remote collaboration using augmented virtuality for · 2004. 11. 30. · kohler, et al., 2003. an...

H. Regenbrecht*[email protected]

T. Lum*[email protected]

P. Kohler*[email protected]

C. Ott†[email protected]

M. Wagner‡[email protected]

W. Wilke*[email protected]

E. Mueller*[email protected]

*DaimlerChrysler AGGermany

†igroup.orgGermany

‡shared-reality.comGermany

Presence, Vol. 13, No. 3, June 2004, 338–354

© 2004 by the Massachusetts Institute of Technology

Using Augmented Virtuality forRemote Collaboration

Abstract

This paper describes the concept, prototypical implementation, and usability evalua-tion of the Augmented Virtuality (AV)-based videoconferencing (VC) system cAR/PE!.We present a solution that allows three participants at different locations to com-municate over a network in an environment simulating a traditional face-to-facemeeting. Integrated into the AV environment are live video streams of the partici-pants spatially arranged around a virtual table, a large virtual presentation screen for2D display and application sharing, and 3D geometry (models) within the room andon top of the table. We describe the general concept and application scenario aswell as the actual hardware setup, the implementation, and the use of the system inits current state. Results of two usability studies with 87 subjects are presented thatshow the general usability of our approach as well as good overall satisfaction. Partsof the work described here were presented as a poster at the second InternationalSymposium on Mixed and Augmented Reality (Regenbrecht, Ott, Wagner, Lum,Kohler, et al., 2003. An Augmented Virtuality Approach to 3D Videoconferencing.Poster at 2nd Int. Symp. on Mixed and Aug. Reality, Tokyo.).

1 Introduction

Favored by the boom of the Internet and by an increasing number ofworldwide merger and acquisition activities, global communication has notonly increased in volume, but has eventually created a technically advanced andactually working communication network. A computer-supported collabora-tive work (CSCW) environment is on its way to an international scale. CSCWuses three main technologies in this context: (1) application sharing, (2) datadistribution and sharing, and (3) videoconferencing (VC). Ideally, remote en-gineers can, for example, share a common view onto their Computer AidedDesign (CAD) system, while manipulating and discussing the same set of dataand communicating via a VC system. But because of the organizational andtechnological immaturity of today’s systems, current forms of global commu-nication utilize telephone lines, email, and maybe computer screen sharinginstead of the CSCW technologies mentioned above.

The goal of our present work is a solution which improves communicationcapabilities by integrating Augmented Virtuality (AV), Virtual Reality (VR),data distribution and sharing technology, and state-of-the-art network tech-nology into one CSCW system.

The kernel of our system is an AV frame. All the components needed are inte-grated into this kernel. Because AV as the underlying technology has the potentialto use the best of two worlds, VR and “real” (physical) reality, we believe that it

338 PRESENCE: VOLUME 13, NUMBER 3

could serve as an ideal environment for natural communi-cation in a synthetic data world and network.

According to Milgram, Takemura, Utsumi, and Kishino(1994) AV lies on the more virtual segment of the reality-virtuality continuum. AV enhances virtual worlds with realworld components or technical aspects. In our case, realparticipants in a CSCW setup are brought into a virtualscenery with spatially aligned video and audio representa-tions. We will show that this approach can overcome thelimitations to natural communication present in today’svideoconferencing systems. By combining AV with theother data channels needed for collaborative engineeringtasks, real CSCW can be achieved.

After the discussion of related activities in this field wedescribe our general system concept, as well as the cur-rent state of implementation, the cAR/PE! system pro-totype. Based on a first prototype system, results ofinitial usability evaluations are presented. We drawconclusions about technical and psychological aspects ofthe system and describe future activities derived fromthese conclusions at the end of this paper.

2 Related Work

The main impulse for the development of thecAR/PE! system comes from our own previous work inthis field. We have developed and tested an AugmentedReality (AR) system called MagicMeeting (Regenbrecht,Wagner, & Baratoff, 2002) which allows a number of four(but not limited to four) participants to discuss virtual 3Dgeometry on a physical table as if the virtual object were areal one. The participants wear head-mounted displays(HMD) with video see-through capabilities, and sitaround a table on which the virtual objects are placed.Each user sees the real world in a video-mediated way aug-mented with spatially aligned virtual objects. Integratedinto MagicMeeting are 2D and 3D data, specialized in-teraction techniques, and distribution techniques basedon Inter- and Intranet technology.

The main advantages of this system are found in thevery natural way of communicating and interacting, andin the integration of all data needed for a successfulmeeting scenario. The main disadvantages are the use of

cumbersome hardware (in particular, the HMDs limitface-to-face communication), the robustness and qualityof the tracking system, and the inherent need to havethe meeting at one physical location.

Similar projects dealing with collaboration and AR(e.g., Billinghurst, Kato, Kiyokawa, Belcher, & Pou-pyrev, 2002; Schmalstieg, Fuhrmann, Szalavari, & Ger-vautz, 1996) are also investigating colocated AR insteadof remote collaboration.

Collaborative Virtual Reality systems, such as thosedescribed in Benford, Brown, Reynard, and Greenhalgh(1996) and Frécon, Smith, Steed, Stenius, and Stahl(2001), and multiuser computer games connect VR orgame engines mainly over the Internet and distributethe users’ interactions over the network. The represen-tations of the users are either symbolic (e.g., a cube ortext stands for a person) or avatars (dynamic 3D geome-try) such as Active Worlds (www.activeworlds.com).The whole common display and interaction space is apurely virtual one. These environments can be highlyrealistic and vivid, but are more often synthetic or ab-stract and cannot really be a substitute or alternative fora communication space known from the real world.

CSCW systems vary widely in the functions they pro-vide. The typical classification in a time-and-space ma-trix (see Johansen, 1988) distinguishes between supportof local or distributed (space) working contexts and be-tween synchronous or asynchronous (time) work situa-tions. CSCW systems like IBM Lotus Notes or Mi-crosoft Outlook utilize the desktop metaphor fromtoday’s graphical user interfaces and bring them into adistributed, collaborative context. The main shortcom-ing of these systems is that they lack synchronous, dis-tributed communication possibilities while offering theabove-mentioned functions in a user friendly environ-ment. Often, simultaneous telephone conferencing isneeded to substitute for this lack of synchronicity.

VC systems, on the other hand, have improved thequality of communication over the last years substan-tially and have partially integrated CSCW content (e.g.,chat and whiteboard) with video and audio. ProfessionalVC stations require higher bandwidth or the availabilityof alternative telecommunication channels, but providehigh-quality A/V (audio/video) channels. In most

Regenbrecht et al. 339

http://www.activeworlds.com

cases, application or 2D data sharing is offered as anoption. These (expensive) systems are typically used inspecial rooms or facilities in large companies. Use ofthese systems is increasing only slightly. In general, eachparticipant has to go to the (physical) VC facility nearhis or her working environment to attend the meeting.The communication channel is provided by a TV-set-sized A/V device. Research is underway to improve thespatial impression of these devices by adding a 3D com-ponent to them.

