Although much progress has been made at the rendering stage, architectural modeling, particularly at the early stages of design, represents one of the most difficult challenges for the interactive graphics software industry. Currently available architectural design software is difficult to use at the preliminary design phase. Most offerings are derived from two-dimensional drafting systems requiring precise input and supporting only harsh direct lighting and exact geometric display. What is needed is a "back-of-the-envelope" environment, with the ability to sketch, doodle, and erase, to proceed in an iterative, non-linear fashion, and to perform these operations in a full three-dimensional domain. Since design is an iterative process and it is frequently necessary to work conceptually in two dimensions as well as three, it is important to be able to migrate freely from one domain to another. It is just as important to be able to extract plans, sections, and elevations from 3D sketches and models, as it is the other way around.

We must also be able to shift freely from the three-dimensional sketches to approximate surface models composed of primitive elements to solid models with precise definitions. We must be able to progressively move from the simple line drawing and direct lighting routines used in standard graphics hardware to photorealistic representations including shading, shadows, textures, and global illumination, as well as provide all of the perceptual cues necessary to disambiguate complex models. Finally, we have to do this in real-time, providing dynamic walkthroughs, since movement is really necessary to evaluate space. No software is currently available which provides the thread of continuity between these widely varying levels of representations.

Our specific objectives include providing answers to the following questions.

• How do we create meaningful functional links between symbolic, graphical, and realistic visual abstractions and representations?
• Early, creative-stage abstractions are tentative, and lack the specificity of later, more formal representations, yet the generality of symbols, sketches and rough diagrams enable a fluidity in thinking that is often hampered by the detail of formal plans and specifications. How can we create a spectrum of abstractions and representations from the discrete forms that currently exist, and how can we support free movement back and forth along this spectrum?

For more information, please visit the web pages for Architecture and Fine Arts at the Program of Computer Graphics.