APS News

Physics and Technology Forefronts

Multiple Disciplines Converge in Embedded Wireless Sensor Networks

by Jennifer Ouellette

Schematic diagram of an embedded wireless sensor network
Schematic diagram of an embedded wireless sensor network

Imagine a world in which buildings can detect their own structural faults and respond to earthquake tremors in real time. Public health officials continuously monitor contamination levels in water supplies and can even trace contaminants back to their source. When bacterial levels in coastal waters get too high, surfers, swimmers and fishermen are alerted immediately. Every aspect of modern life would be monitored via wireless linkages.

This might strike some people as a futuristic scenario ripped from the pages of the latest science fiction novel, but it is becoming a technological reality. At the Center for Embedded Networked Systems (CENS), an interdisciplinary team of researchers from six California institutions combine microsensors, actuators, robotics, low-power electronics, and wireless network technology into compact, integrated packages for distributed monitoring and control.

These so-called embedded wireless sensor networks can be designed to monitor and collect information on everything from plankton colonies, endangered species, soil and air contaminants, traffic flow, medical patients, and buildings and bridges. They are yielding an unprecedented level of hard scientific data. And they can do so far more cheaply, with greater energy efficiency, than the instruments now in use, which are linked by wires and power lines. The emergence of such networks is due in part to the explosion in wireless-enabled consumer electronics, and the continued emphasis on miniaturization.

CENS Director Deborah Estrin strives to foster a multidisciplinary innovation model that seeks to create a tight “feedback loop” between users and innovators. For many users, the result is “a tremendous innovation in their ability to understand the physical processes they study.”

The CENS collaboration involves researchers at the University of California, Los Angeles (UCLA), the University of Southern California (USC), the University of California, Riverside, CalTech, the University of California, Merced, and California State University at Los Angeles. The center’s work involves everything from fundamental communication theory to embedded computing, networking, electrical engineering, sensor technology and statistics principles. For UCLA’s William Kaiser, there is a clear physics aspect to the project. “We are making measurements, and we need to optimize the information return relative to all the different sources in our environment, so we’re relying throughout on physics principles,” he says.

CENS is developing systems that can be used to characterize natural phenomena in two primary environments: terrestrial ecosystems over a broad range of climate types in North America, Central America, and Australia; and aquatic systems, such as coastal marine environments, urban rivers, and streams.

Terrestrial Ecosystems
In California’s San Jacinto Mountains, robotic arrays of sensors and cameras–dubbed “treebots” –move up and down cables attached to trees to actively monitor changes in temperature, humidity or sunlight. The treebots are individual nodes in a Networked Infomechanical System (NIMS), which combines robotics with multiple environmental sensors, including actuated imaging systems and a wireless network. The treebots are highly autonomous and can communicate with other devices. They have their own servers and can use wireless net links to send sample information and other measurement data back to the home laboratory, located in this case at UCLA.

Treebots are more flexible than fixed nodes. They can be raised and lowered as needed to collect data at different levels of the forest canopy. Full-motion cameras mounted on high observation towers track wildlife and changes in plant growth, while a “nestbox” collects time-lapsed images to document wildlife nesting activity. There is even a microclimate array to collect climate data above and below ground.

Because the robotic nodes are constantly changing location, the treebot system also boasts a capability Kaiser calls adaptive sensing. Adaptive sensing involves finding the right data at the right time, thanks to self-configuring systems that adapt to unpredictable environments where pre-configuration and manual intervention aren’t possible. “Any time you put down a number of static nodes, they are in fixed locations, which limits the spatial resolution you can achieve,” he says. “But with a robotic node that can move on command, you can get 3D precision resolution.”

Taken together, the network will help scientists understand the subtle changes that take place over time in light, humidity and CO2 levels, not to mention the growth of individual leaves and branches. Statistical techniques are vital to achieving this. “One needs to be able to characterize the incoming data stream from the sensors and use that to adaptively adjust the way in which a sensor operates, or where it moves to collect more data,” says Kaiser.

Other new technologies that have been incorporated into the treebot’s sensing network include a thermal mapping device developed by Kaiser’s UCLA colleague, Phil Rundel. It maps the surface temperature of objects as the robotic system scans. Rundel has used the device to study, for example, the thermal properties of unique alpine plants that inhabit very high elevations, and must therefore withstand extreme temperature differences between the very cold atmosphere and the much warmer ground. Rundel has also developed a laser mapper, which enables the treebot to scan the forest and reconstruct the shape of objects in 3D.

Aquatic Ecosystems
Richard Ambrose heads the public health program at UCLA and uses embedded wireless sensing networks to characterize urban streams. One area of interest is the problem of excessive algae. When too many nutrients get into the water, whether it’s a river, lake, or pond, the result is excess algae.

