Intel's next challenge: Mastering physical world

SANTA CLARA, Calif.--Intel researchers are tinkering with a number of technologies that could help solve some of society's pressing problems--like finding a parking space.

IrisNet, a project currently under development by Carnegie Mellon University and Intel, is a sensor network designed to provide up-to-date information about the outside world. Cameras hooked into the network off the Oregon coast, for instance, relay information about anomalies--such as the sudden presence of a fleet of fishing boats or a whale--to remote PCs.

Similarly, wireless, or wired, IrisNet sensors can be placed in a parking lot to monitor available spots, said Phillip Gibbons, one of the researchers. A driver with a cell phone or an Internet-enabled car coming to an airport, for instance, could log on to a Web site and determine which long-term parking lots have the best availability.

"We are doing for live data what Google does for content," Gibbons said.

Using technology to better navigate the obstacles of real life is one of the dominant themes of Intel's research efforts and one of the main topics at its "Research Day," which took place here Wednesday. Also called proactive computing, the effort involves developing machines that can obtain, filter and send information with as little human interferences as possible.

Some of the proactive efforts revolve around improving current networks and computers, while others seek to use computing in new fields, like home health care. In one experiment, a glove with radio frequency identification (RFID) technology was given to seniors. The data obtained from the glove could let remote health care workers anticipate needs or get warning of sudden problems.

"We are interested in expanding the landscape for silicon," Intel Chief Technology Officer Pat Gelsinger said.

The various projects range in size and scope. PlanetLab, for instance, is a worldwide network that allows developers to try out distributed applications.

Invite Michael Kanellos into your in-box Senior department editor Michael Kanellos scrutinizes the hardware industry in a weekly column that ranges from chips to servers and other critical business systems. Enterprise Hardware every Wednesday.

Netbait, which runs on the PlanetLab network, is an application that detects sudden bursts of unusual Internet traffic on a global scale, said Paul Brett, an Intel researcher. Ideally, early detection can help identify worm outbreaks earlier, he said.

By contrast, Beehive, under development at Cornell University, and Coral seek to reduce Internet search and access times by pointing Web users to the closest source of the data sought. It can also help stem problems from any outages.

"Every component is unreliable, but as a whole (the Internet) works," Brett said.

Started last year by Intel, Hewlett-Packard and Google, PlanetLab has grown from 150 servers to 350-plus servers. Servers were activated in Canada at the end of last year, while servers in Brazil and India will be activated soon.

Other Intel projects include:

• The Personal Server. This technology is a box about the size of a deck of cards that contains an Xscale processor, flash cards for storing data, and various radios for networking. While riding in a car, a person can play music stored on the server through the car's speakers by transmitting data over an FM signal. Inside a conference room, the same personal server could then be used to beam PowerPoint slides onto a Bluetooth-enabled flat-panel screen, said Roy Want, one of the researchers on the project.

"It takes advantage of the (input/output) around you," Want said.

This summer, about 20 interns will carry these devices around in a trial called Musicology, he said.

• The GrooveBox. This application, being developed in Intel's Cambridge, England, labs, allows a cell phone with a camera to interact with PCs or public kiosks tuned to specific Web sites. Optical symbols are embedded into special Web pages. The cell phone's camera locks onto these and then uses them to send commands to the remote screen.

In principal, the technology functions the same way as bar codes, but bar codes are often too difficult for camera phones to accurately interpret, said Richard Sharp, a researcher on the project.

• Computational Nanovision. This technology combines probability theory with semiconductor testing procedures. In a nutshell, semiconductor testing is expensive, and the application seeks to determine, through probability, which tests can be eliminated or reduced without degrading the quality of the output.

Probability is one of the fundamental pillars of these new, extroverted computing paradigms, Intel's Gelsinger said. Because you are dealing with the physical world (as opposed to finite data sets) getting an exact, fixed answer is impossible, but accurate results can be had.

"Statistics affects everything in computing," he said. "You no longer get the answer. You get the answer with a range of probability."

Like nearly all research projects, not everything shown by the company will come to market, and a substantial amount of work will be required to create commercially viable projects.

In the IrisNet project, the researchers are working on ways for the various sensor cameras to stitch together their images to provide a comprehensive picture of the parking lot, Gibbons said. The sensor cameras are also part of the push toward probability. They only ping a person when something unusual--and hence probably interesting--comes into view.

Researchers also have to figure out an efficient database and caching strategy for maintaining the data, he said.