CCV COLLOQUIUM
MacMillan, Room 117
Wednesday, November 12, 2003
3:00PM
Visualization: New Dimensions, New Domains, Old Questions
Andries van Dam
Vice President for Research
Thomas J. Watson, Jr., University Professor of Technology and Education
and Professor of Computer Science
Visualization leverages the massively parallel computer that is the human visual system, enabling users to see and understand patterns, trends, and anomalies in data. Visualization is an imporatnt, but often underappreciated (and underutilized) aspect of computational simulation. This talk fouses on the applications of visualization in immersive virtual reality, particularly our Cave. I will show examples from Brown research in classical scientific areas and in the humanities and social sciences.
The exceptional performance requirements of virtual reality have led us to collaboration with IBM to use their SGE ( Scaleable Graphics Engine), a network-attached frame buffer driven by a 48-node PC cluster with NVidia graphics cards. I will briefly describe the project, which provides parallelization of rendering to applications that can exploit it, by analogy to parallelism in scientific computation.
Finally, I will speculate about some not-too-distant future developments. First: the future of collaborative environments -- next-generation collaboratories. These will be based on integrated 2D and 3D immersive environments that incorporate the "ubiquitous computing" vision of Mark Weiser and all of its devices, both large and small, with augmented and virtual reality spaces. Achieving this will require seamless, transparent interoperation among these different technologies and form factors, as well as betwenn human experimenters and representations of their remote collaborators.
The second development I eagerly await is one of the grand challenges for simulation science: to build a "virtual human" from the subcellular level on up. My view is that, in addition to the vast amount of research still required to understand how all the subsystens at all the different levels operate and interact; we'll need a new kind of computational data structure. I call these data structures "clip models" -- families of interoperable simulation-based models for each subsystem at all levels that have both geometric structure and behavior.