December 3, 2002 2:15 PM PST
Intel taking Internet to Mars
Intel's vision for "proactive computing"--in which remote sensors will feed information about the physical world to computers for analysis and use by humans--is moving closer to reality, said David Tennenhouse, vice president and director of Intel research, at the MicroVentures conference taking place here this week.
"The technology is at a point where you can go out in certain niches and get a return on investment," Tennenhouse said in an interview. "If that can be successful, you can pull a train of investment."
Additionally, the company is raising the ante in showing how and where the technology can be used. Intel itself, for example, has teamed with the Jet Propulsion Laboratory to use the technology to create an Internet link to Mars. The remote sensors will gather raw data about the planet's environment and beam it up to an orbiting router.
"The hard part is connecting the Mars Internet to the Earth Internet over a whopping delay," Tennenhouse said, pointing out that it will take three minutes for data to travel from Mars to Earth. Most likely, the data will be cached, so the network will function more like an e-mail system.
A different group of researchers is contemplating an undersea network. So far, Intel has shown how the technology can be used to unobtrusively monitor animal habitats and collect forest fire data.
Like nearly all its research efforts, Intel's involvement in proactive computing is geared toward an overriding goal: sell more chips. Automating the process for gathering data will seed the market for more powerful computers and servers.
Intel's communications division will likely also benefit. While Intel probably won't manufacture the remote sensors, it is very interested in building the chips for relay stations that would gather information from the sensors and then pass it on to a computer. These relay stations would likely contain 802.11 chips, flash memory and XScale processors, Tennenhouse said.
The sensors can be used to collect a variety of information. In one experiment, sensors are being used to monitor Alzheimer's patients. The devices remind patients of daily milestones, such as when to prepare lunch, and can send messages to therapists in case a patient fails to accomplish a task. Health care and general industrial customers are two of the initial target markets for sensors along these lines.
Companies such as Johnson Controls and Honeywell currently manufacture sensors that monitor heat, humidity, temperature and other environmental factors. These existing devices, however, largely require wired networks and are more expensive. If Intel's proactive-computing efforts pay off, these companies may be forced to adopt the new technology or work hard to protect their niches.
Several technological issues, though, still need to be ironed out.
"The things are always larger than you want. The battery life is not as good as you want," Tennenhouse said.
Another problem comes in data duplication. If two sensors detect an object, a bird in flight for example, a human operator can't immediately tell if two different sensors saw the same bird, or two different ones. The problem first emerged in the 1940s with the development of targeting systems and has yet to fully be resolved, Tennenhouse said.
And networking standards have yet to be finalized.
"Power is the big problem with 802.11," Tennenhouse said. Some of the start-ups actually use chips from cordless phones to network their sensors. Bluetooth is also under consideration.Nonetheless, 802.11 will likely win out. Philips and other companies are currently promoting Zigbee, otherwise known as 802.11.15.4, as an energy-efficient 802.11 standard. Standards for extending the range of the wireless protocol will also emerge.
"People always underestimate how far down and how far up you can drag something," Tennenhouse said.