Aimed one significant step further is the Office of theFuture (OotF) project at UNC (Raskar et al., 1998). Re-mote participants and environments will be integrated intoone mixed environment simulating the spatial configura-tion of reality from each participant’s point of view.

A similar approach to integrating participants in a spa-tial simulation of reality is the project VIRTUE (Kauff& Schreer, 2002). Here, a combination of advancedcomputer vision and computer graphics methods is usedto provide a comprehensive system for face-to-face vid-eoconferencing of three users. This system gives theimpression of three participants sitting at the same table.

The main disadvantages of OotF and VIRTUE lie inthe very complex hardware and software used. Never-theless, these approaches are the main references for thecAR/PE! system presented here.

The commercially available system Muse-Lite (www.musecorp.com) offers a collaboration scenario whereapplication sharing and presentation capabilities are pro-vided as well as chat and A/V conferencing. The collab-orative environment is set up in a 3D manner similar tothe approach presented in this paper. The main differ-ences are: Muse collaboration is designed for confer-ence-like situations (one presenter, many people in the“audience”); it is Web-server-based, so there is a needto exchange data (including confidential data) via thecompany’s/provider’s site, and it does not allowattention-directed A/V communication.

3 Concept

In this section we describe the general concept ofthe system independently from its current implementa-

tion. One application scenario will serve as a use caseto compare concept and implementation. The require-ments listed are either given by the future customersand users of the system or are derived from the state-of-the-art in industry and research. Finally, the desiredproperties of the target system, future research, and dis-cussed concept alternatives are presented.

3.1 Application Scenario

The cAR/PE! system is embedded in a globalworking environment. Members of a project team situ-ated in different locations need a communication plat-form to coordinate the projects they are working on.The cAR/PE! application scenario is that of a plannedmeeting, not of an ad hoc or spontaneous gathering orsharing of information. Since the field of engineering isvery diverse, the scenario described in the following israther abstract. cAR/PE! provides a communicationplatform for the exchange of a broad range of differenttypes of information.

Two or more engineers are conducting a meeting.During this meeting they can discuss changes or im-provements to 3D objects derived from CAD data, suchas parts, machines, or even entire industrial plants. Sincethe objects are virtual, the engineers do not have tomeet in one location, such as in the actual industrialplant. Also, either the objects can be discussed as awhole, compared to other objects, or the focus of dis-cussion can be set to details. Like in a real meeting, ad-ditional documents are needed for a comprehensive dis-cussion: a slide presentation for the agenda, goals,details and so on; an application sharing capability todiscuss details in the CAD system normally used; andtext and spreadsheet document display or a possibilityfor documentation. Each meeting has a moderator wholeads the session and has certain rights and obligations,depending on the purpose of the meeting, such as as-signing tasks, choosing or manipulating objects to bedisplayed, and so on.

The engineers sit in their regular office, with access totheir documents and computer systems. The cAR/PE!system runs on the same computer the engineers use for


their everyday work. They have been invited to the vir-tual meeting by the moderator of the meeting throughtheir normal IT tools ahead of time. Like for a normalmeeting, information about the meeting can be distrib-uted through mail or shared information spaces in ad-vance to all participants to prepare for the meeting.

3.2 Requirements

The system should be integrated into the existinginfrastructure. This can be divided into the followingthree main aspects. (a) Data chain integration: The databrought into the meeting and the results of the meetingshould fit into the existing technological environment.Types of data are PowerPoint presentations, text docu-ments (e.g., MS Word), spreadsheet documents (e.g.,MS Excel), pictures in electronic form, and 3D geome-try. This geometry is derived from CAD systems (here,Dassault’s CATIA) and should be capable of being inte-grated seamlessly into the environment. (b) Process in-tegration: The invitation and the data selection processshould utilize the same tools as already used by the en-gineers during their daily work. Especially, access to theproduct data management (PDM) system should be aseasy as possible. (c) IT infrastructure integration: thesystem should be set up on top of standard network andtelecommunication (TC) infrastructure already presentor to be provided within the near future. The security ofthe users and the data transmitted has to be guaranteed.

In general the system should be suitable for meetingswith two to ten participants. It should be optimized fora three-user and a two-user setup, which covers the ma-jority (56%) of the use cases for videoconferencing(Aaby, 2003).

Costs of the system should be minimal (much lessthan EUR 5000) because one goal of this project iswidespread use and distribution. A solution is preferredthat uses an ordinary PC already present at the engi-neer’s desk, extended by the cAR/PE! software andlow-cost off-the-shelf hardware.

The interface to the system should be as natural aspossible compared to real world meetings. Communica-tion between the participants and interaction with thedocuments, 3D data, and the environment should be

supported by an advanced interface. All data needed forthe meeting should be integrated interactively into onesingle environment. The whole system runs in synchro-nous, interactive real time.

One of the most crucial technical requirements ismoderate bandwidth consumption. Because this projectis intended to be productive in the midterm future (1 to3 years from now), we have to estimate the bandwidthavailable when the system becomes operative. Today wecan count on ADSL resources (128/256 KBit/s, 768/1536 KBit/s) or equivalent bandwidth on ATM or sim-ilar network infrastructure. In the future we assume thatwe can rely on resources available today in wireless localarea networks (11 to 22 MBit/s). Therefore our systemshould work today in any wireless LAN.

Additional criteria are the following: connectivity andcompatibility with existing systems (especially Internet,TC), accessibility to conferencing and/or communica-tion services, user authentication, accessibility to chatand whiteboard, robustness and accountability, use ofand/or contribution to standards, bandwidth efficiency.

3.3 User’s Workplace (in Real World)

There are an almost unlimited number of possibil-ities of how a participant could use the cAR/PE! systemin the real world, ranging from cell phones or personaldigital assistants to multiprojection systems. We havedeveloped a subset of reasonable conceptual variantswhich fit our application scenario and context.

An obvious option when approaching from a Magic-Meeting point of view is the use of HMDs in combina-tion with a headset at a desk. Here the user sits at his orher desk and wears an HMD with a headset consistingof a pair of headphones and a microphone. A camerafacing the user is placed on the table (or alternatively ontop of the monitor on the table). This camera capturesthe video images of the user and is responsible for track-ing the user’s head. The main advantage of this setup isthe possibility for the user to freely move his or her viewinto the augmented environment. The main disadvan-tage is, like in MagicMeeting, the cumbersome equip-ment and the limited field of view and resolution.

