be sent between obje cts. The progre ss that

makes the virtual environment come alive is then carried by the dis-

tributed simulation in the form of objects sending each othe r message s

and responding to them. We will not discuss technical de tails of the

message-passing mechanism that would break the scope of what is

possible here . Instead, we concentrate on the interpretation of mes-

sage s: How can obje cts understand each othe r?

How do objects understand messages? There is no unive rsal mes-

sage language. The meaning of a message is defined by convention. For

an evolving system, that means that messages de fined at a late r time will

not be understood by obje cts created at an earlier time , simply because

the meaning of the message was not defined then. Therefore , a mecha-

nism is needed that allows interpre tation of a message. An oracle

s .provides a solution Snowdon & West, 1994 , but it requires a central

authority that may be hard to set up in a distributed system. An

alternative is to arm objects with learning skills, but this is again a

difficult problem for which no gene ral solution exists.

How do objects survive in worlds with diffe rent requirements? Up

to this point, we have silently assumed that virtual environments vary

s .only in content i.e ., in the obje cts they contain , not in the ir fundamen-

tal se tup. Actually, a global ne twork of environments require s a se t of

standards that are kept by all site s, or the system will fail to work.

Howe ver, that does not mean that the re cannot be variation in the

conditions of the surrounding provided by the environment to its

inhabiting obje cts. The problem does not arise if only obje cts that are

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 14: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ53 8

designed to live in a particular type of environment are allowe d. But

s .what happens if we allow objects or use rs to migrate between worlds?

For example, a world may define gravity and othe r physical laws to

work on the contained objects. How can an obje ct that doe s not know

about physics behave reasonably in this environment? And vice ve rsa:

How can an object that requires physical laws for fundamental opera-

tions such as movement to work exist in an environment without such

propertie s? The ge neral answer is, if the obje ct is capable of adapting

autonomously to the situation, the problem can be solved. A method for

s .this has been proposed in Snowdon and West 1994 , but it is unclear

whe the r it is ge nerally workable.

We favor an alternative approach: We de fine all requirements not

on a per-environment base but rathe r by having separate groups for

every environmental influence . Membership in such a group means that

the object-environment communication regarding this matter is carried

out. The obje ct itse lf or a local authority decide s whe the r membership

in a particular group is appropriate or not. This leave s the requirement

that the obje ct is capable of ve ry simple fundamental actions, such as

move ment in three dimensions or the ability to display itse lf.

A FIR ST STEP TOWARD GIBSONIAN VIRTUALENVIR ONMENTS

We are currently implementing a distributed virtual environment infra-

structure that we hope will help us to explore the possibilitie s of a

s .Gibsonian cyberspace Schmalstieg & Gervautz, 1995 . The environ-

ment is fueled by a network of serve rs. The serve rs divide the world

samong each othe r into regions that are connected ge ometrically cur-

.rently we are using a regular grid . Inte rest in the obje cts is local, so

that the environment software has to maintain communication channels

s .only with their immediate ne ighbors in information space . This is

important for slow network connections, because it puts a natural limit

on the amount of network communication.

While the serve rs provide a space for the obje cts to live and take

care of the ongoing simulation, use rs can connect to a serve r to be

present in the environment. A network connection is set up between

environment se rver and user client, and information about the environ-

ment’ s content, and changes, is transmitted to the client as needed. In

particular, we are deve loping an optimized protocol for remote display

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 15: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS 539

sof 3-D graphics, based on V RML high-pe rformance 3-D graphics is

.ve ry important for the immersive quality of the environment . Using the

client software, the user is able to inspect the environment and also to

inte ract with it. When a user leave s the region of one server and enters

the domain of another, the network connection is transiently passed on.

Some users may want to contribute to the virtual environment. This

can be done by the creation of new objects or by the creation of new

object types. The behavior of obje cts can be specified using an obje ct-

oriented scripting language. In addition, an obje ct can be configured to

be controlled by an external application, so that arbitrary software

systems can be integrated.

