News Focus
News Focus
icon url

fowler

10/11/02 7:29 AM

#1604 RE: fowler #1602

Interesting article:

Forbes Magazine
Sensors Gone Wild
Thursday October 10, 6:36 pm ET

By Benjamin Fulford


An experiment in the California desert provides tantalizing hints of how a networked world could make everyday life a lot more precise and profound.
Biologists at the James San Jacinto Mountains Reserve in southern California stumbled upon a teeny discovery in the world of squirrels a few months back. It turns out the little guys, heretofore known to exist on nuts and berries, have a thing for moss; they munch on it for the moisture. No one knew; no one had been in the right place at the right time to see it.

ADVERTISEMENT


But the squirrel in California was caught in the act when the animal tripped a 3-inch-square motion sensor that fired up a tiny wireless camera to record the event. The small sensor is one of dozens of digital spectators linked into a speedy wireless network and sprinkled across a 30-acre patch of wilderness 25 miles southwest of Palm Springs. The devices take note of the movements of animals, plant growth and such factors as time, temperature, air pressure, wind speed and changes in carbon use (you can see results at www.jamesreserve.edu).

Animal insights, however, aren't the point of this $40 million experiment. The real goal is to explore the uses of intelligent sensors, a technology whose promise suddenly seems huge. The applications for this "embedded intelligence" are vast and profound. Eventually large swaths of the Earth will communicate with the digital realm using millions of miniature sensors. "We will really be able to instrument the physical world," says Deborah Estrin, a UCLA computer science professor who runs the Center for Embedded Networked Sensing, which oversees the James Reserve project.

Sensors will be placed in bridges to detect and warn of structural weakness, and in water reservoirs to spot hazardous materials. Hospitals will track patients with such things as wireless bandages that warn of infection. Truck drivers will be able to dodge traffic jams based on slow-ups twenty cars ahead. Urban planners will track groundwater patterns and how much carbon dioxide cities are expelling, letting them make better land-use decisions. Farmers will be able to apply fertilizer and water only to the exact bits of the field that need them.

"I like to be conservative about things, but in a way [sensor networks] could be bigger than the Internet. The Net is relegated to a small screen and a keyboard. This will detect who you are and where you are. The whole analog world will interface with the Net," says Clark Nguyen, a professor of electrical engineering on leave from the University of Michigan to develop sensors for the Department of Defense's Advanced Research Projects Agency.

Darpa has made sensors a top priority, putting up $160 million of its own money and $500 million in matching funds from other federal agencies. The Pentagon wants what it calls "hyperspectral" data from a war zone. By spreading thousands of chips from a drone aircraft, the military will be able to follow enemy troops and detect bioweapons and electromagnetic noise.

Early this year a pilotless aircraft sprinkled three dozen cheap wireless magnetic sensors, each about the size of a credit card, along a road at the Marine Corps Air Ground Combat Center in Twenty Nine Palms, Calif. Once they hit the ground, these sensors automatically formed a wireless network and began scanning the environment for magnetic signals. When a vehicle rolled by they could tell from its magnetic signature what kind it was, its speed and direction. The readings were sent wirelessly via the aircraft to headquarters.

Before the sensor revolution can truly take off, researchers must overcome some crucial engineering hurdles in power management and data networking. The progress made by firms such as Intel (NasdaqNM:INTC - News) and Transmeta (NasdaqNM:TMTA - News) in producing low-power chips has helped make processing bits a much thriftier affair, says Estrin. But efficiencies have not kept pace when it comes to transmitting those same bits of data. The James Reserve still uses expensive and short-lived batteries. Estrin wants chips that sip power so slowly they can run on their own for years.

Some new sensors are getting so small--some tiny enough to be invisible to the naked eye--they will be able to run on 100 microwatts. (A microwatt is one-millionth of a watt. Compare that to a Pentium 4, which runs at 75 watts.) At the 100-microwatt level they could gather energy from ambient heat and photovoltaic cells, says Stephen Senturia, a specialist in microsystems at MIT. His colleagues are working on making chips so small they could power themselves like the watches that need only the kinetic energy generated by wrist movements.

Engineers also have to devise new networking schemes to juggle the billions of constant inputs that could easily swamp network bandwidth. This is Estrin's top priority at the James Reserve project. Ethernet and TCP/IP, the Internet's two fundamental protocols for shuttling data around, are poorly suited for low-power wireless sensor networks. The Net compensates for its habit of losing or garbling lots of data packets by constantly resending to the same address, leaning heavily on plugged-in servers at the edge of the network, fast Cisco (NasdaqNM:CSCO - News) routers and abundant bandwidth in between.

Estrin's scheme, still in rudimentary stages, calls for far less indirection to conserve communication power. Data packets are processed every step of the way along the network, not just at the edge. They zip around with no specific address, grouping themselves on the fly based on a time stamp, location or the nature of the data being sent: video, audio, chemical signature. And the network must operate only when necessary. Radios have to turn themselves off when they sit idle. "Even listening uses precious joules," she says.

One matter of little concern is money, thanks to the deflationary forces of Moore's Law. Airbag sensor chips cost only 60 cents each but combine an accelerometer with a microprocessor that figures out whether you are just slamming on the brakes or having an accident. They detect 15 types of information such as speed, deceleration, yaw, vibration and angle and then use algorithms to figure out what it all means. To detect and process all this information using traditional instruments would cost over a thousand dollars.

Once costs decline far enough, sensors will move from commercial and military applications to deep inside your home. The TV, light switches, shower heads and air conditioners could automatically respond to a person's wishes and location. In five to ten years, "people will no longer need to carry ID, keys or money, because sensors will recognize their face," notes Tatsuro Ichihara, a vice president at Japan's Omron , one of the world's leading sensormakers.

The hitch: These always-on networks are a snooper's dream come true, putting each citizen in the same situation as that squirrel in the James Reserve. "There would be a whole suite of sensors in a room that could tell you how many people are there and what they are saying. The implications of this are so huge we need to get sociologists and legal people thinking about this early," says Roger T. Howe, director of the Sensor & Actuator Center at UC Berkeley.