Sensors on the March

Computerworld - They're everywhere. Tiny wireless microelectromechanical sensors—also known as "smart dust" or "motes"—are monitoring temperature, humidity, stress and motion in settings as diverse as crop fields, bridges, factories, warships and the branches of Northern California's mighty redwood trees.
Now, imagine these so-called MEMS implanted in your body, periodically sending joint-load alerts to the orthopedic surgeon who performed your knee or hip replacement. Given extremely rapid advances in the intelligence and flexibility of sensor-based microcontrollers, such "smart implants" aren't all that far-fetched, experts say.
But up until the past year or so, such a scenario was implausible because of limitations in both the power supply and the programmability of most sensors and sensor networks. Sensors normally produce an overwhelming flood of data in a constant stream that steadily depletes their battery power.
With funding from the U.S. Navy, Williston, Vt.-based MicroStrain Inc. is experimenting with piezoelectric materials, which generate electricity as they undergo stress. This way, sensors could collect the power they need from vibrations on a factory floor or from the movement of the person they're implanted in.
Researchers are also fine-tuning software so that sensors deliver summary information, such as alerts or alarms, rather than a steady stream of raw data. This also conserves power.
At Palo Alto Research Center Inc. in California, a team led by principal scientist Feng Zhao is experimenting with an energy-saving "information-driven sensor-querying" algorithm, which would enable sensors to autonomously task themselves to collect and transmit information based on the usefulness of the information.
"It's quite similar to the way humans track information," Zhao says. "You can't pay attention to all stimuli. What we're building is distributed attention for sensor networks. It's the ability to shift and focus attention when new stimuli of interest emerge."
Meanwhile, working with researchers at the University of California, Berkeley, Intel Corp. has created an open-source operating system called TinyOS, which, among other things, enables sensors and sensor networks to report summaries of data or various classifications of information.
"TinyOS renders sensors into programmable routers. You can program what happens close to the sensor and what happens on the network," says David Culler, a Berkeley computer science professor.
This real-time information is then stored in TinyDB, which "can make the physical world like a database," Culler says. "Rather than issue SQL queries to get information out of a database, you issue queries to data streaming from the real world. If you look at the power industry, there's equipment throughout the country that's quite aged, and the ability to watch that