A more promising alternative can be seen as an almost


ideal solution regarding price and comfort. The user sitsin front of a standard monitor that is equipped with oneor more cameras, a set of loudspeakers, and a micro-phone (echo-canceling). There is no need for the userto wear any equipment. A special device or metaphor isneeded for the self-movement within the (virtual) en-vironment. The most crucial tasks when implement-ing this alternative are (1) creating a sufficiently goodspatial-audio impression (loudspeakers) and (2) trackingthe user’s head, which is difficult because there is noequipment and therefore no artificial feature to be rec-ognized. In a first setup the loudspeakers can be re-placed by a headset, and artificial features (like markers)can be attached to the user’s head for tracking. We havedeveloped seven principal workplace variants but focuson the one mentioned here for concreteness.

3.4 Meeting Room (Virtual World)

We have limited the number of possible partici-pants in a meeting to 10 persons. Although it is imagin-able that such a system could be used even in confer-ence-sized setups (several hundreds of persons), wedecided to focus on smaller groups. The main reason forthis is the assurance of scalability while maintaining sta-bility and implementation effort. Most meetings thatapply videoconferencing in our industrial context todayare attended by three persons, followed by two-personconfigurations. But even in setups between two and 10persons, there are a lot of principal differences in thedesign of the meeting room. In real meetings two per-sons usually sit opposite each other, while a meetingwith 10 persons happens around a large table (see Fig-ure 1).

Independent of the number of persons involved, themeeting room should have the following main proper-ties: (1) integration of a virtual projection screen, (2)integration of a space where 3D models can be dis-cussed, and (3) information about the persons partici-pating. The overall impression of the room should coin-cide with the expectations the users have of a real-worldmeeting room. Hence, the main task for the design is toreduce the almost unlimited number of possibilities(e.g., no need for a floor and ceiling, no need for chairsetc., no need even for Euclidean geometry) down to areasonable and convincing result.

3.5 Interaction with the System

The kind and quality of the interaction of the userwith the system is the main design task. We have to dis-tinguish between three phases of interaction: (1) priorto the session, (2) during the session, and (3) after thesession. The majority of interaction takes place in phase2, while phase 1 includes the whole process of invitationand session preparation (especially data selection anddistribution). Phase 3 prepares the protocols and otherresults for later processing or consideration. In this pa-per we focus on phase 2.

The most basic and essential form of interaction is theself-movement of the user within the environment. Theuser should be able to see all parts of the environmentthat are important for self-awareness and for the task tobe performed. In cAR/PE! the main elements to beexplored are (a) the 3D model space, (b) the presenta-tion area for slide display and application sharing, and(c) the other participants. Sometimes all elements (a) to(c) are important to the user at the same time. Beforethe actual session starts, the user can freely explore theenvironment to get an impression of the overall struc-ture of the room and of the subjective relation of his orher “body” to the scene. During the meeting it is suffi-cient to have an on-seat interaction. This means that atleast three components should be controllable by theuser: rotation of the head to the left and right, rotationof the head up and down, and—for overview/focus rea-sons—a change of the field of view, if technically possi-

Figure 1. View of two- and 10-person meeting variants.


ble. With these basic options the user is able to partici-pate in a meeting scenario such as described above.

For the non-HMD-based workplace alternative,which is our current focus, we decided to use a desktop-based 6DOF controller device with constraints on therotation of the (virtual) head and for changing the fieldof view. In addition to that, we have developed a tech-nique called WorldWindow that serves as a substitutefor the restrictions of movement given by the screen-based interface. This technique is similar to “fishtankVR,” but uses computer vision technology. The user’s(real) head is tracked by the same camera used for videostream capturing, and the view through the display(monitor or projection) into the virtual environment isset according to the detected head position (see Fig-ure 2).

With this technology we can use off-the-shelf equip-ment as an interface to virtual and mixed environments.In a later step, this technique can be used with an addi-tional camera on the back of the monitor or with a see-through display for an AR view, too.

The selection and modification of 3D models can beimplemented by the use of discrete actions such as but-ton clicks. The models of choice should be visible, ei-ther in symbolic or realistic appearance, within theroom. After one or more models are selected, they haveto be brought into common view for all participants.The extent of possible manipulations of the objects de-pends on the task, ranging from simple movement(translation and rotation) via scaling or detail zoomingup to assembly or modeling actions. We decided to in-troduce the different manipulation possibilities in severalsteps according to their complexity: (1) simple transfor-

mations, (2) simple assembly/disassembly, and (3)packaging investigations with collision and contact sim-ulation. For the very common task of variant discussion,the implementation of step (1) is sufficient.

Because almost every meeting is accompanied by slidepresentations, this form has to be integrated into such asystem. We believe that a seamless integration of 2Ddata into the 3D environment helps to ease the transi-tion from 2D interfaces used today to 3D interfaces asintroduced with cAR/PE!. Not only slide shows arepart of today’s meeting processes, but all kinds of datausually displayed in 2D, especially office documents(texts, spreadsheets, pictures) and CAD data. That iswhy the system should allow specialized forms of displayand interaction and why it should integrate some func-tionality for application sharing.

For all these reasons we decided to have a presenta-tion wall within the environment which can serve as aplace for 2D data display and application sharing.

For discussion or model manipulation purposes somekind of pointing interaction is needed. A very commonapproach is to use a ray-cast metaphor, which is multi-functional by nature. It can be used to manipulate ob-jects in 3D or simply to point at 2D or 3D objects. Forour application scenario, pointing at the presentationwall or at the object to be discussed is the main require-ment. Both direct and indirect manipulation of the rayand the object is useful in our scenario.

For successful communication, clear, undisturbed,and spatial audio is necessary. The spatialization of thesound should indicate exactly where a sound originates.If a participant next to the user is talking, he or sheshould be able to realize this even without video infor-mation. Mapping of audio and video information in thevirtual space is essential for orientation in the virtualspace. Because the participants are mainly placed on ahorizontal plane in space (e.g., around a table), we canuse this constraint to simplify the implementation of the3D audio component. The 3D audio interface shouldwork with headphones as well as with 2.0 to 7.1 audiohardware.

To support the prior-to-session and after-sessionphases we decided against developing our own solutionsin this field. Instead, we relied on (legacy) systems al-

Figure 2. Schema of WorldWindow technique.


ready integrated into the existing workflow, and developappropriate interfaces to these tools and systems. Themain interfaces are: (a) to organizer, mail, and calendarsystems, (b) to existing product-data management sys-tems, (c) to CAD systems already used, and (d) to doc-umentation and protocol tools.

