Networked Embedded Systems: Sensor Networks and Beyond

When Kris Pister first started thinking about wireless sensor networks at a RAND workshop back in 1992, he didn’t imagine they would become a reality in just a matter of years. “It’s been really incredible to watch it grow from a decade ago, when it really was science fiction, to seeing it mature into an active research area and finally be commercialized. Now there are a whole bunch of companies doing this,” says Pister, an EECS professor at UC Berkeley who coined the phrase “smart dust” to describe individual sensor nodes.

Today, sensor networks are being used to monitor subtle environmental changes to California’s redwood canopy. They’ve replaced expensive and difficult-to-install power and temperature monitoring infrastructure in the Chicago Public Health Department. They’re even being developed to provide surveillance along the U.S.-Mexico border. As academic and laboratory research in the area flourishes, established and start-up companies alike are actively ramping up their research to meet industry’s needs. And this is just the beginning.

“The technology has tremendous applications. Many are still exploratory, but since so many universities and labs are pursuing them, eventually this is going to pick up in a big way,” says Ruzena Bajcsy, UC Berkeley EECS professor and CITRIS director emeritus.

CITRIS director Shankar Sastry predicts: “These network-embedded systems are going to be the infrastructure of tomorrow. They will be as much a part of our lives in the future as the Internet is today.”

A microscopic view of the same microchip.
(Image courtesy of Kris Pister)
CITRIS researchers are working on reductions in the size of microchips as well as how to power and network these devices.
(Image courtesy of Kris Pister)

To reach that point, a number of technical challenges still must be overcome. That’s why CITRIS researchers are engaged in projects ranging from strengthening the underlying hardware and software, so that it’s more reliable and secure, to developing new and revolutionary applications. What they’ve found is it’s often the very same attributes that make wireless sensor networks so convenient and flexible that also generate the obstacles between the current generation and future large scale deployments of networked embedded systems, that is systems embedded physically in our physical infrastructure.

For example, mobility and small size make wireless sensor networks cheap and easy to set up, but they also pose a security risk because the motes are more easily tampered with. Or take power. While the ever-shrinking mote doesn’t ask for much power to do all that sensing, processing, and transmitting, it does require some. Because it’s wireless, that means batteries. And batteries not only limit how small a node can be, but also require frequent replacement.

For the latter, progress is already being made. “With existing technology in industry, to get a 10-year lifetime out of nodes, you might need a D-cell or C-cell battery. With the next generation, you’ll be able to get a lifetime with a fat coin cell battery, and the generation after that it will be shrinking down,” says Pister. His student Ben Cook is publishing a paper at ISSCC (Intl. Solid State Circuits Conference in San Francisco) in February in which he demonstrates that he can operate a radio receiver in the same frequency band as WiFi but at 200 microwatts, about 100 times lower than the best commercially available radios used in wireless sensor networks. Meanwhile Berkeley professors Paul Wright (ME) and Jan Rabaey (EECS) have been working on energy-scavenging batteries, which use “ambient” sources of power, like floor and wall vibrations, to recharge.

Scalability is another concern. “The technology is still somewhat brittle so when you start doing large deployments problems with reliability and security arise,” says Sastry. He says the next generation of networks will need to have fewer false alarms, be more trustworthy, and, in addition to sensing, wield more control over the systems they monitor. For example a sensor monitoring the flow of oil in a pipeline might be able to automatically modify it when needed.

Sastry’s recently completed Network Embedded Systems Technology (NEST) project has produced an experimental platform on which researchers can test out their solutions to these types of problems. Similarly, the Dynamic Ad-Hoc Wireless Network (DAWN) research being spearheaded by UC Santa Cruz computer engineering professor J.J. Garcia-Luna-Aceves brings together scientists from seven leading universities to develop scalable, ad-hoc or peer-to-peer wireless networks, which could be used for communication in battlefields and emergency situations.

Driving all this innovation is the promise of new applications. As people dream up new ways to use sensor networks, they will adapt the technology to meet their needs. That, in turn, opens up new opportunities.

Bajcsy’s current work using sensors to monitor elderly people is a good example of this cycle. “We’ve been redesigning these sensors so that, in addition to the 3D accelerometers, they’ll have gyroscopes that can detect the posture of these people and hopefully—we don’t know yet—detect when they are losing balance so we could alert them when they are about to fall. We’re also interfacing it with a Bluetooth so it can make a call for help if indeed a fall does occur,” she says. Although her motivation for these modifications is very specific, it’s not difficult to imagine how these applications might be beneficial in very different scenarios.

“There’s going to be a real revolution when we solve the location problem,” says Pister. He’s referring to the fact that sensor networks have been successfully combined with GPS technologies to track movement, but GPS doesn’t work indoors. Hoping to solve this conundrum, graduate student Steven Lanzisera is developing a tool that would enable the nodes to calculate their own position to within three feet based on their proximity to other nodes in their network. Fire Information and Rescue Equipment (FIRE), a system designed by Berkeley mechanical engineering students that uses sensor networks to help firefighters safely navigate high rises, is already making use of similar technology.

Bajcsy is excited not only by the social impact that these new tools will have, but the economic and educational impacts as well. “We in computer science are worried about the decline of students coming into our departments. Well, this technology is much harder to outsource than software is. You can have these gadgets made abroad, but the installation and maintenance is local, so you will need the skills of engineers,” she says.

To help train the future workforce, CITRIS is launching a new Service Sciences Engineering and Management degree program that is inspired partially by the sensor web agenda. This program is expected to open its doors in Fall 2006. Meanwhile a number of start-up companies such as Pister’s Dust Networks, UC Berkeley computer science professor David Culler’s Arched Rock, and Moteiv, which was founded by three Berkeley engineering graduate students, are adapting the technology to make it easier for industry to use.

“What we in the universities had done was make it easy for other academics to build applications and do research in this area, but it wasn’t easy for companies to get in. What we set out to do at Dust Networks was to build a networking product that would be as easy to use as pushing the WiFi card into your laptop,” says Pister.

With every advancement and new contribution, the day when sensor webs become an integral part of life grows closer, though it’s difficult to say just how they will be used. As Sastry points out: “With sensor webs there’s no single killer app, because with sensor webs there are thousands of killer apps.”