Gaining detailed knowledge about the specifics of the relationship between excessive nutrients and too much algae is difficult in the real world because there are so many factors that influence how much algae grows and where it grows, including sunlight, which water substrates contain the highest concentrations of nutrients, and how fast the water flows. There’s even a time factor, since algae can soak up nutrients like a sponge and store them for later.

Using NIMS, researchers can observe all the variations in flow and concentrations of contaminants within a stream. One of the more interesting findings is that concentration levels of nitrogen and phosphorus vary with the time of day the samples are collected. “NIMS gives us the possibility of collecting data on a temporal scale that we would never be able to get otherwise,” says Ambrose. “We can track the dynamics of the nutrients so we can understand that relationship better.”

Currently, his cable system spans the stream in cross sections moving in 2D, with sensors that can move across and along the stream. The ultimate goal is to have sensors that can also move down into the stream to take measurements at different depths to give a 3D picture rapidly.

CENS collaborator Tom Harmon at UC-Merced is using NIMS to better understand the origin of toxic material, algae, and bacteria in marine/coastal areas, using a system of fixed buoys and robotic boats that automatically move to take samples of the environment. The goal is to determine how environmental change — whether natural or induced by human activit y— can lead to excessive growth of toxic bacteria or algae.

UCLA’s David Caron is using a similar approach. His team has developed a version of the robot system that can operate underwater, scanning a stream cross-section to determine what kinds of contaminants are flowing past. But the system also precisely measures the flow in all directions, including eddies and other circulations. This allows them to compute not only the concentration of the contaminant in the San Joaquin River, for example, but at what rate it is flowing downstream.

On the Horizon
The potential applications for such systems extend far beyond the study of complex ecosystems. DARPA is interested in using them to monitor battlefield conditions. Embedded sensing could also be incorporated into concrete bridges to monitor vibration, stress, changes in temperature, even cracking.

Attaching nodes to water or power meters in residential neighborhoods could make existing meter reading methods obsolete. Placing nodes along a sensor-equipped highway would enable police to better monitor traffic flow.

The next step, according to Estrin, involves more widespread proliferation of embedded wireless network systems among scientists — in every discipline. She cites NEON as one example of a continental multiscale sensor network. NEON is being designed to track changes in various environments, from urban and suburban areas to more rural and wild settings. She also hopes to combine this new observational capability with remote sensing and existing GIS-based modeling facilities. Another example of a distributed sensor-based observatory is Earthscope, which connects thousands of stations to map Earth’s interior and study crust deformation, searching for clues to the planet’s early evolution.

Ultimately, rampant proliferation should bring the center’s innovation “feedback loop” full circle. The miniaturization and wireless trends in consumer electronics paved the way for developing prototype embedded sensing networks for scientific applications, but as they proliferate, more consumer-oriented urban sensing applications will emerge. Estrin foresees a day when individuals make use of acoustic, imaging, or personal-health-monitoring sensors, communicating with and through their already omnipresent cell phones: “That’s when we’ll start to see this proliferate out into non-scientific applications.”

Sound Bytes
UCLA’s Mark Hansen has never been the sort of statistician who stays inside what he calls the “physics box.” In addition to his work with CENS developing statistical algorithms for embedded wireless sensing networks, he is an accomplished multimedia artist. And his science feeds directly into his art.

In the late 1990s, when Hansen was at Bell Labs, the company revived a program in art technology that originated in the 1960s, that teamed up engineers and scientists with New York City artists–among them Robert Rauschenberg and Andy Warhol. Thirty years later, Hansen hooked up with artist Ben Rubin and produced the “Listening Post” [www.earstudio.com]: a multimedia installation that is also an experiment in sonofication — the process of turning raw data into sound, instead of plotting it onto a graph.

Hansen and Rubin built a data stream using text from online bulletin boards and chat rooms, as well as tracking users’ Web browsing activities. This data was then processed by a voice synthesizer to “score” the video portion of the installation: a long panel of 231 small text displays, each about the size of a candy bar.

The end result is a visual and aural representation of data flow that proffers snippets of connectiveness, random glimpses of people interacting in the virtual world at any given moment. “Whether we like it or not, the flow of data exists to regulate our movements through the world, our behaviors,” says Hansen.

“The Listening Post” was featured on National Public Radio and won a 2003 Webby Award for Net Art. It will be up and running again at the San Jose Art Museum this summer, with the data stream expanded to include snippets from blogs, news sources, even Wikipedia. Hansen and Rubin are now working on delivering real-time data feeds to live actors, instead of using voice synthesizers and a grid of text displays — bringing the human element back into the technology.

APS encourages the redistribution of the materials included in this newspaper provided that attribution to the source is noted and the materials are not truncated or changed.

Editor: Alan Chodos
Associate Editor: Jennifer Ouellette
Staff Writer: Ernie Tretkoff