3.6 Visualization of Participants

For the impression of sitting in a real meeting,graphical representations of the participants are inher-ently needed. In most collaborative VR-based systemsthe users are either represented as abstract geometry oras avatars. Abstract geometry (like a cone) seldom givesthe impression that another person is acting on the re-mote site; avatars are not that convincing unless theyhave a very high level of detail and behave kinematicallycorrectly. In VC applications, the participants are dis-played as videos in a window on the desktop. This pro-vides a good impression of the participant but unfortu-nately not of the spatial relationship within the meeting.In our setup we place video-planes in a 3D space. Theplanes display live video in 2D but follow the move-ments of the remote users in space. If a remote userturns his/her head or applies the WorldWindow tech-nique, the connected video-plane moves accordingly inthe virtual space. This provides each user with informa-tion about what his or her partners are looking at. Evenwhen the user moves freely within the room, we canfollow the movements.

The video and audio streams should be synchronizedas well as possible, although fluent audio with low la-tency is more important than high quality of the videostream. Investigations on the correlation between videoand audio quality have shown that a higher objectivequality in audio can significantly improve the subjectivequality of video and vice versa (Kohlrausch, 2003). Inour design we focused on providing good audio qualityaccompanied by sufficient video quality.

A status wall is integrated in the meeting room wherenames and videos of the participants are displayed. Thisenables a user to check his or her own visibility withinthe meeting and therefore to the other participants (likea mirror).

4 Implementation

Taking the general concept as a guideline we haveimplemented a prototype system with the following ad-ditional requirements. (a) The system should becomeproductive as soon as possible. We believe a very earlyprototype can serve for much more precise discussionand decisions about future requirements, directions, andoptimizations. (b) The overall architecture and ap-proach should serve as the basis for further develop-ments. (c) Even the first prototype should consider theindustrial context and infrastructure to be introduced inthe future. (d) The system should definitely be low costto achieve broad acceptance and to have a chance ofintegration.

4.1 System Overview and Architecture

With our implementation of the system we con-nect multiple cAR/PE! stations to an Ethernet network.The network is divided into four channels: (1) the videochannel transmits the video data of the participants; (2)the audio channel transmits the voice streams; (3) theinteraction channel is the carrier for all kind of interac-tions by the users, for the initiation and synchronizationof the stations, and for the transmission of the applica-tion sharing data; (4) the 3D data channel is used toinitially distribute the geometry of the cAR/PE! roomitself, for the initial distribution of the 3D models to bediscussed in the meeting, and for changes or ad hoc dis-tributions of 3D models. To avoid network overloadimmediately prior to or during the meeting, all alreadyknown data (especially 3D models) are distributed be-forehand (off-line; at the time of writing of this paperthis is still done manually). These data are stored locallyat each station. The invitation process is responsible forconsistency. We decided on this semireplicated data ap-proach for bandwidth, performance, and latency rea-sons. For a discussion of pros and cons of centralizedversus replicated architectures see Schmalstieg, Reit-mayr, and Hesina (2003).

The stations are connected point-to-point. We de-cided on this option for security and infrastructure rea-sons.


4.2 Station Hardware

Each cAR/PE! station is equipped with a standardPC (Pentium IV, 2.4 GHz, 512 MB RAM, nVidia Ge-Force4, bt878 card), or alternatively, a high-end note-book computer; an analog color camera (720 � 576,25Hz) equipped with a wide-angle lens; a 17� TFTmonitor; a keyboard and mouse; a SpaceMouse/Magel-lan (Logitech/3Dconnexion) 6DOF controller; and astereo headset with mono microphone (see Figure 3).As an option, loudspeakers (stereo or surround) can beused. All components are off-the-shelf. For the trackingof the user’s head (WorldWindow) an additional head-set with markers was needed in the first prototype. Thecurrent second prototype system uses computer vision-based face-tracking methods instead of markers, whichis even more robust.

4.3 Station Software

The application presented in this paper is based ona Virtual Reality system architecture implemented in ourin-house VR system “DBView” (Sauer, 2001). DBViewprovides all the essential components for VR applica-tions. The core of DBView is a scene graph that man-ages all objects of the current scene. The scene graph isbased on Open Inventor (Wernecke, 1994), a wide-

spread scene graph API developed by Silicon GraphicsInc. Open Inventor is available for all major operatingsystem platforms. Within the kernel, the scene graph isnot accessed directly, but is encapsulated by an objectmanagement system. Integrated into the kernel are alsoa module manager to handle dynamically loaded appli-cation modules, a configuration management, and amessaging module. The latter connects—and allowscommunication among—modules and external applica-tions of a shared environment. The cAR/PE! softwaresystem runs on each station and is implemented as anapplication module. All application-specific functionalityis provided by this module.

The current implementation is based on MS Win-dows 2000 mainly because of the better integration intoour enterprise context.

The module interface allows the VR kernel to be ex-tended by adding special functionality or task-specificapplication logic. We use this module interface to ex-tend the VR system towards our 3D videoconferencingsystem, for example, video capture/streaming/visualiza-tion, audio streaming and (3D) display, WorldWindowtechnique, and cAR/PE! application control.

4.4 Interaction

For our current usability scenario (see below) weneed certain control of the system. We have two types ofusers: a moderator and two other participants of the meet-ing. All users are able to change their viewing directionwhile seated at the table. This is done by using the Space-Mouse, which is constrained to allow rotation around twoaxes and a change of the field of view angle between 1°and 160°. In addition to this movement the WorldWin-dow technique (implemented by a special module) can beapplied. Both movements are transmitted to the other par-ticipants and control the orientation of the video plane(see Figure 4). With this approach, the user can see wherethe other participants are looking.

Each user can bring him or herself back to the origi-nal home position and can toggle WorldWindow track-ing on and off by pressing SpaceMouse buttons.

Also implemented for each user are two differentpointing interfaces. The first one utilizes the Space-

Figure 3. cAR/PE! hardware setup.


Mouse to send a virtual ray from the pointer’s ownvideo plane (avatar) into the environment. The secondone tracks a marker physically mounted to a desktopmouse (similar to the approach described in Regen-brecht, Baratoff, and Wagner, 2001, which also sends aray). With both approaches the users are able to point atthe presentation wall, at the 3D models, or at eachother as an additional form of interaction. The authorshave tested both methods and have realized that theseare too cumbersome to use. For this reason we did notprovide it for the usability study. In addition to thecommon interaction forms for all users the moderatorcan invoke additional functions via the keyboard: load-ing 3D model alternatives, SpaceMouse control (transla-tion and y rotation) of the model on the table, changeof slides (pictures) on the presentation wall, and takinga snapshot of the moderator’s current view to be dis-played on the presentation wall. All interaction results ofthe moderator are distributed by our messaging inter-face and are therefore visible to all participants.

The application control also allows configuration ofthe (premodeled) cAR/PE! room for use as a virtualvideoconferencing facility: pose of each video plane (in-cluding own one), loading of different models andslides, and access restrictions for non-moderators.

This basic but distributed functionality is sufficient toperform a simple meeting within the cAR/PE! environ-ment.

