FORT COLLINS -- Colorado State University computer scientists are changing the future of computing by creating an entirely new class of adaptable computer systems capable of processing large files and complex data sets at blazing speeds -- up to 600 times faster than the quickest Pentium chip. The unique high-speed, low-power computers already are being used to improve defense satellite recognition systems, enhance military night vision software and improve the speed of transmitting images over the Internet.
The key to this new technology, developed by Colorado State computer science researchers Wim Bohm, Ross Beveridge and Bruce Draper, is a powerful chip called a Field Programmable Gate Array, or FPGA. The FPGAs greatly increase a computer's speed by reconfiguring the hardware circuits to directly match the needs of a particular software program. The systems essentially create programmable hardware by allowing users to repeatedly download new programs directly onto a computer's circuits.
"We are developing a very different paradigm for computing that has the potential to dramatically change the industry within the next 20 years," Beveridge said.
These cutting-edge computer systems have piqued the interest of the U.S. military. For the past four years, the Defense Advanced Research Projects Agency funded the development of this innovative technology. Two projects have been completed for the military that are now being adapted by government labs for defense purposes.
The first is a ground-based target recognition system with night vision that automatically detects objects such as enemy tanks and displays results on a screen inside a U.S. military vehicle, describing both what the object is and its position. The detection process works about 600 times faster than the most powerful Pentium chip.
"If you are in a military operation in the middle of the night trying to determine if an enemy tank is in the vicinity, you don't want the answer in an hour or even a minute, you want it right now," said Draper. "A system with an 800 MHz Pentium chip displays the image in 65 seconds; our system takes less than one second."
The second project is an automatic target recognition pre-screener program that allows Air Force satellites to scan images of large land masses and quickly recognize vehicles and facilities for further military observation. The new system recognizes objects about 40 times faster than previous systems. Deployed on a satellite, it can also be upgraded from Earth.
A commercial application currently being supported by Honeywell focuses on improving wavelet technology to compress large images. Wavelets are a primary tool used to condense JPEG images for transfer over the Internet. With this new technology, researchers are making this data compression 10 to 100 times faster and cheaper than the chips currently being used.
The Colorado State research team is now looking into other potential commercial applications, including incorporating the technology into PCs, resulting in desktop machines that would be virtual supercomputers. Other possible applications involve using this technology to improve electronics such as digital televisions.
"There are currently several digital television formats that are consistently being changed and upgraded," said Bohm. "By using FPGAs, any updates or changes in formats could be downloaded directly to the TV. Consumers could easily obtain the latest technology without having to buy new electronics."
The Colorado State computer science team created this new technology by building a compiler that maps algorithms directly onto computer circuits that can then be downloaded onto an FPGA. The ability to download a program directly into hardware circuits, coupled with advances in the digital chips, results in FPGA circuits that can handle many simultaneous processes hundreds of times faster than on a normal computer.
For complete information about this technology and its applications, including in-depth overviews, diagrams and technical information, visit the Web at http://www.cs.colostate.edu/cameron/index.html.
Cite This Page: