Sensor Technology - Sensor Networks

Imagine a network that links itself together autonomously, reacts to its environment and transfers information at lightning speed. What sounds like something out of a science fiction novel actually exists. Specifically, it is a self-organizing sensor network, consisting of sensor components called nodes that can independently determine their location, communicate wirelessly and create a network without any outside support. To accomplish this, each node is equipped with modules for location positioning, communication, data-processing and power supply units.

The first such sensor networks are still relatively simple. They measure parameters such as gas concentration, acoustic signals, temperature, brightness, humidity and acceleration. Using these measurements, it is possible to detect forest fires, the damage caused by an earthquake or the amount of dangerous chemicals present at a production facility (see Pictures of the Future, Spring 2003, Sensor Networks). Five years ago, scientists at DARPA (Defense Advanced Research Projects Agency) in the U.S. began the Smart Dust project. The objective was to enable thousands of miniature sensor units to observe enemy troop movements without being detected. Acoustic, magnetic and seismographic sensors would enable the units to register troop and vehicle movements.

The sensors would then filter the raw data and only forward relevant information. With this in mind, researchers at the University of California at Berkeley have developed sensor nodes measuring just a few cubic millimeters (see box). These nodes have to communicate via targeted laser beams because the components for radio communications are still too large and use too much energy for such an application. In contrast, in the Scatterweb project the sensor nodes, although larger, can already communicate wirelessly and configure themselves independently for the most part. "Our Scatterweb can be used for a range of applications, as the nodes can register light, vibrations, temperature, air pressure, motion and other parameters," says Jochen Schiller, Professor of Technical Computer Science at the Free University of Berlin. "Today’s sensor networks display great differences in terms of their ability to network themselves autonomously, integrate into systems such as the Internet or Ethernet and achieve the necessary level of flexibility should reprogramming be required."

Scatterweb sensors have different ranges depending on the data they are transmitting: Neighboring nodes can be recognized via radio up to a distance of four kilometers; movements can be registered up to ten meters away. Marine biologists at the University of Umeå in northern Sweden are using Scatterweb technology to monitor the Baltic Sea. The sensor nodes, which are installed in buoys, measure temperatures at different depths. A so-called multi-hop technique is used to conserve the energy needed for transmission. Here, the signals are routed from buoy to buoy until they reach land.

Whereas the Scatterweb project is primarily designed to provide a development platform, Siemens Corporate Technology (CT) is focusing on self-organization solutions that enable sensor nodes to set up communication networks on their own. Using special local positioning radar technology, each sensor would measure the distance to its neighbors and thus determine its own position. The sensors also need to be able to find out for themselves where they can send their data. They must also organize data processing operations autonomously and be able to interpolate so as to forecast data in the future. "It’s already possible for them to make a spatial forecast," says Dr. Rudolf Sollacher, head of Neural Data Processing at CT and an expert on self-organizing sensor networks. For example, a sensor node in a building can also estimate a temperature profile for those areas where no sensors are located.

Fireman in a Sensor Network. Siemens researchers plan to present largely self-organizing sensor networks in November 2004. In this particular scenario (see illustration above) a fireman with a display inside his helmet and a portable PC enters a burning house. The PC is equipped with an integrated sensor node with an Internet interface. Numerous sensors located in the building transmit data via radio over distances of 30 to 100 m. The sensor network displays the temperature in his immediate area and guides him to the fire step by step. The display on the helmet could later be replaced by 3D headsets that display information about smoke concentration or the presence of toxic gases, for example.

Today, sensors are already registering data on temperature, motion, brightness and noise in buildings. Siemens Building Technologies recently launched the world’s first security system with bi-directional radio communication on the market. In addition to smoke detectors, it consists of devices that recognize when glass has been broken, motion detectors, door contacts and a module for controlling lights, shades and other equipment. The SiRoute radio units can also independently find a way to get their data to the central control office via other components in the event of a radio disturbance, sabotage or if distances are too great. Because inactive elements are automatically put into an energy-saving "rest mode," the batteries last up to four years. The entire system is operated by a remote control unit the size of a credit card. Plans call for similar sensor networks to autonomously register tension and cracks in the materials of buildings and tunnels, and report this information to maintenance teams.