In a second prototype system, which is currently fin-ished, we have extended the functionality of the firstprototype. The main differences are as follows. (1) In-stead of utilizing the buttons of keyboard and Space-Mouse we have included icon buttons to control theenvironment. These are placed within the virtual roomand are spatially attached to the user’s view. The stan-dard computer mouse is used for interaction. (2) Usingthe mouse to point at the presentation wall and at the3D model is also possible. The pointer is representedwithin the space as a 3D object with different colors foreach participant (to be able to distinguish them in thecollaborative setup). (3) The SpaceMouse is also used tocontrol a clipping plane through the model (in threeswitchable directions). (4) Application sharing on thepresentation wall is possible (with low performance).

4.5 cAR/PE! Room

The AV environment described in the conceptchapter has been implemented in its first phase as asetup for three participants. The room was modeled

Figure 4. Visible viewpoint control.


with a standard 3D modeling program. The main ele-ments of the room are textured with images derivedfrom a radiosity program using “real” lamps for thelight simulation. This leads to a more realistic overallimpression of a meeting room.

The room consists of the main elements shown inFigure 5. The presentation wall is intended for applica-tion sharing and slide presentation display. For bettervisibility, the wall is slightly tilted.

In the center of the table at which the virtual partici-pants sit, alternative 3D models can be placed (see Fig-ure 6). All alternatives are shown on the left-hand wall.The information wall shows live video streams of eachparticipant as well as status information.

This room (model geometry) serves as a first test bedand can be substituted by alternatives by simply loadingother (Inventor/VRML) files.

5 Empirical Work

Two different experiments were conducted fromApril, 2003 to August, 2003 using our first prototypesystem as described above. The first experiment aimedat general aspects of usability as well as at first insightsinto the concept of this CSCW tool. The second experi-

ment was concerned with the comparison of cAR/PE!with a real meeting situation and with a standard CSCWtool with a graphical user interface, Lotus Sametime(www.ibm.com). Experiences and results from an openformat questionnaire that led to several improvementsof the system will be discussed.

5.1 Experiment 1—Overview

The goal of this usability case study was to collectuser experiences with a first prototype. The focus of in-terest was to see whether the cAR/PE! system offers auseful way of supporting a meeting with participants indifferent locations. Since the function of this system isto enable communication with other participants therewere two main dimensions of interest: (1) how commu-nication between the participants evolved and was expe-rienced; and (2) the direct interaction with—and useof—the actual system.

(1) Communication involves the exchange of infor-mation between an emitter and a recipient. In a typicalmeeting situation, information is sent directly by anemitter, for example by speaking or through nonverbalcommunication such as facial expressions, or the infor-mation is distributed in the form of objects such as awritten report or a presentation board.

We wanted to know if communication would arise atall, and if so, how participants experienced the verbalexchange of information as well as the visually perceivedinformation. A basic necessity for communication be-tween persons is to direct someone’s attention towardsa person or an object. This includes a person’s percep-tion of the partner’s direction of attention. Since thesystem offers the possibility to present 3D objects, wewere naturally interested in the perception and use of3D objects. Besides the primary goal of communication,exchanging information, we were also interested inwhether the system would enable the experience ofpresence and of social presence, which are believed tobe connected to the effectiveness and efficiency of com-munication (Nowak, 2001; Lombard, 2001).

(2) The direct interaction with the system focuses onthe usability of the system. This involves movement inthe virtual space (the change of the personal direction of

Figure 5. Schematic overview of cAR/PE! room.


http://www.ibm.com

view) and the manipulation of objects. Aspects of inter-est are orientation in the virtual space, feedback infor-mation about movement, and usability of input devicesand of other information displayed by the interface.

The data were collected through a questionnaire andoffer a first look at strengths and weaknesses of thesystem.

5.1.1 Method. Participants. The participants inthis experiment were 27 adult volunteers (19 men and 8women, mean age 29.2 years, range 22–42 years). For3 participants German was not the mother language,but all had lived in Germany for at least 18 months. Theparticipants were recruited from our R&D departmentand were placed in groups of three. All but one partici-pant had a college education or were still studying. Ran-domization of groups was aimed at but not totally ac-complished, due to organizational problems andcancellation of appearance at short notice. The modera-tor of each group was assigned randomly. The partici-pants were not paid, and anonymity of data was pro-vided.

System. In this case study, the cAR/PE! system wasused by three participants and the task was to decide onthe most aesthetic out of five car models. The decisionhad to be supported by all three members. One partici-pant was assigned to be the moderator for the meeting.The virtual car models were placed on one side of thevirtual meeting room and could be moved by the mod-erator to the middle of the table. The moderator alsohad the possibility to put screenshots on the presenta-tion board. One side of the room contained a presenta-

tion board with a hypothetical agenda for the meetingand some information about the use of the system andthe task. Participants were given 20 minutes to solve theproblem.

Setting. Each person was sitting in a different room infront of a screen. A 3D-Mouse (6DOF controller Space-Mouse) as a navigation/manipulation device, a stereoheadset with mono microphone, and a headband withtracking markers attached for the head tracking, wereprovided. The participants’ movement in the virtualspace was restricted to looking up and down, rotatingtheir view left and right 360° and changing their field ofview. This could be done with the 3D-Mouse or bytracked head movements. Keys on the 3D-Mouse wereassigned a fixed viewpoint and turned the head-trackingon or off. Only the moderator could change the carmodels with the keyboard and move them with the 3Dmouse. The exchange of slides on the presentation walland the screenshot taking were also activated with thekeyboard.

5.1.2 Materials. Instructions about how to usethe system and the task description (one version for themoderator and one for normal participants) were pro-vided in paper form and were available to the partici-pants during the entire session. After the session theparticipants were asked to fill out a questionnaire. Itcontained the following parts, derived from the abovementioned questions of interest.

A. Personal data such as age, education, and former ex-perience with AR/VR systems or VC (5 items).

Figure 6. Different inside views.


B. General questions about usability (based on Lewis,1995), for example, “It was easy to use the system” (4items).

C. Questions about the audio quality (5 items).D. Questions about the video quality, design of the in-

terface (9 items).E. Questions about navigation and interaction of the

system (15 items).F. Questions about the impressions of the virtual meet-

ing room (5 items).G. Questions about the task-related functions (6 items).H. Questions about presence (taken from a study by

Schubert, Friedmann, and Regenbrecht, 2001) (11items).

