Doree Duncan Seligmann Multimedia Communication Research Department Bell Labs firstname.lastname@example.org
My interest in virtual environments originates in work on 3D intelligent systems [IBIS, COMET, KARMA] and multimedia systems [Rapport, MR, N-ICE]. My current research and testbed system [Archways] is directed at finding new ways to make complex multimedia systems as transparent to use as the telephone has become.
Archways automatically generates a 3D virtual environment to support remote multimedia interaction using knowledge-based graphics techniques and 3D spatialzed sound. There are many reasons why people are able to conduct natural conversations over the telephone. One is that there is little ambiguity about what is going on: you know if the connection is up, who is on the line, and even some indication about what the other person is doing. Not so with multimedia systems. Instead, users spend a substantial amount of time interrupting the natural flow of their conversations to query about the state of the system ("can you the video?", "who is typing into the shared document?", "are you still there?", etc.).
Archways creates the 3D environment to provide this kind of feedback. Some screen shots are available at: http://www.cs.columbia.edu/~doree/archways/archways.html I'll add some more later with some captions.
The idea is that people are interacting using various media (e.g. text-based chat, realtime media, multiplayer games, bullentin boards, etc.). Archways is notified how these media are being used (using an infrastructure and protocols called MR that makes this possible). The type of information ranges from each keypress event into a shared document to a new chat group being created. Driven by these events, the Archways server decides what objects to create/modify/destroy, assigning to each, a set of characteristics and a location. The Archways clients/browsers (running at each endpoint) create customized displays for each user depicting this enviornment, showing parts of the real world augmented with virtual places, the people and objects in both, enhanced with annotation objects showing both the connectivity and associations between the various particpants, devices, and services.
Each of the graphical objects include methods, constraints, and information that is used to automate their display, geometries, and rendering parameters, as well as functionality; they also change automatically to reflect changes in the virtual contexts. Thus, each object is not only a visualization, but is also a means for control as well as a mechanism to display content. For example, I have a set of little Sony video monitors on top of my computer monitor in my office. The environment depicts me in my office as well as those devices I can currently use. I can use these graphical objects to control the video. But at the same time, anyone in the environment can *enter* my office, see what devices (media) I have available, and even (permissions allowing) opt to view the actual video sent to those monitors or see how they are connected to other objects in the "real world."
We are currently adpating our system for a variety of applications. They are: multimedia conferencing and collaboration, virtual tourism, a virtual classroom, and live theatrical performace.
Here are dome issues which might be interesting topics of discussion which directly touch upon the system we have built.
1) Should virtual environments try to *hide* the real world by ignoring the real locations and environments of the users in them, or should they be environments that represent real world places along with virtual places and objects? In fact, the environment which enables us to interact remotely or electronically *is* a virtual environment. If you and I are on the phone together, we are each *in* two places: where we are, in the real world, but also, the virtual place that is defined for our conversation. Similarly, when we use devices to interact remotely, i.e. the cameras that capture our "talking heads" for a video conference; the keyboards we use to provide inut to shared electronic objects, those devices are also present with us in the virtual place. Is this duality important or useful to represent? We use suqare soap bubbles hovering over the representation of the real world to depict these virtual places.
2) Representation: what level of detail is necessary? How do we decide what events are important to inform other users? Is it useful to know, drawing, for example from your MOO, the actual time-of-day of each participant? What other information should be depicted or made available? If someone is "looking" at me, should I know about it? Each view can be augmented to show such information. For example, our system uses "cables" to show how video devices are connected, or which keyboards have control over a shared document. Similarly we use animation. For example, the keys on the keybaords are animated and letters travel through the cables to the appropriate windows. These mechanisms can ehighten the sense of "presence" in the virtual enviornment.
3) Is it necessary to coordinate at the endpoint (i.e. for each participant) of all the disparate data streams that represent whole entities? For example, my identity, my actions, my voice, my devices, etc. Is there an advatage to redundantly represent the same information in different ways? In Archways, we use 3D sound to reinforce the 3D locations of each object and person in the environment. How do we decide whether to use text, graphics, sound, etc or combinations thereof? In previous work (COMET) a media coordinator initially tagged which information would be conveyed in text, graphics or both. But, these assignnment could fail. When the user interacted with the system, indicating, for example, that he still did not "understand" (by querying for further information), the media coordinator reassigned the information (with different priorities).
4) Should a virtual environment consist of disparite, but "equivalent" views? A traditional 3D shared environment suffers many of the restrictions of a real environment: not every person's view is *ideal*. Should the objects that make up the virtual environment have the ability to modify themsleves for each particular participant, but at the same time maintain some kind of consistency (such as a fixed 3D location) overall? Thus, each person may have a completely displays of the virtual enviornment (not only perscpective, but "look and feel"). Thus, each view is customized for the particular user. For example, objects can be senstitive to the way they are being viewed. For example, we represent a shared display of application windows on a conference table. However, this display reorientates itself so that it is always upright for the particluar user (as user moves around). Similarly, textual labels reorientate themselves and change in scale to remain legible. Also, each person may be interested in different aspects of the virtual environment. Their view can be augmented with additional information without disrupting the scene for the others. This is, of course, analogous to a person accessing supplemental information about the objects in a MOO.
5) Is it useful to provide (and coordinate) multiple views of virtual environment and objects? Archways can generate different views that are each constrained by the type of information they are designed to communicate, such as, showing the other people in a virtual place or showing a person's local environemnt. These views automatically change as the environment changes (e.g. growing to include more people as they join, or more devices as they are inserted). Also, eac object is imbedded with information so a view can be created to facilitate access to its control mechanisms.
OTHER ISSUES that could be of interest:
1) What are the advatages of a 3D model?
2) What are the tools we will have to build to allow people to design and customize the virtual environment?
3) Persistence and history. One advantage of a virtual environment is that we can change time. Should we make the walls talk?
4) When the context of a situation changes, should not the environment as well? Should the notion of sessions be imposed on shared virtual environments allowing for disjoint activities in the same virtual places? For example, consider all the different types of interaction that can be simultaneously supported in one virtual classroom.