Digital Dust Sweeps Away Traditional Networking

Disposable sensing and communications systems converge into a cubic-millimeter bundle.

Advances in miniaturization, integration and energy management show that a complete wireless sensor/communication system can be merged into a package the size of a grain of sand and networked. Applications are far-reaching—from military sensor networks to industrial quality control.

The goal of the research, known as the Smart Dust project, is to explore the limits of miniaturization and power consumption in autonomous sensor nodes. These devices could be used to monitor weather patterns or placed on a cereal box before shipping so that a store manager could determine whether the product was exposed to moisture during delivery. These tiny intelligent nodes, or motes, are possible because of the recent convergence of microelectromechanical systems, wireless communications and digital circuitry. The project, funded by the Defense Advanced Research Projects Agency, Arlington, Virginia, is being conducted by researchers at the University of California–Berkeley.

According to Seth Hollar, a graduate student at the university, the time needed to develop a complete Smart Dust system initially was on the order of years. Scientists required an alternative means to test basic behaviors of the dust in a more timely manner. Hollar and his colleagues created larger prototypes, which they call commercial off-the-shelf (COTS) dust, to figure out what computational power would be needed for the smaller devices. “We needed to try out the COTS devices and test their abilities to see if what we wanted to implement would actually work on the large scale first,” he says. “COTS dust created on a cubic-inch scale could serve as a platform to run a variety of algorithms to test various behaviors that smart dust would exhibit.”

Hollar demonstrated that the devices can perform such tasks as measuring weather data and sending it to a central source for analysis. The algorithms for this programming have given researchers the base to determine processor capabilities for the newer and smaller smart dust devices, Hollar shares.

Smart dust, like its larger counterpart, is being designed with four basic components: power, computation, sensors and communication. Each dust mote will need to survive on extremely low power while tracking and communicating the data it receives.

Brett Warneke, a Berkeley doctoral candidate, is working on the computing element of the mote. “My main focus is building a microprocessor that really pushes the limits of energy consumption,” he says. Circuit techniques, logic level and architecture are areas of study that are being investigated to help keep energy use low.

“We do a lot of power cycling,” Warneke says. For example, each dust component is turned on while needed then turned off. This gives researchers more control over the power cycle. Even the microprocessor will be in an off mode as needed, he notes. Only a small unit will stay alive to awaken the rest of the system when necessary. After the devices are activated, they will operate from very low voltage power supplies, currently running between 0.3 volt and 1 volt.

At its present stage of development, smart dust requires a battery the size of one used in a hearing aid, which composes 70 percent of the mote’s total size, Hollar says. In later versions, the dust will use solar power, allowing the mote’s size to shrink dramatically. “Power will always be a concern in terms of how long you can keep the device running,” Hollar shares. “When the device is actively communicating, the question becomes from what distance can the unit communicate with other devices? But this is not a show stopper,” he adds.

Communications is another consideration in the device’s design. The Smart Dust team chose to use free-space optical communication rather than radio frequency communication because this method requires significantly less energy per bit, university researchers say.

The optical design employs a microelectromechanical system (MEMS) device that allows passive optical communication, Hollar relates. Low-power communication between a dust mote and a base station receiver is possible by using a microfabricated corner cube retroreflector (CCR). The CCR is made of three gold-coated polysilicon mirrors positioned at right angles. When a laser shines on the mirrors, the light is reflected back to the source, Hollar explains. This method, which consumes little power, is an effective way to send information.

A base station receiver that collects the information from each mote is an important aspect of the communication system, Brian Leibowitz, graduate student, says. He is tasked with building the receiver that sits at the base station and gathers the data from each dust device throughout the network. The implementation techniques are tricky, he shares. Leibowitz compares the receiver to a video camera. Each dust mote flashes light back at the base station to communicate its data, and each mote’s lights blink during different portions of the video image. A computer then processes the video stream and decodes all of the data.

“We’re going to take this concept one step further and build a microchip that is similar to a video camera except in every pixel in the camera you will have a complete communications receiver,” he says. Each pixel has the electronic capability to monitor its own light level, detect incoming data and correctly receive the information. The challenge is to make the receiver’s circuits small enough and at a low enough power to put one copy at every pixel in the camera, he explains.

Each pixel in the receiver requires a photo detector and circuits designated to process the light received by the detector. “One of the issues is that we’re restricted to a certain size for the pixel,” Leibowitz says. “If the circuits take three-quarters of the allowed area, then there is only one-quarter of the area left for the photo detector.” This results in a 75 percent loss of light because it is striking a portion of the chip that does not have a photo detector on it, he says.

One of many methods for distributing these tiny motes—now measuring only 63 cubic millimeters with a first-generation delivery date of next summer—is via a silicon maple seed design, experts say. A honeycombed layer of silicon 0.1 millimeter thick is the base for a 3 x 10 millimeter winglet. With a cubic millimeter of silicon attached, these wings auto-rotate as they drop, much like a maple seed would fall.

Additional delivery methods include unmanned aerial vehicles that could drop the devices from 30,000 feet and act as interrogation platforms, artillery shells and individual placement for industrial operations.

Despite the challenges, smart dust, equipped with many different types of sensors—such as for temperature, light and humidity as well as an accelerometer—will have limitless applications, Hollar says. “The realization of smart dust’s potential lies in using large numbers,” he notes. For example, hundreds of dust motes could be distributed over a few acres of land and could monitor temperature at hundreds of different points as well as provide information on traveling heat waves or moving cold fronts. This information, coupled with humidity data, could help farmers monitor crops at a detailed level or aid observation for environmental protection efforts.

Military applications also are possible, Hollar points out. With the ability to detect and record temperature, light and sound, the dust could be scattered throughout a surveillance area. Likewise, if a large vehicle such as a tank would pass by an area, the dust would detect the vehicle’s vibrations and upload the sensor data when queried by a laser interrogator system. “While no one sensor could definitively detect a passing vehicle, the aggregate of sensors would provide enough spatial information that, when processed, would produce vehicle detection,” he says.

Hollar also has conducted work with MEMS accelerometers—devices that measure vibrations—and smart dust applications. The project involves a virtual keyboard for which a person applies motes to each fingernail in a nail-polish form or like press-on fingernails, he relates. Currently in glove form, the keyboard is able to recognize all 26 letters of the alphabet and can read gestures at one character per second. “The idea is to shrink the glove into fingertip devices . . . and then you could be anywhere and work on a computer,” Hollar says. “Check your e-mail on the subway or do some work while you’re at a boring meeting,” he quips.

Smart dust eventually will be made for pennies apiece, Hollar says. “The true power of this is that you have many of them,” he shares. “The power is in the numbers.” Hollar also emphasizes that every mote does not have to work for an application to be accurate or successful. “The whole picture involves looking at the sensor data collected, processing it, and figuring out a higher level understanding of what’s going on. The motes are the medium to capture that sensor data.”

Warneke notes that the diversity of applications brings about a change in thinking. “I think that new modes of thinking are generated with this project because the devices are disposable,” he says. “We can distribute thousands of them and keep the cost down. This is a new way of looking at networking because we have disposable elements and the data from any particular node is not that important. In traditional networks, every piece of data is important.”