Sensor networks might also be able to facilitate the monitoring of patients. Intel and the Alzheimer’s Association plan to develop a wireless sensor network that observes the behavior of Alzheimer patients and sounds an alarm should anything unusual happen. The network will be able to register the patient’s location and remind him or her to take their medicine. Sensors distributed throughout hospital rooms and on the body could also be used to monitor pulse and temperature.

No one knows whether sensor networks will also be placed inside human bodies in the foreseeable future. Sensors that can be swallowed do in fact exist today. These are used to measure temperature or provide color images of the digestive tract. The biggest technical problem here, according to Sollacher, is not miniaturization but involves communication between the sensors, since they need an autarkic energy supply, especially if the sensor network is to remain in the body for a long time. Sollacher can imagine using "passive elements that obtain the energy they need from outside sources or from the body itself."

If self-organizing sensor networks are to be employed on a massive scale, the costs of the nodes, their energy requirements and size will have to be reduced. A Scatterweb node today costs around 50 €, for example. "What we need are sensors the size of a matchbox that cost 20 €," says Schiller. Such a reduction in price presupposes production in large lots, he says. Mass production would cut costs even more. As far as power is concerned, the sensors of the future will have to use ambient energy, since a battery change will be impossible. "Solar cells are one option," says Schiller, "but the nodes could also exploit temperature differences or vibrations." The main thing is that only those sensor nodes that have enough energy transmit data—for example, only those that are exposed to sunshine at a given moment.

"Communication should be minimized as much as possible, and computers’ built-in intelligence has to be increased, since both measures decrease the energy requirement," says Sollacher, who is convinced that the biggest gains in energy efficiency can be achieved with the data processing unit. Experts are also expecting smaller nodes to appear. Nevertheless, grains of sand that emit radio signals, as envisioned by the Smart Dust concept, will remain science fiction—at least for the time being.

Sylvia Trage

The Free University of Berlin has developed a miniature Internet in which data from individual sensors is collected and sent out over a sensor network. Dubbed "Scatterweb," the sensor network was made public at this year’s Hanover Fair. Any Web browser can be used to provide access to Scatterweb. The network is very flexible and freely programmable even when in operation. A variety of sensor nodes were developed. An embedded sensor board (ESB) is pictured above. ESBs are small devices whose surfaces measure 4 × 5 cm and contain numerous sensors for parameters such as brightness, noise levels, vibrations and motion. A microphone, speaker and infrared transmitter/receiver are also integrated into the units. An ESB "at rest" only requires 8 µA of power, while an active one consumes between 8 and 12 mA. When used with a conventional AAA battery, an ESB will last between five and 17 years, assuming it sends 25 bytes of data every 20 seconds. ( www.scatterweb.net)

The EU projects "Eyes" and "Bison": The "Eyes" project (2002 – 2005) focuses on energy efficiency. "Bison" (2003 – 2005) deals with biologically modeled sensor networks. Here, the focus is on robustness, self-organization and self-repair.

Great Duck Island: On this island off the coast of Maine, the College of the Atlantic (COA), Intel and the University of California at Berkeley are using a sensor network to observe a rare type of storm swallow. The sensors measure temperature, humidity and air pressure in the nests and surroundings. A sensor base station driven by batteries that last up to a year is connected to the Internet. The second network generation, containing 105 nodes, was installed in summer 2003, then again 60 hatching nest sensors and 25 weather sensors were added.

Sensor networks at UC Berkeley: Researchers at the University of California at Berkeley are striving to pack digital circuits, laser-based wireless communications and MEMS (micro-electro-mechanical systems) into one tiny system. Plans call for a complete sensor node—consisting of a microcontroller, storage unit, sensor, radio transceiver and power supply—to be integrated into a volume of 1 or 2 mm³. Whereas the old Flashy Dust Mote model was 138 mm³, the new type of Golem Dust Mote is 11 mm³ and 5 mm long.