 |
PALO ALTO, Calif., April 25, 2003 — The Palo Alto Research Center (PARC) has an illustrious history of technological innovation, including Ethernet networking and the personal computer. In the future, it hopes to add wireless sensors to its list.
The famed research center, spun off by Xerox Corp. in 2002, hosted the second Information Processing in Sensor Networks workshop this week amid anticipation of significant funding grants from the National Science Foundation.
 |
“This is the hot area” for research funding at the NSF, John Cozzens, program director of NSF’s Signal Processing Sensor Program, told Small Times.
Cozzens said the NSF received more than 1,000 research proposals for its Sensors and Sensor Networks initiative. In May between 55 and 60 of these proposals will receive grants as large as $2.5 million over periods of up to five years. The NSF expects to award a total of $34 million, more than a third of the $110 million earmarked for Information Technology research. Some of those projects are likely to include microelectromechanical systems (MEMS) research, and some may include nanotechnology.
Cozzens called the PARC-hosted conference, which drew 180 participants, the single most important gathering for scientists working on various aspects of sensor networks. He said technology advances have generated intense interest in sensor networks. “The magic words are surveillance and security,” he said.
The NSF’s emphasis on sensor networks is just one of the research initiatives in the field, which together should mean good things for the small tech world. While today’s rudimentary commercial efforts at sensor networks don’t typically use small tech, the near-term future will likely see a number of approaches that use MEMS. Nanotechnology is further out, but may prove useful — as a power source, data storage and even communications filters.
Still, while government-funded research helped create the Internet, cellular phone networks and other mass technologies, funding is not a guarantee of commercial viability, said Steve Jurvetson of venture capitalist Draper Fisher Jurvetson (DFJ). “You could argue that if the field is still ripe for grants, it’s not applicable for VCs,” he said. At the same time, he said, knowing what areas are drawing research dollars could suggest ancillary fields worth funding. DFJ has invested in some sensor firms, most notably Ember Corp.
Most believe miniaturization, whether through conventional methods, MEMS or nanotech, will drive the spread of sensor networks. But plenty of issues need working through, on the hardware, software and social fronts. For instance, Kris Pister, CEO of Dust Inc., a Berkeley, Calif. startup that sells networks of sensors, called motes, said in his opening keynote at the IPSN conference that the major issue in wireless sensor networks is the power needed to communicate data.
Pister said that a 3-mote network could send 10 alarms a day every day for four years. But a 300-node network would last only two weeks. “We can build large networks, but not with a very long lifetime,” he said. He said the typical Dust deployment was less than 100 nodes, and the biggest network he’s aware of remains the 800-node demonstration network he built for an Intel Developers Forum. In fact, Dust itself hasn’t made 1,000 motes this year, though it is generating enough revenue to fund itself, and has turned down at least two term sheets from venture capitalists.
Pister also said that sensors with high-resolution, 1sq.mm video cameras may come to market within two years, but it isn’t clear that customers will pay more for these than for current 5sq.mm cameras on the market.
Pister, who’s been skeptical of the use of nanotechnology in sensor networks, also said he’s changed his mind. Pister said advances in nanotech sensors, computation, storage and communications filters may lead to the ability to build motes on the nanoscale, though not for at least five years. He also thinks nanotech-based power cells offer intriguing potential.
Other unanswered questions involve how companies will filter the data generated by large networks of sensors, nor what sorts of applications might run on them. Feng Zhao, the PARC scientist who heads its research efforts in collaborative and MEMS sensors projects, says the lab has developed several types of applications for them, to learn about how data works in a dynamic routing environment, and also what role, if any, TCP/IP will play. He says there is no viable killer application for sensor networks right now.
“We need a small set of driving applications” for sensor networks to be useful, Zhao said. Once they are, he expects to see large companies come in and aggressively license technologies and push them into volume production.
Zhao said the size and number of the NSF grants will accelerate the development of standards for sensor networks, laying the foundation for widespread commercialization. For instance, now sensor networks using hardware from different manufacturers cannot easily communicate.
“NSF is doing a big thing” for the industry, he said.
The NSF is not the only government source of research funding for sensors, and there are a number of companies and operations like PARC funding the work. All this work is predicated on creating what Pister predicts will be tiny sensors everywhere from “the boardroom to the bedroom.” That raises obvious social issues, which are being discussed now, well before they’re feasible.
Cozzens said the said NSF is well aware of the privacy issues such networks pose. “Look, look, it’s a trade-off — do you want to be secure or not?” he said. “This technology will make us more secure, but there is a price for all this.”
For now, though, much work remains before sensor networks make their way into the factory, let alone the bedroom or the boardroom.