Philip Levis, CS; Vladlen Koltun, CS
If you haven’t yet been invited to send a digital representation of yourself to a business meeting or a family reunion in a “virtual world,” it may be because these richly graphical online environments are hamstrung by technical and economic limitations that constrain their reach. So far, virtual worlds have been built by just a few companies, using proprietary technologies that cannot grow in the same free-flowing way as the traditional Web.
October 31 2008. If you haven't yet been invited to send a digital representation of yourself to a business meeting or a family reunion in a "virtual world," it may be because these richly graphical online environments are hamstrung by technical and economic limitations that constrain their reach. So far, virtual worlds have been built by just a few companies, using proprietary technologies that cannot grow in the same free-flowing way as the traditional Web. As a result, while millions of enthusiasts see them as providing unprecedented richness to online interaction, they're stuck in niche status.
"Three-dimensional virtual worlds might be the next computer application platform," says Philip Levis, an assistant professor of computer science and electrical engineering. "So how do we design technology, systems, and networking abstractions to support such an application?"
To develop technologies that will make virtual worlds more open, scalable and ripe for innovation, Levis has teamed up with computer science Assistant Professor Vladlen Koltun and computer science and electrical engineering Professor Pat Hanrahan, the Canon Professor in the School of Engineering. Last month the team, along with colleagues Michael Freedman and Tom Funkhouser of Princeton University, received two grants totaling $800,000 from the National Science Foundation to fund their research.
In the early going, two of the concepts they'll employ to open up the world of virtual worlds are ones that would make the ancient Greeks proud: geometry and democracy. In this case, infusing the Internet with the ability to handle geometry-oriented data would allow it to efficiently model the inherently spatial nature of virtual worlds-a lot of them are like vastly social online games with big areas in which to wander. Democracy, meanwhile, refers to inventing user-friendly software that will make virtual world creation easier for everyday users without special expertise or deep corporate pockets.
Space and sensibility
At first blush, a virtual world like a city or even an online big box discount store might seem overwhelming from a data perspective compared to even a large Web site. After all, if these worlds are to immerse visitors in graphical and social realism, then every car and pedestrian on the street, or every item on the shelf will have to be visually rendered and ready for interaction. But virtual worlds offer a way to prioritize all this data that comes from their mimicry of the real world; proximity determines what's most important.
In other words, if a virtual world and the underlying network understand where a user's digital representation, or avatar, is located within the world, they can concentrate their resources on the data pertaining to whatever is close by. In a gladiator world, the lion that is five feet from a player's avatar deserves more of the system's attention than the lion that is 100 yards away. Finding a way to apply this filter would make the networks running virtual worlds operate much more efficiently and better able to handle millions of users.
"Suddenly we have this metaphor that is very clear," Levis says. "Only things that are close to you can interact with you. We can use these real-world metaphors to make systems-level decisions."
So to make this insight concrete, the team is working on developing a network protocol based on three-dimensional geometry. The protocol would provide anyone building a virtual world with a ready-made standard for handling proximity data smoothly.
The idea is not completely new, but perhaps the best precedent was too far ahead of its time to be implemented, Koltun says. Back in the 1980s, the U.S. military sought to develop technology for virtual battlefields and experimented with something like a geometric protocol. But at the time, networks were just too slow for the idea to be realized. Similarly, previous researchers have created ways for databases to store data based on geographic coordinates.
The free (virtual) world
Creating geometric infrastructure technologies that will allow virtual worlds to run efficiently and scalably would be an important accomplishment in its own right, but Koltun, Levis and Hanrahan also want to make these technologies widely available so that anyone can exploit them. That's also the driving philosophy behind their work on making software that will ease the creation of content to populate these virtual worlds.
Democratizing the creation and implementation of virtual worlds will have several benefits, Koltun says. First of all, participants in virtual worlds will not be beholden to just a few companies, as they are now. Lowering the barrier to entering the virtual world business will inspire a greater diversity of virtual worlds, just as open infrastructure standards will make them more compatible. Secondly, having many virtual worlds will allow more people to participate in them. The capacity of the virtual world space-much like the capacity of the Web-won't depend on the limited budget or workforce of an oligarchy, but instead on the breadth of the user community.
"Anybody should be allowed to create a virtual world," Koltun says. "I don't need to ask anybody to set up a Web page. It should be similar with virtual worlds and right now it's not."
Koltun's solution has been to remove the expertise needed to create the many objects that give a virtual world its richness. Last year, for example, he produced a program called Dryad that allows users to create trees without having to draw or sculpt, or even imagine them. Instead Dryad includes a multidimensional database of tree attributes (e.g. trunk thickness, bark texture, leaf shape) and can dynamically generate an infinity of trees with different combinations of those attributes. To "create" a tree, users merely float above a forest, heading in the direction of the stand of trees that meet their fancy. Then they can select the tree that seems closest to what they want and use menus to tune its parameters to the desired outcome. Each user's idea of a perfect tree can then be exported in a standard 3D graphics format.
Dryad is completely intuitive and requires no artistic ability, but under the covers a formidable amount of mathematics handles the transitions and combinations among the scores of dimensions in the attributes database. With the new grant, Koltun plans to move on to creating software for generating people, animals, and buildings. Surprisingly, the buildings look like they'll be the hardest to implement, he says. The reason is that buildings share very few common attributes and so the intuitive transitions among the trees in Dryad won't be nearly as easy to produce. Virtually all trees have a trunk that splits off into branches that then terminate in leaves. But the similarities between a tin Quonset hut and the New York Public Library pretty much end at the existence of a door.
Still, Koltun, Levis and Hanrahan are undaunted because they see an opportunity to make the Internet a thriving universe of virtual worlds. "We're designing an infrastructure for a new communication medium that could potentially be used on a planetary scale," Koltun says.
