Building a wireless nervous system

Could an environmentally aware network spot a forest fire in its early stages? Monitor a battlefield for the position of enemy troops? Detect damage to an oil pipeline, averting a disastrous leak?

Someday, wireless sensors may observe battlefields or assess damaged oil pipelines. In the near future, there's plenty of work to be done before they're even ready to manage an office building's air conditioning.

Dozens of universities, venture capitalists, technology start-ups and major corporations are working on an experimental technology that could give computer networks millions of tiny electronic feelers. But they must untangle some thorny technical and economic problems before the technology takes off.

Central to these wireless monitoring systems are special miniature sensors--"smart dust"--programmed to observe their environments and wirelessly relay information back to a central computer system. What sets these sensors apart from a standard sensor--such as a thermostat--is their ability to link to each other in intelligent clusters and collectively process and transmit data.

Each device, also called a "mote," contains a processor, a small bit of computer memory, and a low-power radio transceiver. These devices can also come packaged with their own sensing gear and power source--all miniaturized to the size of a bottle-cap.

The devices are expected to dwindle to the size of an aspirin or grain of rice over the next several years, at which point they could be dropped into waterways to detect pollutants or embedded into asphalt in roads to monitor traffic patterns. Imagine scattering thousands of these minute devices around buildings, bridges, factories and fields, giving people the power to observe the world on a finer scale than ever before.

"I do strongly believe this technology is going to be revolutionary and will get to the point that it has a huge impact on people's daily lives," said Rob Conant, vice president of business development at Dust Networks in Berkeley, Calif.

By the end of the decade, the Internet and other networks will sprout hundreds of millions of such devices, according to Harbor Research, which tracks the fledgling market. The market for equipment and services related to wireless sensor networks could be in excess of a billion dollars by that time, the research firm has forecast.

Dust Networks, whose founder coined the term smart dust, is one of a handful of start-ups working to make this vision a reality. Others include Crossbow Technology in San Jose, Calif., Ember in Boston, and Millennial Net in Cambridge, Mass.

Intel and Accenture are also active in the field, as are leading research universities including the University of California at Los Angeles, Carnegie Mellon University, the Massachusetts Institute of Technology and the University of California at Berkeley.

The U.S. government has taken an interest in the technology as well, with the Defense Department--through its Defense Advanced Research Project Agency (DARPA)--the National Science Foundation and the Central Intelligence Agency each funding efforts.

Small steps
Despite all the brainpower involved, wireless sensor networks have largely remained in trial phase for the past several years, due to cost and technical barriers. Dust, Crossbow and others are focused primarily on breaking into three markets that represent low-hanging fruit: defense, building automation and industrial equipment.

Outfitting sensors already in air conditioning, lighting and security systems with wireless networking gear could make buildings safer, more comfortable and less costly to build and maintain, says Dust Networks' Conant.

Building automation is ripe for the technology because it's already so sensor-intensive, he said. "Any sensor can become a wireless sensor by adding a mote," Conant said. "It's like taking a PC and turning it into a networked PC."

In defense, the name of the game is gathering intelligence about the battlefield, troops and equipment. The technology could also help asset-intensive industries--such as petroleum and chemical companies--monitor their facilities and detect damaged or worn-out equipment, such as a crack in a pipeline.

Tyco Thermal Controls, an industrial pipe heating company, hopes that wireless sensor networks will reduce the installation costs of its temperature-monitoring gear. The company hopes to replace expensive, maintenance-intensive wiring with a wireless system from Ember, which it plans to integrate into the equipment it installs in oil refineries and chemical plants.

"Because of the toxic environment, wiring gets very expensive," said Ken McCoy, general manager of the electronics unit of Tyco Thermal. "Wireless is a natural." The company is still testing the technology and has not incorporated it into any customer sites.

Big mistakes
The slow start to finding commercial success has already led to the early demise of one promising venture, San Diego-based Graviton.

The company, which raised more than US$60 million in venture capital over its five-year life span, counted Kleiner Perkins Caufield & Byers, the CIA's venture capital arm (called In-Q-Tel) and Royal Dutch/Shell Group among its prestigious investors. Its assets were sold last year to a company called Xsilogy for an undisclosed amount. But according to Graviton founder Michael Nova, the company fetched a measly US$1 million from its buyers.

Nova, who left the company in a management shakeup a year and a half before it was dismantled, said Graviton had the wrong business model. And, he warned, many of the players that remain in the space are grappling with business-model questions as well. In Graviton's case, the company tried to do too much, he said. It wanted to provide sensor devices, networking equipment, as well as installation, data processing and reporting services.

"Trying to sell a total solution was a mistake," Nova said. "It was too far out there in terms of what people wanted to do."

The companies that are still around, including Dust and Ember, have instead focused on the enabling equipment--sensors and the networking gear to connect them. In fact, Ember just makes the networking component and doesn't produce sensors at all.

Sweating the small stuff
Another set of hurdles to wider adoption of wireless sensor networks involves the cost, size and energy efficiency of motes. At US$50 to US$80 a piece, motes are too costly today for many of the applications its inventors envision. But researchers expect the price to fall to as low as US$1 in the next five years or so.

Many sensors run on batteries--which can create a maintenance hassle if users are expected to replace them frequently. Dust Networks, which relies on batteries to power its sensors, has come up with a design that keeps the devices running for five to 10 years before they run out of juice. Eventually, developers want to tap solar and kinetic energy to power the devices.

"Surely it could be better, and it will get better over time," Conant said.

Other gaps to fill include developing operating and network standards along with tools for filtering and analyzing the anticipated onslaught of sensor data.

The Institute of Electrical and Electronics Engineers, a professional association, has formed a working group to iron out wireless communication protocols for sensing networks, focused on the 802.15.4 specification.

Meanwhile, an open-source program called TinyOS, developed at the University of California, Berkeley, has emerged as a popular operating system for the devices.

Analysts are confident the obstacles can be overcome with time. And they're not deterring the hundreds of pilot projects already under way in the military and at companies that sell industrial equipment, including Honeywell and General Electric, according to Glen Allmendinger, president of Harbor Research. "There is clearly a groundswell of activity going on," he said.