Sensor Sensibility

Monday, 07 May 2007

Imagine a future in which billions of tiny computers are embedded into buildings, streets, fields, or even our bodies. These devices might monitor weather, traffic, crop conditions, the progression of diseases, or a host of other variables. The tiny computerized sensors would spontaneously organize into networks that could adjust their structures and functions in response to the information that they pick up. That future might be just around the corner. Researchers have already deployed networks with dozens of matchbox-size sensors in a wide range of applications. Sensor networks are tracking the movement of zebras in Kenya and determining bullet trajectories in military field tests. Coming soon, many engineers predict, are cheap sensors the size of dust particles. Those high-tech specks will measure temperature, vibration, noise, light, and more. The question, the engineers say, is not whether these smart-dust sensors will someday pervade our environment, but when. For smart dust to be useful, however, engineers must figure out how to build a global view from the information provided by millions or billions of individual sensors. For example, suppose that agricultural researchers scatter a million battery-powered, smart-dust sensors by helicopter to monitor water levels across a cornfield. Without knowing where each sensor has landed, how would the researchers determine whether the sensors' combined range leaves gaps? Or imagine that engineers have deployed a sensor network to keep track of boats in a harbor. If each sensor reports how many boats it detects, how can the engineers keep an accurate tally without knowing how many sensors have counted the same boat? To tackle these questions and others, researchers are drawing on techniques from topology, the study of shapes. Analyzed by mathematicians for more than a century, topology has until recently had few real-world applications. Yet topology, which pieces together the global structure of a space from local snapshots, is exactly what sensor-network engineers need, argues Robert Ghrist, a mathematician at the University of Illinois at Urbana-Champaign. "Topology is good for finding hidden features inside a space that you can't see very well, that you don't have all the information about," Ghrist says. "Figuring out the structure of wireless sensor networks is the kind of problem topology was meant to solve." Erica Klarreich

 

< Prev		Next >