I. Questions about social presence (taken from a studyby Nowak, 2001) (16 items).

J. Follow-up questions with an open answering format(4 items).

The questionnaire contained a total of 80 items, withdifferent answering possibilities, ranging from open for-mat answers to yes/no answers to 7-point rating-scaleanswers with anchors like “agree”/“fully disagree.” Allitems use the 7-point rating-scale answer form with 1 �agree and 7 � totally disagree (in total 47 items), ex-cept open format questions, yes/no questions, anditems about presence, which were adopted from Schu-bert, Friedmann, and Regenbrecht (2001) and had a7-point Likert-scale ranging from �3 to 3 with anchorslike “not at all � �3” and “totally � 3.”

Because this case study’s goal was not to create andevaluate a questionnaire for this type of system, we usedseveral items that have been tested in previous studiesbefore and added some for our specific purpose. So asnot to change the reliability and/or validity of theadopted items we left the original answering categorieswith the disadvantage of a more heterogeneous layoutand possible confusion of the participant about thechanging answering possibilities.

5.1.3 Procedure. A single-case design was used,each group having the same task and the same condi-tions. Analysis of data was descriptive, extracting ten-

dencies of strengths and weaknesses of the system aswell as building hypotheses for future experiments.

The participants were seated at their designated sta-tions and were told to read the written instructions. Thesystem was already running so that participants wereable to directly test what they were reading and get usedto the system. After approximately 10 minutes the realsession was started. During the entire session an investi-gator was present for each participant if help wasneeded. After agreeing upon a solution the participantswere given the questionnaire and a pencil. Total timefor the experiment was approximately 60 minutes.

5.1.4 Results and Discussion. Data of questionswith the 7-point rating-scale were transformed into a3-point scale to avoid bias of central tendency and forbetter interpretation since we were interested in generaltendencies of opinions/attitudes about the system. Thedata were analyzed by calculating the mean and thestandard deviation for each item (except yes/no ques-tions, for which only percentages were used) and per-centages, which is used in the following because of bet-ter explanation.

Overall, the cAR/PE! system was rated easy to use by88.9%, and user satisfaction was good (66.7%). Inter-pretation of the data in reference to (1), whether thesystem enables the exchange of information, is morediverse. Exchanging information verbally was ratedgood by 55.5% with only 7.4% rating it poor. This issatisfactory, as it is known that audio delay is a crucialfactor when using a system like cAR/PE! for enablingcommunication, but the rating indicates that the qualityof the audio transmission has to stay at this high level orshould even improve. Participants’ prior experience withvideoconferencing systems or Augmented or VirtualReality systems had no effect on the perceived audioquality.

The virtually displayed car models, the presentationboard, and the video image of the other participantswere perceived without any problems by the majority(59.3%, 88.9%, and 85.2%), although the models lackedreality in appearance (51.8%). For this task cAR/PE!offered all needed tools to complete the task. Althoughthe quality of the video was perceived as good (59.3%),


participants did not look at their partners very often(only 37% did), and did not have the impression thatthey were looked at by their partners (only 14.7%). Thismay be due to the problem that participants were look-ing at the screen and not directly at the camera, andthus eye contact could not be made. Additionally, thetask demanded looking at the discussed car models,whereas information between participants was ex-changed verbally. There were also problems knowingwhat the other participants were looking at or wherethey were looking, which indicates problems at estimat-ing somebody’s focus of attention. Since the partici-pants’ image was always “looking” in the same directionindependently of the position of the avatar, the positionof the avatar was the only indicator for where a personwas looking or what he or she was looking at. This maybe improved by, for example, a pointing device.

A different question is to what extent the system gen-erated a sense of being within the virtual environment(presence) or a sense of being with another person inthat environment (social presence, Biocca, Harms, &Burgoon, 2001). Participants did not have a very strongsense of being or acting in the virtual world, althoughthe virtual environment did attract their attention(55.5%), and the layout of the room was perceived asbeing clearly arranged (70.3%). In general the data sug-gests that there was little sense of presence. This couldbe due to various reasons and must be tested with amore detailed experiment in the future. Concerning theevolution of social presence, the data shows that thepartners were perceived as real (92.6%) and were willingto listen to one another (74%), which confirms the as-sertions about the good video transmission quality andaudio quality. On the other hand, the session was notassociated with a real meeting (70.4%), and it did notgenerate the feeling of being in the same room as theother partners (66.7%). Once again this may be due tothe type of task elaborated and the restricted time. Wesuggest a longer working session and a task requiringmore use of functions and therefore interaction/move-ment with the system to test presence and social pres-ence.

The direct interaction with the system (2) was gener-ally rated positively (77.7%). This also holds for more

specific aspects of the interaction, such as understandingthe functions of the keys on the 3D mouse or keyboard(89.9%), time needed to learn to control the system(rated sufficient by 80.6%), or orientation in space.Other aspects of navigation suggest further improve-ment; for example, switching between different fields ofview was complicated for 49.5%. This may be due to thenovelty of the 3D mouse, the head-tracking, or to thedelay of the system to user inputs. The head-trackingwas not accepted widely, which may be due to slow re-action time of the system, uncomfortable headbands, ornovelty of the system and the movements in the systemimplicated by one’s own movement. This is indicated bythe difference in the use of the head-tracking betweenparticipants with experience with Virtual and Aug-mented Reality systems and participants with no experi-ence (used 17.06% as opposed to 6.2% of the time).

The design and layout of the room could be im-proved, although 51.8% said they were able to workeffectively in this configuration. Data showing that par-ticipants were having problems changing their field ofview suggest the search for a more effective configura-tion or a different solution, such as a different field ofview.

5.2 Experiment 2—Overview

A second experiment was conducted comparingcAR/PE! with either a real meeting situation or LotusSametime, a conventional CSCW tool. The experimentwas conducted with engineers from DaimlerChryslerproduction plants in Germany who are potential usersof the cAR/PE! system. Although the focus of this ex-periment was the evolution of mental models in virtualenvironments and will not be discussed in this paper,results from an open format questionnaire and experi-ences from this realistic setting and sample will be dis-cussed.

For better comparability, the setting of the secondexperiment was not based on the task of the discussionof 3D models within the cAR/PE! environment, be-cause there are no operative meetings nowadays discuss-ing virtual models in 3D. Meeting situations in produc-


tive sessions deal with 3D models on the basis of 2Drepresentations displayed on a screen.

5.2.1 Method. The goal was to create a highlyrealistic setting for engineers in reference to their tasks.Two different tasks were created from this background,one being a cost optimization task, the other a designtask where participants had to consider all changes whenconverting a standard side door into a frameless sidedoor for a convertible. Participants solved one of thetasks with cAR/PE! and the other either in a real meet-ing, sitting around a table, or with Lotus Sametime.Time to solve the tasks was about 10 minutes with anadditional 5 minutes of introduction on how to use thesystem. After performing both tasks, participants weregiven an open format questionnaire with demographicquestions and questions concerning data security, im-provements of the systems, the experimental design, andwhether they would like to use the cAR/PE! system fortheir work. Time for the entire experiment totaled ap-proximately 70 minutes.

Participants. 60 participants were recruited fromDaimlerChrysler production plants. Motivation for par-ticipating in the experiment was interest in cAR/PE!since all participants worked in distributed projectteams. The average age was 37 years (range from 24 to63 years). 94% were male. Average employment atDaimlerChrsler was approximately 11 years. All partici-pants were from the production planning or develop-mental departments or working directly together withthese departments.

Setting. The cAR/PE! system was similar to that ex-plained under 5.1.1. Changes were: (a) the head track-ing was absent and (b) a pointing device for the presen-tation board for each participant was installed. Pointingdevices were used through the mouse. Three additionalviewpoints were installed, one for each of the remainingtwo participants and one for the presentation board,using the keys on the 3D space mouse. The presenta-tion board displayed photos of a side door and its cover-ing, which was used as a help for solving the designtask. The real meeting situation was realized by placingthe participants around a table, and presenting a realside door and its covering next to the table. The meet-

ing condition with Lotus Sametime was realized on thesame personal computers used for cAR/PE!, showingthe same photos of the side door and the covering as inthe cAR/PE! session.

5.2.2 Results and Discussion. Although resultspresented here are only exploratory, the sample theywere derived from and the setting under which partici-pants tested the system were highly realistic and there-fore offer valuable insights into engineering work prac-tices and their requirements regarding a system likecAR/PE!. Since the questionnaire had an open formatfor answers—except the demographic questions andthe one asking whether participants would like to usecAR/PE!—answers that were mentioned frequently arepresented.

Overall handling of the system was rated good, al-though some participants had problems using the 3Dspace mouse (mentioned by six participants). 92% of theparticipants would use a system like cAR/PE! for theiractual work. In total, 30 participants mentioned the au-dio quality and rated it to be improved, with only oneperson rating it good. Experience with bad audio qual-ity led to technical improvement of the audio qualityafter conducting half of the experiment, resulting inonly 8 (from the mentioned 30) complaints about thequality after the change. As pointed out in the first ex-periment, audio quality is a crucial factor and deserves ahigh priority in further developments. The poor audioquality may also be a reason why more structure con-cerning who is speaking was mentioned by 6 persons,suggesting, for example, visual cues around the avatarfor the person speaking. We suggest that further im-provement of audio quality will solve some of the diffi-culties observed because direct transmission of languagecan offer the structure missing presently. 18 persons intotal mentioned the configuration around the table as aproblem, suggesting, for example, possibilities tochange the configuration to predefined positions. This isespecially important when discussing, for example, slideson the presentation board, because observations showedthat participants hardly changed their positions, onlyhaving the presentation board in their field of view. Thecurrent seating configuration makes it impossible to


look at the presentation board and at the other groupmembers at the same time. This can be solved by eitherchanging the seating configuration or by additionallyplacing the participants’ live video around the presenta-tion board. Another disadvantage concerning the seat-ing configuration was identified when participants talkabout a 3D model. Since the model is placed in themiddle of the table, each person has his or her own viewof the model, making it difficult to explain his or herview to the other members since they cannot see theside of the model the person is talking about. Thiscould be solved by seating the persons in a straight line,all having almost the same view on the model.

In addition to comments about the configuration ofthe virtual meeting room and its usability, participantsmentioned a lot of functions they would need in theirdaily work. Some had been planned since beginning thecAR/PE! project anyway, for example integration ofdocument sharing or a white board for drawing, whileothers were new. What seems to be of major importancein design meetings is the meeting protocol, since it listsfor all parties or members of the meeting the tasks to becompleted by the next meeting. Eleven participants fa-vored an additional screen for the protocol. Since spon-taneous integration of one’s documents was also men-tioned by 9 participants, we suggest a virtual desktop beintegrated in the meeting room for each participant,displaying his or her documents.

Overall testing in the above mentioned setting provedcAR/PE! to be a good basis for further development,also confirming very good system stability. To accountfor the diversity of tasks and settings used only in theengineering field we propose to develop a basic systemfor all and to add task/setting-specific features andfunctions.

6 Summary and Future Work

We have presented an AV prototype system for3D videoconferencing with three remote participants.We were able to prove our concept and implementationwith a focus set on the evaluation of the general setup

and on the usability of the system. While emphasizingthe communication aspects in such a face-to-face meet-ing incorporating 2D and 3D documents, the criticalmain parameters have been tested. The potential forfuture productive use could clearly be shown: all partici-pants in the usability study were able to communicatewith almost no restrictions.

Because of the general success of our approach we aregoing to develop this prototype towards a system foroperative use within our enterprise context. Later on wewill provide a system which can be used by externalpartners (such as suppliers), too. Our future focus is setto the following aspects:

On the basis of our current prototype system we aregoing to substantially improve the system performanceand functionality. Because of the limitations and thecost of bandwidth resources we have to optimize theamount of data to be transmitted. While the data trans-mitted prior to a session are less of a problem, the inter-action, video, and audio channel have to be optimized.This is especially so because the system will be intro-duced into the enterprise context within the next year.Since this is earlier than planned, we have to cope withlower available bandwidth than initially estimated. Inthe near future we are heading for an ADSL-like basedinfrastructure (0.8 MBit/s downstream and 0.1 MBit/supstream); therefore, we will limit the interaction mes-sages sent through the network to a minimum, com-press the video data with the implementation describedabove, and add an alternative POTS (plain old tele-phone system) interface to the system. With this ap-proach we separate the audio channel from the network(while maintaining the 3D audio effect) and can signifi-cantly reduce network load.

The interaction functionality with the system is goingto be extended according to the need of our users. In afirst step we will primarily focus on the simulation of atraditional slide-based-presentation meeting scenario,where the 3D data serve as an add-on. In the secondstep, interaction with the 3D data out of the closelycoupled PDM systems can be manipulated. We are alsoconsidering (depending on the customers’ needs) theintegration of Augmented Reality content into themeeting environment.


Much conceptual work and prototype implementa-tion has to be done regarding the integration of the sys-tem into the real processes of our enterprise. On thetechnical infrastructure level we have to hook up thesystem to existing (secure) networks, to VC system pro-tocols, and to the telephone system architecture alreadypresent. Interoperability has to be maintained with re-spect to different hardware platforms, operating sys-tems, and applications. On the workflow level we haveto integrate the system into collaboration tools (such asLotus Notes and Sametime), different in-house projectmanagement tools, documentation systems, and work-flow and change management systems. A very crucialaspect is the integration of the output of a cAR/PE!meeting into the systems (e.g., protocols and modifica-tions of the data within the session). The data integra-tion level has to consider problems such as the conver-sion of VR data, the close coupling to (distributed andheterogeneous) product-data management systems, andthe integration of version management. From both theworkflow and data integration level, future requirementswill be specified regarding (1) authorization and au-thentication (e.g., determining who is allowed to initi-ate a session/partake in a session) and (2) access con-straints of the PDM systems, especially with externalpartners (e.g., to deal with situations where the sessionmanager selects objects for discussion in a session, butat least one of the participants is not allowed to access,or see, one or more of these objects).

