In the CAVE, high-resolution stereoscopic images are generated by a multiprocessor Silicon Graphics workstation or multiple workstations and rear-projected onto several walls and front-projected onto the floor, using mirrors to fold the optics. Participants wear stereo glasses to view the images in 3D. A head-tracking device is affixed to the glasses worn by the navigator, and a hand-tracking device with input controls--called the "wand"--is used to interface with the environment.
"The reason it works very well is because, unlike a head- mounted display, the whole world does not have to turn when you turn your head," DeFanti explains. "The CAVE is like the real world. The perspective changes subtly, but that is a much less radical phenomenon than changing the world."
The capability of sharing findings with others, however, is the most significant benefit of the interface, he says. "It preserves the teacher-student/salesman-customer relationship."
The CAVE may be used both to explore precomputed datasets and to interactively steer supercomputer simulations.
The first CAVE was built at EVL and publicly displayed at SIGGRAPH 92 as part of Showcase. There were subsequent demonstrations at Supercomputing '92 and '93, with a 64-node CM-5 available onsite for simulations at the latter.
NCSA's CAVE measures 10 x 10 x 9 feet and is housed in a specially modified room at the Beckman Institute for Advanced Science and Technology. Echoing the "Holodeck" on Star Trek: The Next Generation, a grid crisscrosses the ceiling and walls near the CAVE. Images appear on two or three walls and the floor; sound emanates from four speakers in the corners, providing ambience and data-driven feedback.
The images are computed by a four-processor SGI Onyx system with three Reality Engines. The Onyx connects via HIPPI to NCSA's supercomputers, of which the CM-5 and SGI Challenge are being emphasized for virtual environments.
Sound is produced by the Ensoniq ASR-10 synthesizer and the Yamaha 990 signal processor. The demand for virtual sound cannot be met by existing commercial products; however, NCSA is developing interactive sound synthesis techniques using an SGI Indigo as a dedicated server, says Robin Bargar, who is coordinating audio development. Sound is digitally placed in three dimensions utilizing an MX-16 system similar to those used by Lucasfilm. For more details see sonification in virtual reality (VR).
Other CAVEs have been built at the ARPA Enterprise (Arlington, VA) and the Argonne National Laboratory (Argonne, IL, near Chicago).
Back to A virtual environments toolset
Forward to EVENT