http://www.eet.com/at/news/OEG20030128S0028
Companies test prototype wireless-sensor nets By R. Colin Johnson EE Times January 29, 2003 (1:34 p.m. EST) PORTLAND, Ore. � Self-organizing wireless-sensor networks, a realization of the Pentagon's "smart-dust" concept, have reached the prototype stage worldwide. The smart sensors, or Motes, were created by the University of California at Berkeley and Intel, and are being tested out worldwide today. "At this stage, there are over 100 groups around the world that are using the combination of our open-source Motes with the TinyOS [operating system] and TinyDB [database]," said Berkeley professor David Culler, who is also director of Intel Research's "lablette" in Berkeley. Researchers at the Defense Advanced Research Projects Agency (Darpa) proposed the smart-dust concept four years ago. The idea was to sprinkle thousands of tiny wireless sensors on a battlefield to monitor enemy movements without alerting the enemy to their presence. By self-organizing into a sensor network, smart dust would filter raw data for relevance before relaying only the important findings to central command. The prototype Motes consist of an application-specific sensor array board married to a generic wireless controller board, both in a hermetically sealed enclosure. Once the design has matured, single-chip realizations will begin to downsize the wireless sensors to a volume less than a cubic millimeter. To facilitate the self-organizing of Motes into a sensor network, the researchers created TinyOS and TinyDB as well as a host of Tiny applications and a simulator. "In a sensor net you've got many tiny devices deployed in the world, and they're connected as a network. TinyOS is the framework for building up the operating system capabilities needed for the sensor network � the networking capabilities, localization and support for applications. TinyDB then aggregates the data at the next layer up," said Culler. Since Culler's team and Intel have offered the Mote hardware and the TinyOS and TinyDB software as open source, researchers worldwide have jumped on the bandwagon to begin building civilian and military applications of smart-sensor networks. And a few are attempting to downsize the boards with custom chips that approach the smart-dust vision in terms of size. "Many companies are now in the demonstration stage, and there are other companies looking at driving down the size of the Motes, which may end up being little nuggets that have the sensor, the radio, the processing and the power storage all integrated into incredibly small devices," said Culler. "Rather than put more and more stuff on one chip, they're asking, 'What's the smallest complete computing device we can make?' " In addition to the high-profile military application, wireless-sensor networks could be put to legions of civilian uses, from environmental monitoring to providing heath care monitoring for the elderly while allowing them more freedom to move about. "This is a new class of computer," said Culler, who predicted that in 10 years, the emerging compute class "will be well-established, and we'll have a whole new set of challenges." To see around corners Darpa has historically funded many application areas for military purposes that have ended up having even wider civilian use, the Internet being the most famous example. Wireless-sensor networks were likewise originally a military concept for the battlefield of the future, enabling soldiers to "see around corners" and to sense the threat of chemical and biological weapons long before they get close enough to cause harm. When the project began about four years ago, Culler said, "people were putting a sensor or two on a few laptops and calling it a sensor network � either that, or they were just running simulations. The view was, 'If only there was a widely available platform, then this whole research area could really take off.' " Meanwhile, Darpa smart-dust prototypes used off-the-shelf components and commercial radio transmitters. "I looked at their prototype and said, 'That could be the starting point of the platform we need,' " Culler said. "So we redeveloped the architecture to make it modular with separate boards for sensors and power, and built TinyOS around it. That was the point at which we formed the collaboration with Intel." Darpa supplied the funding "to produce an open-source embedded platform in this wireless space, called the Network Embedded Systems Technology Program, and Intel offered its expertise too." UC-Berkeley then contracted with Crossbow Technology Inc. to build several hundred Motes for testing. Although the hardware design was made public, Crossbow is still allowed to manufacture it for anybody. "Intel licensed some of its networking technology to Crossbow, so that now lots of researchers in industry labs and universities can not only pick up the open-source OS and database from us, but they can also pick up the hardware from Crossbow to quickly put their whole kit together," said Culler. Since Motes are so tiny � usually with memory measured in mere hundreds or thousands of bytes � there was also a great need for a tiny operating system. TinyOS fit the bill by providing a set of modular software building blocks. Designers choose the components they need, and file sizes are as small as 200 bytes. The secret to getting a network to self-organize, said Culler, is to rely on local rules. When these run on many Motes simultaneously, the result is a hierarchically arranged network. In self-organization, "you're making decisions that are based on local rules. After deploying thousands of sensor Motes and hundreds of gateway Motes throughout the environment of interest, a simple rule might be: 'Establish a routing path to the nearest gateway.' Groups of sensor Motes would then form trees to the nearest gateway Mote," yielding "a hierarchical organization popping out right from the beginning," said Culler. Since much of the value of such networks comes from having many sensors sensing the same thing, the local rules running on each Mote may be different in a manner that is transparent to higher levels in the hierarchy. "You may have sensors that run on energy they have harvested, such as solar or vibrational, and their actions may be based on how much energy reserves they have," Culler said. "So a more complex rule could take into account the energy capacity of a Mote and the loss of links that have powered down. These and similar rules are sometimes called quality-of-service routing, where you have various factors going into your decision-making about how to self-organize the network." The modular, component-based approach is used at all levels in this self-organizing sensor network. For instance, the hardware of the Motes separates the application-specific aspects from the generic aspects. In particular, the sensor, the energy source and the physical packaging can all be separated, so they can be specialized for particular applications. The common parts � processor, radio and power distribution system � are kept on the generic board. "The operating system itself is also component-based," Culler said. "These things are very small, and you want them to be very robust, so again you only want to take what you need. Essentially, you are building an application-specific operating system just by picking and choosing from the various components that are available." Only when necessary TinyDB is also modular, but in addition it interfaces at a higher level, appearing to the OS as if it were an external application layer. Because its job is to aggregate data, however, instead of relaying raw data to its gateway TinyDB only communicates to the gateway when absolutely necessary. "Part of the power you get by having a little computer next to every sensor, rather than just a dumb instrument with a wire back to a big computer, is that you can do a lot of the processing down inside the network," Culler said. "That is very important, because you may be sensing a tremendous amount of information but you really only want to pass along the most important information. TinyDB does this in-network query processing, which is largely application-independent, making it very easy to do a whole range of data collection operations."
