UMBC dissertation proposal: Rendering Massive Models

UMBC PhD student Mark Bolstad is presenting his dissertation proposal on Rendering Massive Models at 9:00am Monday, October 15th in room ITE 352. If you are interested in computer graphics and what a dissertation in the area is like, you might attend. In general, thesis and dissertation proposals and defenses are open to the public and attending some is a great way to learn about an important aspect of graduate study. Here is the dissertation abstract.

Whether it is for the quest of increased visual realism in cinema, or in the processing of the latest scientific data sets, the sizes of models being rendered are becoming larger. As this trend continues, the number of elements within a scene or a single object will exceed four billion, thus requiring 64-bit data structures to reference the components of the object, and consequently, new techniques and algorithms to efficiently renderer the object.

This dissertation will present a system that will have the ability to render models of extreme complexity. The system will be designed to generate photo-realistic images of complex scenes, but is not strictly limited to those types of images. The system will support procedural shading, global illumination, and scene geometry that exceeds four billion primitives, whether it is in a single complex objects, or in a scene with a large number of simple primitives. Some of the unique contributions in this proposal are:

• Rendering models of extreme complexity: This will be the first rendering system that will be capable of rendering scenes with geometric complexity exceeding four billion elements.
• Stochastic algorithms for geometry processing: A unique stochastic algorithm is proposed that reduces the total amount of geometry processed by the renderer with little to no impact on the visual quality.
• Acceleration structures for improving the performance when rendering scenes of extreme complexity: Many of the acceleration structures used for rendering do not scale well with large models. This dissertation proposes modifications to existing acceleration structures to not only increase their efficiency with respect to large models, but to also identify the best candidate for use with a scan-line algorithm.

Dissertation committee:

• Dr. Marc Olano (chair)
• Dr. Penny Rheingans
• Dr. Hillol Kargupta
• Dr. Charles Nicholas
• Dr. Howard Motteler

Posted in Events, UMBC Research | No Comments »

UMBC Multicore Computing Center

With support and collaboration from IBM, UMBC has established the UMBC Multicore Computing Center (MC2) to investigate applications of new parallel processing technologies, including the Cell Broadband Engine (CBE) developed by IBM, Sony and Toshiba. We will integrate 12 IBM BladeCenter QS20s, each with dual 3.2-GHz CBEs into our existing Bluegrit supercomputing cluster, which includes a a 116 core PowerPC cluster. The new processors will be connected by Gigabit Ethernet and 20-Gbit/second Infiniband links.

So, what does this have to do with games? A lot.

The CBE was jointly developed by IBM, Sony and Toshiba and first used in Sony’s PS3. Game consoles can take advantage of high-performance computing to support their graphics and the CBE is what makes the PS3 special. But looking beyond speeding up 3D graphics, there are many more demands that games will make on processors — managing large artificial worlds, making computer controlled game characters more intelligent, adding speech and language processing, etc.

The UMBC MC2 will provide our game programs with a unique asset — a chance for students in the computer science track to learn about multicore and cell processors and, importantly, how to write programs to take advantage of them. Software engineers with experience with these new processor technologies are in high demand in the game industry as well as for any applications that have high computational demands.

Posted in Parallel programming, Programming, UMBC Research | No Comments »

Making machines, and computer games, *really* intelligent

People are remarkably good at “muddling through”, at not getting completely stuck, at continuing to make some progress even when things don’t go as planned. Machines, on the other hand, tend to fail in dramatic fashion when they are faced with unusual circumstances.

UMBC professor Tim Oates is working on building more robust intelligent systems through metacognition, which is the ability to think about your own thinking. Meta-cognitive systems can notice when things are not going well (as opposed to just plowing ahead with whatever they were doing, which is what most systems do today), reason about possible sources of the problem, and try various repairs.

Professor Oates and his colleagues have constructed a meta-cognitive computer player for the tank game Bolo that learns from its mistakes and adapts its knowledge, tactics, and strategies as it faces new challenges in the game and more capable human or automated opponents. For more information see the ALMECOM pages and some of the detailed papers on metacognition in Bolo.

Posted in AI, CS, UMBC Research, research | No Comments »