It's a midsummer evening off the foggy coast of Maine. A small, black bird called a storm petrel swoops into an underground nest. Instantly, a tiny computer inside the burrow registers the bird's arrival and sends an alert through a wireless network. Throughout Great Duck Island, 200 similar computers linked together are tracking the weather, winds, and movements of these enigmatic birds, which for mysterious reasons prefer this island to all others. Naturally, the project, sponsored by Intel Corp., interests ornithologists. But it's far more important to the computing world: Duck Island is a laboratory for a new stage of computing known as sensor networks.
Take a look at the computer under that nest, and you would think technology was marching backward. It has minimal computing power and is capable of only two or three simple jobs, such as detecting heat given off by the bird's body and measuring weather temperatures. Primitive stuff. Yet a far-flung galaxy of these simple machines, each equipped with eyes or ears, chemical, motion, or temperature detectors, is anything but simple. The nodes pass information from one neighbor to the next, as though skipping across lily pads in a pond. And at a control center, a far more sophisticated computer harvests all this data and makes sense of it.
Already, companies from British supermarket Tesco PLC to Shell Oil Co. have deployed first-generation systems to monitor inventories and check the status of pumps at gas stations. That's just the beginning. Within five years, these sensor computers could be shrunk to the size of a grain of sand and deployed over much of the globe, resulting in thousands of new networks. Look for them to be scattered across farms and battlefields to monitor minute chemical and temperature changes and slapped onto trucks and shipping boxes to trace inventory automatically.
Sensor networks promise a mammoth extension of the Internet. To date, the Web has been a showcase for the human brain. It specializes in the words, numbers, music, and images that mankind produces. With sensors, the network stretches to the far vaster field of global activity. This means such networks can cover every single thing that moves, grows, makes noise, or heats up. Potentially, much of the world will be bugged. Moreover, those bugs will be doing most of the work. "Most of the data traffic won't be between human beings this time around but between these silicon cockroaches," says Bob Metcalfe, the networking pioneer who has invested in Ember Corp., a sensor-network startup in Boston.
The sensor revolution has its beginnings in two basic forms. Retailers are experimenting with the simple variation, a so-called radio frequency ID tag, or RFID -- a next-generation bar code. Take a radio reader into a warehouse packed with RFID-tagged goods, and each item promptly identifies itself, all of them joining to create an up-to-date inventory. These tags cost 20 cents but could cost only pennies within two to three years. Far more sophisticated sensors now cost $50. They create their own networks, like the one on Duck Island, and perform trickier jobs, such as running computations.
These diminutive sensors already are stirring up fears of government and corporate intrusion. Billions of tiny tags could track everything from people's shopping patterns to movements of suspected criminals. Privacy concerns will likely grow as sensors make their way into payment systems, are placed inside tires as ID devices, and are tacked onto traffic lights and signs. In June, privacy groups led protests over planned tests of RFID tags in a Wal-Mart store. The retail powerhouse canceled the project, saying it wanted to focus first on a plan of getting tags on most of its shipping cases by 2007. "There hasn't been an effort by the industry to explain to any Joe Schmoe what this technology can and can't do," says analyst David Krebs of researcher Venture Development Corp.
Companies are eager to overcome such obstacles -- and cash in on the huge potential of sensor networks. York International (YRK ) Corp., which manages ventilation systems for more than 60,000 customers, plans over the next five years to install hundreds of thousands of networked sensors on its clients' air-conditioning units. These will monitor temperatures and automatically send updates to York's offices. That should lighten the workload for York's 2,000 technicians, boosting productivity by as much as 15%.
For all the promise, plenty of work remains. One key is to make sensors with an energy supply that will last for years. This means tapping solar or kinetic energy, or tweaking systems so that they draw only tiny bits of battery energy. Another challenge is to hammer out industrywide standards for sensor hardware and software.
One emerging standard may be TinyOS, an open-source operating system developed at the University of California at Berkeley. Backed by Intel and a host of universities and startups, TinyOS works on fewer than 8 kilobytes of memory -- the size of a short e-mail. "It's extremely flexible, and it's open-source, which is important with the market still in a chaos phase," says Mike Horton, CEO of Crossbow, which has used TinyOS in test networks to track car parts.
For now, sensors are too expensive for most massive rollouts. It will take technical improvements and volume orders, such as those expected from Wal-Mart Stores Inc. and its suppliers, to drive down pricing. But the U.S. military already is laying out plans. In tests, it has dropped tubes containing 10 to 20 sensors from an airplane. Once they hit the ground, the tubes explode, sending the sensors flying. Then they establish connections among themselves and monitor enemy vehicle or troop movements, sending the data wirelessly to planes overhead.
Other early applications are more mundane. Tyco Thermal Controls LLC is turning to wireless sensors to cut down on the cost of laying wires in its pipe-heating systems. Such wiring accounts for two-thirds of the expense of installing monitoring systems. During the next three months, Tyco plans to launch customer tests.
As sensor systems grow more sophisticated, researchers hope to dispatch them on more complicated missions -- and into our everyday lives. Intel is experimenting with systems in health-care centers to help patients with faulty memories -- reminding them to eat a meal or drink water. No doubt, wonders like these in the home are still 7 to 10 years off. Still, the sensor revolution is off and running. While the last 50 years have been dominated by a march to ever more complex computers, the next few decades will see the rise of simple sensors -- by the billions.
By Heather Green