Helping the blind get around could involve less help from canines and more help from computers in the future.

That’s just one of the many technological advances that investigators at the University’s Center for Future Architectures Research could make possible. The $28-million University-led research center, which opened Jan. 17, involves researchers from 15 different universities.

“One medical application we are developing is a device that can perceive images for a blind person,” said Associate Engineering Prof. Valeria Bertacco, a lead researcher at C-FAR.

Bertacco described a glasses-like device with cameras that identify objects in a person’s surroundings.

“Each identified item is then coded in a sound sequence and the sound … (is) played for the user through the earpieces,” she said. “The result is that a blind person can identify the objects in the surrounding space based on the sounds that he or she ‘hears.’ ”

C-FAR aims to create the “technology of 2030” by improving devices’ use of semiconductor components, called transistors. Bertacco, who leads reliable systems research at C-FAR, called these tiny components the “building block” of any electronic device.

However, C-FAR researcher Krste Asanović, a professor of computer science at the University of California, Berkeley, said the tiny computer chips are reaching their limits in size and ability.

“We’re hitting the physical limits,” Asanović said. “They’re not getting much faster or much smaller.”

Transistors switch electronic signals within electronic devices, providing the processing power for a machine. They are used in every modern electronic device — often with billions in an individual machine.

Engineering Prof. Scott Mahlke, who’s part of Asanović’s team, said transistors used to be able to shrink by half each year, allowing more of these components to fit on a single chip. The rapid increase in the density of transistors dramatically increased the overall computing power of electronics over the past half-century.

“We lost the ability to double speed,” Mahlke said. “They just can’t get any smaller. Once the computer industry loses the doubling, what do we do? We can’t rely on circuits. We need better architecture that uses the existing transistors better that provide the gains people are used to.”

Because transistors may not be able to shrink further within the next five years, devices will require better, specialized architectures to improve overall performance and functionality, Asanović said.

“One of the big trade-offs in processing is ‘more specialized, less general purposes,’ ” Asanović said. “How are we going to make specialized devices widely usable? You want to specialize more but you want to retain (general functionality).”

Mahlke said cell phones that are aware of their environment could be a product of specialized technology currently under development. For instance, one’s cell phone might be able to instantly translate a sign written in a foreign language.

“You can look things up on the web, but it’s not really aware of your presence,” Mahlke said. “Cell phones are going to be a lot more aware of your surroundings. It could whisper in your ear, ‘That’s Mr. Jones over there and you don’t like him.’ Today’s phones couldn’t keep up.”

Moinuddin Qureshi, associate professor of computer systems and software at the Georgia Institute of Technology, is a researcher for the C-FAR storage team. He studies ways to make tomorrow’s phone “keep up” in terms of memory capacity; noting that future memory systems could be 100 times faster than current systems.

Qureshi and his fellow researchers are focused on packing more bytes, or units of memory, into the same physical space to allow more advanced and complex systems.

These futuristic operating systems could then be used to instantly restore data and programs in use before power was lost. The research also focuses on improving power efficiency — perhaps one day enabling heavy users to go a day without toting around a power cable or extra battery.

Leave a comment

Your email address will not be published.