Sensors Everywhere

HP isn't the only organization trying to couple RFID and motes. At Edwards Air Force Base in Southern California, NASA is preparing to use smart sensors and RFID tags to monitor hazardous chemicals (see "RFID Lets NASA Monitor Hazardous Materials," Jan. 10, 2005).

Most pilot tests so far have been modest, but companies including Boeing, Chevron-Texaco, Honeywell, Motorola, and Siemens all are exploring the technology. In December, Japanese tech conglomerate Fujitsu Ltd. disclosed a research agreement with Xerox subsidiary Palo Alto Research Center to explore equipping buildings with networked earthquake sensors, outfitting cars with wireless sensors to avoid collisions, and more. At a summit of world leaders in Jakarta, Indonesia, following December's tsunami, plans for building a network of sensors in the Indian Ocean to warn of undersea earthquakes were at the agenda's fore. But the fast money in the next few years probably will be made delivering wireless sensor nets to big companies as cheaper replacements for routine maintenance and monitoring of operations, such as controlling lighting or providing security around a building or at a border.

Intel, for example, has outfitted an Oregon chip-fabrication plant with 200 wireless sensors that constantly monitor vibration levels on some of the factory equipment and report when a measurement falls out of spec. The effort covers only a fraction of the plant's 4,000 measurable parts but has replaced some rounds by a technician who gets to each machine only every two or three months, Intel Research associate director Hans Mulder says.

General Electric Co. this month completed a test of sensor-outfitted shipping containers that can detect tampering, and it's developing products that could use mesh networks to secure apartment buildings and industrial areas. And Bechtel Group Inc., the largest U.S. engineering and construction company, may within a year or two start testing sensor nets that use a new standard, IEEE 802.15.4, that lets motes self-assemble into a network without programmers specifying what route the data takes. Bechtel has built wireless sensors into projects such as London's subway system and expects the technology to have applications in smart buildings, defense contracts, and chemical plants, infrastructure architecture manager Fred Wettling says. "We see this just starting to take off."

By 2008, there could be 100 million wireless sensors in use, up from about 200,000 today, market-research company Harbor Research says. The worldwide market for wireless sensors, it says, will grow from $100 million this year to more than $1 billion by 2009.

If the technology is to lead to new applications that can open up new markets, sensors' data has to be readily consumed by widely used business software. More industry standards are needed so software vendors have common ways of pulling sensor data from networks that contain sensor nodes of varying intelligence, made by different manufacturers. Researchers also are working on embedding software into sensors to make them more selective about what data is transmitted back to base or to condense information to conserve even more power. Without that capability, large sensor networks could quickly overwhelm themselves, and back-end computers, by draining bandwidth and battery power trying to transmit a flood of data from the field.

But there will still be plenty of data to analyze, and big tech vendors and consultants are going after the emerging market for computer systems and software to do just that. HP, IBM, Intel, Microsoft, SAP, and Sun Microsystems all have recently formed research or product groups to refine and commercialize the technology. For example, SAP is working with Intel and other companies to make sure its applications can consume and analyze the sensor data. SAP and BP also are participants in a European Union-funded project scheduled for this year to build "smart items" such as chemical barrels that broadcast warnings when they're inappropriately stored.

The tech giants already cede the mote market to companies such as Dust, which last year landed funding from the CIA's venture-capital arm, In-Q-Tel. Ember Corp., a startup based on work at MIT, has attracted an investment from Ethernet inventor and 3Com Corp. founder Bob Metcalfe. And Crossbow Technology Inc. has been selling to BP and supplying Intel's projects. "We're not competing with startups," HP's Pradhan says. The big IT vendors want to supply huge volumes of chips for motes, then sell installation services, but their primary goals are to develop new applications for sensor nets and sell software and consulting services. One big problem to solve: the lack of software tools that can program whole networks of sensors in one shot.

"For every dollar the big systems integrators and IBM make on sensors and installation, there's $10 to be made on the management of the data that comes out," says Kris Pister, a professor of electrical engineering and computer science at UC Berkeley, who founded Dust in 2002 and serves as its CTO. IBM, which plans to spend $250 million during the next five years on the technology and has created a "sensors and actuators" business unit, predicts wireless sensor nets could represent a $6 billion overall market by 2007, with the bulk of profits from software that helps the devices better communicate and lets customers interpret data from them.

"Sensors are just a part of an ecosystem of wireless devices," says Feng Zhao, a senior researcher at Microsoft who joined the company last year from PARC to head up a new sensor nets research group on Microsoft's Redmond, Wash., campus. His test bed is parking level P2 of building No. 112, where a handful of sensors detect the size, speed, and magnetism of everything that crosses the garage's threshold, triangulating data from video images and magnetic readings of staffers' cars. At a remote PC, a researcher can analyze the day's traffic by logging on to a Web site and posing queries using standard Web-programming techniques. It works in a restricted scenario and with research prototypes, Zhao says, but "we need to figure out how to organize these systems and develop interesting applications for them" for real-world use. "For all these apps, writing software is very challenging. That will probably be a stumbling block between sensors and killer apps."

"It's kind of like the beginning of the Arpanet days for this sensor-net technology, where there's no killer app yet," says Teresa Lunt, manager of the computer-science lab at PARC. A PARC research project called "smart matter" aims to embed sensors in the environment, and the center has done experiments with Darpa funding, including using sensor nets to track a mock military tank based on its signature sounds. At current prices, though, minus the sensors attached to them, wireless motes are still impractical for most large networks, Lunt says. "But they've served as a placeholder people can use to envision applications with the understanding that they'll be replaced by better technology," she says. "They've been igniting people's imaginations."

Sensor proponents predict a day when superhighways will be salted with motes that help drivers avoid collisions, bridges report when they're seismically stressed, and networks of video cameras pick terrorists out of a crowd. That's a long way from turning down the air conditioning when it gets too hot.

Illustration by Eboy