C ONCLUSIONS

We have discussed the possibilities of implementing three -dimensional

virtual environments in a large networked infrastructure . Our concept is

inspired by the idea of cybe rspace in the Gibsonian sense. Such a

worldwide virtual environment will have to be three-dimensional, highly

inte ractive , spatially continuous, multiuse r-capable , shared, persistent,

scalable , and democratic. Clearly, a system with such demanding prop-

e rties cannot easily be extrapolated from an existing network paradigm

such as the WWW. A new concept is needed that provides solutions to

de sign issues and semantic que stions. For the environment as a whole , it

is important what sharing of the content between multiple use rs means

and how multiple worlds can be connected. For the inhabitants or

obje cts of the environment we have suggested how they can be struc-

tured, how they communicate, and how they can exist in multiple virtual

environments under diverse conditions. Once logically sound and tech-

nically feasible answers to these problems have been provided, an

simplementation of Gibsonian virtual environments may become vir-

.tual? reality.

R EFER ENCES

Blau, B., C. E. Hughes, J. M. Moshe ll, and C. Lisle . 1992. Networked virtual

environments. SIG G RAPH Sym posium on Interacti e 3 D G raphics, pp.

157 ] 160.

Carlsson, C., and O . Hagsand. 1993. DIVE } A platform for multi-user virtual

s .environments. Comput. G raphics 17 6 :663 ] 669.

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4

Page 16: IMPLEMENTING GIBSONIAN VIRTUAL ENVIRONMENTS

D. SCHMALSTIEG AND M. GERVAUTZ54 0

Fahlen, L., O . Brown, and C. Carlsson. 1993. A space based mode l for user

inte raction in shared synthetic environments. Proceedings of InterCHI, pp.

43 ] 48.

Gelernter, D. 1992. Mirror worlds. New York: Oxford University Press.

Gibson, W. 1984. Neurom ancer. New York: Be rkley Communications Group.

Groller, E. 1995. Nonlinear ray-tracing: V isualizing strange worlds. Visual Com -Ès .put. 11 5 :263 ] 274.

Honda, Y., K. Matsuda, J. Rekimoto, and R. Lea. 1995. V irtual socie ty:

Extending the WWW to support a multiuser interactive shared 3D environ-

ment. Proceedings, VRML’95 Sym posium , San Diego.

Macedonia, M. R., M. J. Zyda, D. R. Pratt, P. T. Barham and S. Zeswitz. 1994.

NPSNET: A network software architecture for large-scale virtual environ-

s .ment. Presence 3 4 :265 ] 287.

Macedonia, M., M. Zyda, D. Pratt, D. Brutzman, and P. Barham. 1995. Exploit-

ing reality with multicast groups: A ne twork architecture for large -scale

virtual environments. Proceedings VRAIS’95.

Pesce , M. 1995. VRML } browsing & building cyberspace. Indianapolis: New

Riders.

Schmalstieg, D., and M. Gervautz. 1995. Towards a virtual environment for

inte ractive world building. Proc. GI Workshop on Modeling } Virtual

Worlds } Distributed Graphics, Bonn, pp. 49 ] 56.

Singh, G., L. Se rra, W. Png, and H. Ng. 1994. BrickNet: A software toolkit for

s .network-based virtual worlds. Presence 3 1 :19 ] 34.

Singhal, S., and D. Cheriton. 1995. Exploiting position history for efficient

s .remote rendering in ne tworked virtual reality. Presence 4 2 :169 ] 194.

Snowdon, D., and A. West. 1994. AVIARY: Design issues for future large -scale

s .virtual environments. Presence 3 4 :288 ] 308.

Zyda, M., D. Pratt, J. Falby, C. Lombardo, and K. Kellehe r. 1993. The software

required for the computer generation of virtual environments. Presence

s .2 2 :131 ] 140.

Dow

nloa

ded

by [

Rye

rson

Uni

vers

ity]

at 0

0:59

10

Oct

ober

201

4