A Vast Web of Tiny Sensors

UCLA's Deborah Estrin discusses how networks of small, smart monitors can change the world and how we understand it

Deborah Estrin has large hopes for some tiny technology. One of the leading researchers in the fast-emerging field of sensor networks, Estrin believes that hundreds, even thousands of tiny wireless sensors working intelligently together to monitor temperature, water, and structural changes in the world around us will not only make companies more productive, but help protect the environment. These networks of tiny machines, now about the size of matchboxes, can extend the Internet to cover conceivably everything that moves, grows, makes noise, or heats up.

As the director of UCLA's Center for Embedded Networked Sensing (CENS), a two-year old group funded by a $40 million grant from the National Science Foundation, Estrin is working hard to make this "Internet of things" a reality. She's overseeing research designed to make devices and their networks smaller, cheaper, more energy efficient, and smarter, so they can be used to detect changes in the environment or assess the health of patients.

The center is conducting experiments in the lab and the field, including installing sensor networks throughout the 29-acre James Reserve in Southern California to study the nature preserve's ecosystem. Estrin spoke with BusinessWeek Internet Editor Heather Green about the promise of sensor networks, the work that needs to be done, and concerns about funding for innovative research. Following are edited excerpts from their conversation:

Q: How did you become interested in sensor networks?


What's interesting about this field are some of the same things that fascinated me about the Internet, which was by far the biggest distributed system to have emerged in a dispersed manner. This technology has the same potential. It's fascinating when you have thousands of distributed points that will have to work together. Making systems work in complex physical environments will be highly challenging. At the same time, the potential of using this technology to do things like monitoring our exposure to air containments and allowing scientists to observe the interaction in complex environments will have a great global impact.

A: What do you see as the potential of this technology?


We've had remote sensor technology for a long time. For instance, sensors on a satellite can see a huge field of view. But when you have a remote sensor, every pixel can represent kilometers. If you want to understand a process on a much smaller scale, you have to get up close and see that variability first hand.

Contaminated water is a good example. The soil in the ground is hard to observe. But the truth is, you're not trying simply to look at it, you're trying to understand how the contaminated water is transported or to see whether the contaminants are getting into the groundwater, and that's hard to track. The ground is heterogeneous. Sensor networks can reveal processes that were unobservable. It's the ability to bring the equivalent of a microscope [to these situations]. It could help us get a handle on how we exist in this world in the long term.

Q: What's the challenge of creating these kinds of networks vs. something like the Internet?


You have to put sensors in a highly distributed way -- that means that those devices have to be small and have to be wireless. You can't plug them in somewhere or have them sending out wires. You have to make the systems intelligent and have them tell you when something interesting is occurring, instead of just sending out continuous streams of data.

Q: Why are scientists excited about distributed sensor networks?


They're excited to observe things they just can't see right now. Water contaminants are one example, but there's also the whole issue of land use that's starved for data. Often you're making policy decisions without real insight. If the land use is too fragmented, then you can't protect an environment or species you're trying to help.

Right now, we're planting forests in exchange for carbon-distribution rights. It's poorly understood because we only have models of CO2 coming off the top of the forest, but it's coming off the sides of forests as well. Or we can do structural monitoring of bridges and buildings and know when joints are losing their integrity or beams and floors are torquing. These things are hard to observe without measuring many more points. Sensor networks are like the ability to run a CAT scan on the body instead of just doing a blood test.

Q: How far along are we in developing these kinds of systems?


We have some very exciting prototypes around where we're making these autonomous, self-configuring networks, and verifying that they're doing what they should do. This is very hard. We trust our computers now for many things, but we don't trust them to be that autonomous.

These systems are much more complex and highly distributed, and they're measuring all kinds of things, like the soil and the ground and having different kinds of interactions. We're working on research that has to do with the integrity of the system: How do we put in more intelligence and sophistication in the networks to allow them to be autonomous?

Q: What's the importance of these networks being autonomous?


Autonomous means that every device doesn't need human interaction. To make these systems functional, there will be many types of modalities that will trigger some sample collection, or some acoustic symbol processing. For these systems to scale and be usable, they have to be without the care and feeding we give to our laptops.

No one is interested in just getting feeds of data. You want to have answers to the questions you're asking. So you only want the network to send back some information when it exceeds a certain value. For instance, you tell a sensor you're only interested in knowing when it sees something different from its neighbors or you want to see the output of a image sensor only when its neighbors have all seen their sensors flat-lining. The more sophisticated you can get, the less data you will see, the more intelligence you will put on the network.

We're also working on sensors that have robotic capabilities. They don't have to travel on the ground but on aerial cables to enable robotic nodes to move around autonomously according to where interesting things are happening.

Q: What are some of the concerns you have about innovation and research?


This field started out because of original funding through DARPA (Defense Advanced Research Projects Agency). We were very fortunate to be able to develop and acquire a National Science Foundation center in this area. They have picked up the ball on the funding.

But what has happened is that DARPA is increasingly classifying their work and not funding university research. There's a big concern in the industry about where the next seeding is going to come from. How did Google get started? It was from government funding at Stanford. I'm concerned overall as to where information technology funding for that seeding 10 years from now will come from.

Q: Why is classifying research a problem?


They're classifying the research upfront. How do you get students involved in the next generation of information technology if that work is being classified? There's a vacuum being created in terms of funding of the research that's the driver of info tech and of innovators.

    Before it's here, it's on the Bloomberg Terminal. LEARN MORE