All future work will be accompanied by usability stud-ies targeting more detailed questions and hypotheses.

Acknowledgments

We would like to thank Birgit and Martin Kohlhaas, ThoreSchmidt-Tjarksen, Michael Feltes, Hartmut Schulze, JoergHauber, Denis Brenner, Bertrand Muguet, Michael Mueck,Alexander Dick, Kai Boeswetter, Ralph Schoenfelder, andVolker Amann for their contributions to the cAR/PE! systemand for their comments on earlier versions of the paper. Alsowe would like to thank Gregory Baratoff for his special sup-port and all participants in our usability study as well as theanonymous reviewers.

References

Aaby, C. (2003). Marker analysis of video conference systems.Proceedings of ICOB Workshop, Berlin, Germany. [CD-ROM]

Benford, S. D., Brown, C. C., Reynard, G. T., & Greenhalgh,C. M (1996). Shared spaces: Transportation artificiality andspatiality. Proceedings of ACM Conference on Computer Sup-ported Cooperative Work (CSCW’96) (pp. 77–86). Boston:ACM Press.

Billinghurst, M., Kato, H., Kiyokawa, K., Belcher, D.,Poupyrev, I. (2002). Experiments with face to face collabo-rative AR interfaces. Virtual Reality Journal, 4(2).

Biocca, F., Harms, C., Burgoon, J. K. (2001). Criteria for atheory and measure of social presence. Retrieved February6, 2003, from http://www.mindlab.msu.edu/Biocca/papers/2002_scope_conditions_social_presence.pdf

Frécon, E., Smith, G., Steed, A., Stenius, M., Stahl, O.(2001). An overview of the COVEN platform. Presence:Teleoperators and Virtual Environments, 10(1), 109–127.

Johansen, R. (1988). Groupware—Computer support for busi-ness teams. New York: Free Press.

Kauff, P., & Schreer, O. (2002). An immersive 3D video-conferencing system using shared virtual team user environ-ments. ACM Collaborative Environments, CVE 2002.

Kohlrausch, A. (2003). What can we learn from perceptionresearch about audio-visual interactions in immersive broad-cast and communication systems? Proceedings of the ICOBWorkshop, Berlin, Germany. [CD-ROM]

Lewis, J. R. (1995). IBM computer usability satisfaction ques-tionnaires: Psychometric evaluation and instructions for use.International Journal of Human-Computer Interaction,7(1), 57–78.

Lombard, M. (2001). Resources for the study of presence:Presence explication. Retrieved January 24, 2003, fromhttp://nimbus.temple.edu/�mlombard/Presence/explicat.htm

Milgram, P., Takemura, H., Utsumi, A., & Kishino, F.(1994). Augmented reality: A class of displays on the real-ity-virtuality continuum. Proceedings of Telemanipulator andTelepresence Technologies, SPIE Vol. 2351, 282–292.

Nowak, K. (2001). Defining and differentiating copresence,social presence and presence as transportation. Retrieved Jan-uary 24, 2003, from http://nimbus.ocis.temple.edu/�mlombard/P2001/papers.htm


http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=0722-7833()2351L.282[aid=6050515]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=0722-7833()2351L.282[aid=6050515]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1044-7318()7L.57[aid=2656757]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1044-7318()7L.57[aid=2656757]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()10L.109[aid=4971253]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()10L.109[aid=4971253]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1359-4338()4L.2003[aid=6050516]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1359-4338()4L.2003[aid=6050516]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1359-4338()4L.2003[aid=6050516]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1359-4338()4L.2003[aid=6050516]http://www.mindlab.msu.edu/Biocca/papers/2002_scope_conditions_social_presence.pdfhttp://www.mindlab.msu.edu/Biocca/papers/2002_scope_conditions_social_presence.pdfhttp://nimbus.temple.edu/mlombard/Presence/explicat.htmhttp://nimbus.temple.edu/mlombard/Presence/explicat.htmhttp://nimbus.temple.edu/mlombard/Presence/explicat.htmhttp://nimbus.ocis.temple.edu/mlombard/P2001/papers.htmhttp://nimbus.ocis.temple.edu/mlombard/P2001/papers.htm

Raskar R., Welch, G., Cutts, M., Lake, A., Stesin, L., &Fuchs, H. (1998). The office of the future: A unified ap-proach to image-based modeling and spatially immersivedisplays. ACM SIGGRAPH 1998.

Regenbrecht, H., Baratoff, G., & Wagner, M. (2001). A tan-gible AR desktop environment. Computers & Graphics,25(5), 755–763.

Regenbrecht, H., Wagner, M., & Baratoff, G. (2002). MagicMeet-ing—A collaborative tangible augmented reality system. VirtualReality—Systems, Development and Applications, 6(3), 151–166.

Regenbrecht, H., Ott, C., Wagner, M., Lum, T., Kohler, P.,Wilke, W., et al. (2003, October). An augmented virtualityapproach to 3D videoconferencing. Poster session presentedat the Second International Symposium on Mixed and Aug-mented Reality, Tokyo.

Sauer, J. (2001). Virtual reality in der produktentwicklung

[Virtual reality in product development]. VDI-Berichte[Reports] Nr. 1614.

Schmalstieg, D., Fuhrmann, A., Szalavari, Z., Gervautz, M.(1996). Studierstube—An environment for collaboration inaugmented reality. Proceedings of CVE’96 Workshop

Schmalstieg, D., Reitmayr, G., & Hesina, G. (2003). Distrib-uted applications for collaborative three-dimensional work-spaces. Presence: Teleoperators and Virtual Environments,11(1), 52–67.

Schubert, T., Friedmann, F., & Regenbrecht, H. (2001).The experience of presence: Factor analytic insights. Pres-ence: Teleoperators and Virtual Environments, 10(3),266–281.

Wernecke, J. (1994). The Inventor Mentor. ProgrammingObject-Oriented 3D Graphics with Open Inventor [Release 2,1994] Boston: Addison-Wesley.

http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()10L.266[aid=5253398]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()10L.266[aid=5253398]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()10L.266[aid=5253398]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()11L.52[aid=6050517]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=1054-7460()11L.52[aid=6050517]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=0097-8493()25L.755[aid=6050518]http://www.ingentaselect.com/rpsv/cgi-bin/linker?ext=a&reqidx=0097-8493()25L.755[aid=6050518]

remote collaboration using augmented virtuality for · 2004. 11. 30. · kohler, et al., 2003. an...

Documents