Intel Envisions 'Smart Dust'
Every sci-fi geek has heard of smart dust – tiny nanomachines that float around, collecting data and transmitting information. They played a role in Neal Stephenson's novel The Diamond Age, where he talked about clouds of smart dust floating invisibly in the atmosphere, occasionally getting into fights and leaving smart-dust carcasses that could eventually cause lung damage.
Smart dust isn't quite reality yet, but bigger versions – where the dust motes are small computer boards – are already in use. Motes are among dozens of ideas being pursued by Intel Research, a relatively new R&D arm of Intel Corp. (Nasdaq: INTC) devoted to technologies that have feasible applications but aren't yet serious businesses.
Intel executives gave the press a first introduction to Intel Research on Wednesday, at a reception near the company's Santa Clara, Calif., headquarters. Other technologies on display ranged from audio-visual speech recognition – where the computer reads your lips as well as interpreting your voice – to chips for analyzing DNA and proteins.
The motes fit into a larger Intel dream of "proactive computing" (never mind that there's no such word). This would be the next step after ubiquitous computing: With so many computing devices surrounding us, we're going to want them to talk with each other rather than having to interact with a human all the time, says David Tennenhouse, vice president of Intel's corporate technology group.
Part of the idea is to cut down on information overload from cell phones, PDAs, and whatever other devices invade your life in the future. "You want to have oceans of computers standing behind those, working behind the scenes," Tennenhouse says.
The life-cycle of an Intel Research project starts at a university. Intel has placed R&D offices near campuses such as U.C. Berkeley or the University of Cambridge, and it seeds those campuses with grants to pursue interesting-sounding research. As a line of study develops, Intel may increase the grants and start doing parallel projects internally.
The Intel-internal projects were on display at Wednesday's event. These are still in the research phase – they have conceivable applications, but it's not clear whether they'll become Intel products. Many could become areas that Intel simply encourages because they drive the need for microprocessor technology, the company's interest in video games being one example.
For the researchers, these projects present a chance to dabble without the usual market pressures. That's one reason why 13-year Intel veteran Ralph Kling, a low-power electronics expert, joined the mote team nine months ago. "You don't get to define new architectures that often," he says.
An Intel mote is a small computer board – about one-third the size of a credit card – that houses a sensor, a microcontroller, and a radio. The idea is to get these devices to create ad hoc networks among themselves, transmitting data to some central server or gateway. This would allow remote monitoring of conditions such as temperature or humidity; the idea works with chemical and biological sensors as well (anthrax detection was a popular example among the researchers).
The problem is that these devices have to be low-power, because they'll often be in places where it's inconvenient or impossible to be changing batteries all the time. That means they have to rely on very short-range transmission schemes, such as Bluetooth.
In an environment like a factory floor, many of the motes would be several network hops away from the gateway. Data would hop from mote to mote, with the chances of packet loss increasing at each hop.
Intel's demonstration on Wednesday showed the idea of placing a few 802.11 motes into the network. These acted as hubs, with other motes using their 802.11 brethren as gateways to the real gateway, thus cutting down the number of hops for most traffic. It also lowers the amount of work each mote has to do – since they wouldn't be forwarding one another's data in those multihop scenarios – which in turn lowers their power consumption and increases battery life.
So, what will it take for these motes to become truly dust-sized? The biggest challenge turns out to be the battery, because most of the other parts will shrink as technology advances.
"Sensors are very shrinkable using nanotechnology. CPUs [central processing units] we know will shrink. Memory, the same thing – it's all Moore's Law, essentially," Kling says.
Moore's Law doesn't apply to the radio, because it's made of analog components, which don't necessarily shrink with new generations of manufacturing technology. But Intel's plan is to move most of the radio into the digital realm and even handle many functions in software – the so-called "software-defined radio."
That leaves the battery, and because the battery's going to be slow to shrink, it might not be worth striving for dust-sized motes just yet. "Do you want to spend a lot of time designing a package that's going to be dominated by the battery?" Kling says.
Nearly 100 teams are using the motes, Tennenhouse says. They're being used to monitor birds' habitats on Great Duck Island off the coast of Maine. More practically, they could be used to monitor temperature throughout a building. Or an airplane could spray them over a forest fire, and they could report temperature levels for particular locations, giving firefighters a map of what's going on.
Among the most likely applications is one Intel itself could use. The company has up to 4,000 sensors in each of its manufacturing facilities, or fabs (fabrication facilities). They monitor the vibrations of fab equipment, detecting early signs of wear and tear so the machines can be fixed before they break. Great idea, but the 4,000 sensors have to be read by hand – a job that could be automated by motes.
The motes are still in the prototype stage. Intel's hope is to condense the board into a single chip, which doesn't appear likely until after 2005.
— Craig Matsumoto, Senior Editor, Light Reading