cartome.org

1 May 2002

Economist, 2/2/2002, Vol. 362 Issue 8258, Special section p8, 2p, 1 graph

DESIRABLE DUST

How smart sensors can monitor the real world

WHEN peripatetic futurologists such as John Gage of Sun Microsystems really want to impress their audience, they talk about "smart dust". The concept is indeed intriguing. The dust in question is made up of tiny, wireless sensors that could be dispersed anywhere--say, over a battlefield to find out where enemy troops are, or whether chemical or biological weapons have been used.

This is not science fiction. Researchers at the University of California at Berkeley led by Kris Pister, a professor of electrical engineering, are already working on a smart-dust prototype the size of a small nailhead. A klunkier version can already be bought from a Silicon Valley start-up, Crossbow Technology.

It may be decades before smart dust is dispersed over real battlefields, but the tiny devices are at the forefront of an important technological trend that is often underestimated: the spread of sensors and tags. New manufacturing processes, wireless technology and intelligent software are making them ever smaller, smarter and, most important, cheaper. As with microprocessors and lasers in earlier decades, the novelty is not that these sensors exist at all, but that they have suddenly become cheap enough to be used in ordinary everyday products, says Paul Saffo, director of the Institute for the Future, a Silicon Valley think-tank.

Electronic nose, anyone?

Perhaps to an even greater extent than better software, wireless sensors and tags play a big part in making the real-time economy happen. Whether they are tiny thermometers, miniature microphones, electronic noses, location detectors or motion sensors, all of them provide ever more information about the physical world, meaning that firms can add ever more cells to their corporate spreadsheets. Small wonder that forecasters expect the technology to become a huge business. The global market for sensors alone will reach more than $50 billion in 2008, according to Intechno Consulting, a consultancy in Basle, Switzerland.

Sensors are nothing new in expensive machinery, where an unexpected breakdown can be costly. What makes them so much more useful is that they are increasingly connected. General Electric has long been equipping some of its aircraft engines, turbines and locomotives with all kinds of sensors, but until recently the data were not available in real time. Now all the information is regularly transmitted via satellite to a GE remote-monitoring centre. If there is something wrong with a jet engine, for instance, the facility identifies the likely cause and tells the airline about it.

Remote monitoring also allows classic old-economy companies to deliver high-tech services. For example, SKF, a Swedish bearing manufacturer, uses sensors to measure the vibrations of machinery at strategic positions and feed the data into analysis software that can determine when a bearing is about to fail. Users can then arrange for its replacement. This is particularly useful in manufacturing processes where downtime is expensive, such as in car factories or paper mills.

Monitoring systems are also appearing in more everyday settings, for example in ice-dispensing machines in supermarkets. Sensors in "Ice Factories", operated by Dallas-based Packaged Ice, are able to obtain real-time data on dozens of conditions. Is the ice at the right temperature? Are the bags sealed correctly? Has a light bulb burned out? If something has gone wrong, the machines automatically alert the company so it can send someone round to solve the problem.

Similar systems are set to become ordinary parts of cars, air conditioners and household appliances. One firm that is working on this is graviton, a start-up backed by a group of investors including the CIA's venture-capital fund, In-Q-Tel, and Siemens, a German electronics giant. To build ever cheaper sensors, graviton uses technology known as MEMS (short for microelectromechanical systems), devices built with techniques developed in semiconductor manufacturing.

The tiny sensors are, in essence, microchips that convert analogue data about anything physical--pressure, light, gas, genes--into bits and bytes which they communicate wirelessly to a network. Graviton has yet to prove that its technology can live up to its promises, but if it does, some of its investors will certainly become customers. Royal Dutch/Shell is reportedly studying how to use graviton's sensors to monitor its oil refineries and pipelines.

We know where you are

Tagging technology is getting cheaper as well. Today it is mostly used in logistics and manufacturing. WhereNet has developed a system of matchbox-sized wireless tags and readers that allow objects to be located within about 3m (ten feet), making it much easier to keep track of them. American Airlines has recently installed the system in its huge cargo facility at Dallas-Fort Worth airport. Ford has been using it for some time so workers on assembly lines can call for more parts and to track new cars in its yards.

Consumers will also get the benefit of using wireless tags, albeit for the moment only the cheaper "passive" ones, which must be activated by a magnetic field generated by a reader. In a few years, smart labels called radio frequency identification tags are likely to replace today's ubiquitous bar codes so that groceries, for instance, will no longer have to be scanned in individually. Shoppers will just push their carts through a reading station. Likewise, the refrigerator could one day know when the beer supply is running low and re-order automatically.

However, all these fancy sensors and tags need software to make sense of the data they deliver. Just knowing that a piece of machinery is running hot does not help much. GE uses sophisticated statistical methods and historical data to decide whether it is a clogged fuel filter or just bad weather that has reduced a locomotive's horsepower. Similar tools tell a railway company how many of its locomotives need servicing at any one time so that it can schedule the work at its maintenance centres.

All these sensors will generate a phenomenal quantity of data, raising the spectre of a huge information overload. One solution is to make the sensors smart enough to communicate with each other and process much of the information automatically. Mr Pister and his Berkeley colleagues have developed an operating system for their smart-dust motes that lets them form wireless networks without human intervention.

Such self-organising technology might one day make another futurologists' dream come true. Sensors will combine their skills with effectors, tiny devices that can manipulate matter, making it possible to create "smartifacts"--smart materials and in telligent artifacts. One example often mentioned is turbulence-reducing aeroplane wings covered in billions of silicon microflaps and tiny wind sensors. Another is bridges made earthquake-safe using arrays of effectors and sensors. These sound highly desirable, but other potential uses of sensors are less benign.