The Future of Remote Sensing
Look into the future of the digital world and you might find the digital world is looking right back at you. Advances in remote sensing are giving computer networks the eyes and ears they need to observe their physical surroundings. Sensors detect physical changes in pressure, temperature, light, sound, or chemical concentrations and then send a signal to a computer that does something in response. Scientists expect that billions of these devices will someday form rich sensory networks linked to digital backbones that put the environment itself online. The goal, says David Tennenhouse, the former chief scientist at DARPA and current director of research at Intel, will be to use dense arrays of networked sensors to extract as much "information per unit volume," about the environment as possible.
Much of the research driving small, inexpensive sensors is found in the area of MEMS, short for microelectromechanical systems. Scientists working with MEMS are creating tiny electronic features from silicon, some of them smaller than a red blood cell. MEMS is common in the computer chip industry but the technology extends to sensor design as well. For example, Kris Pister, a professor at the University of California at Berkeley, is developing a sensor he calls "smart dust" designed to be so small it literally floats in the air. These minute devices are self-powered and contain tiny on-board sensors and a computer on a scale of just five square millimeters -- roughly the size of an aspirin tablet. Pister is confident he can reduce their size to a single millimeter by 2001 and to airborne dust-like dimensions by 2003. The idea is to use them by the thousands in interconnected networks that communicate with each other using wireless signals. The environmental possibilities are highly varied: Pister envisions smart dust "motes" sprinkled out of airplanes monitoring the atmosphere or hovering in the dark recesses of factory stacks monitoring pollution, or used in farms to measure soil chemistry and pesticide levels. It's currently possible to pack the motes with the computing power of the first Intel computer chip -- just 200 microns long (one micron = one millionth of a meter) -- for about 10 cents. However, continuing advances in MEMS are expected to push the price down below a penny sometime in the future.
Smart dust motes are powered with batteries, which has raised some health concerns because the batteries contain toxic metals like lead and cadmium. But Pister dismisses contamination questions because the amounts of heavy metals used are so small. And when asked if inhaled motes could pose a health threat, he says, "Even if you did inhale them, they are too big to be absorbed. You would just cough them up."
The Eyes in the Sky
The most far-reaching environmental profiles will come from satellite-based remote sensors in space. Scientists at the National Aeronautics and Space Administration (NASA) envision rich flows of environmental data coming from grids of complementary satellites circling the globe. A number of satellites are already monitoring the global environment now. For example, the bus-sized Terra satellite, sent into orbit by the Earth Observing System (EOS) at NASA in December 1999, is measuring 16 of 24 parameters known to play a role in determining climate. Among these are aerosols, clouds, temperature, vegetation, and radiation. One of the most popular applications for satellite images involves linking them to Geographic Information Systems (GIS), a combination used to study global land use patterns; track oil spills, forest fires, and deforestation; and monitor the health of coral reefs.
To increase resolution, NASA is equipping its satellites with so-called "active" sensors that obtain data from lasers and radar that are shot down to target areas and reflected back to an on-board detector. One example is the Vegetation Canopy LIDAR, which beams lasers directly into forest canopies to investigate an elusive parameter scientists see as a kind of holy grail: the global carbon cycle. According to senior NASA scientist Jim Closs, measurements of the global carbon cycle will provide the clearest long-term picture yet of carbon dioxide fluctuations and their influence on global warming. Advances are also being made in hyperspectral remote sensing, which extracts greater amounts of data from reflected radiation than currently used technologies. These sensors yield precise measurements of chemicals in the atmosphere and, according to Closs, will allow scientists to double, or even triple, the number of parameters currently monitored.
The number of data satellites produced in a month now would have taken ten years to generate a decade ago, leading to difficult questions about how the data are going to be managed. Current processing capabilities are keeping up with the flow, but just barely. Mark Gray, a senior programmer at NASA, says the terabyte of raw data NASA collects from satellites every day -- equal to one trillion bits of information -- is beginning to strain computational capacity. To improve its storage capabilities, NASA is piggy-backing on commercially driven improvements by partnering with private-sector companies including Oracle and Silicon Graphics Inc.
But satellite sensors will comprise only part of a broader distributed network with its roots on earth. In the future, millions of low-cost devices embedded throughout the environment will add to the data-management challenge. At the University of Southern California's Information Sciences Institute, researchers are working to solve the problem by designing systems that transfer intelligence to the sensors themselves. This approach is based on a growing technology called "intelligent multitasking." Ramesh Govinda, a computer scientist at the ISI, suggests that embedded intelligence would eliminate the need for a centralized data processing facility. Each sensor would carry a microcomputer and some communications abilities, providing for collaborative signal processing and the ability to make "group decisions" about which data to send and when. For example, sensors in a factory could be designed to respond to toxic releases and spills, perhaps by shutting down an industrial process. If the release extended beyond industrial perimeters, factory-based sensors would communicate with networks of municipal sensors, in turn, initiating a series of protective actions directly within the community.
Eventually, remote sensing data will be accessible to the ordinary citizen on the street. For example, someone with a handheld wireless device will someday be able to access satellite data from the Internet, overlay it with GIS coordinates, and obtain on-the-spot atmospheric information for any location on the planet. Tennenhouse suggests that intelligent homes in the future will link to atmospheric satellite data and monitor their own internal environments accordingly. The key to all these applications is they will take humans out of the loop -- the sensor networks will be intelligent, proactive, and able to respond to environmental changes at lightning speed without human intervention. In this fashion, remote sensors will enable computers not only to "see" their environment, but also to shape their physical surroundings.
Next section: Autonomous Electronic Markets