In the late 1990s, the rapidly dropping costs of microprocessors, of wireless network interfaces, and of sensors led some researchers to propose a powerful vision: a vision that tiny, wirelessly connected, computerized sensors could be scattered about like grains of smart dust and that these 'motes' would self-organize into a network that would allow us to weave intelligence into the physical world.1,2 This would allow us to intelligently control a diverse array of physical systems, making them more efficient, less power hungry, and more responsive to human needs. For example, we could reduce the costs of heating and lighting a building, providing these services only to occupied rooms; could measure every tremor in an earthquake zone, predicting large quakes; or could let computers sense blood sugar levels and control insulin pumps, making life more pleasant for diabetics. Thousands of researchers were inspired by this vision to work on many aspects of these 'wireless sensor networks,' making this a rich field of scientific inquiry.
However, one aspect of wireless sensor networks has detracted us from realizing this powerful vision. This is the need to provide power to sensor motes. It has turned out that powering a sensor using batteries makes them large, expensive, and unwieldy. Instead of scattering them to the winds, they need to be very carefully sited, so that batteries can be replaced from time to time, and so that energy would not be wasted on expensive wireless packet transmissions. This has greatly reduced the scope of wireless sensor networks, making them more of a niche technology than one would otherwise expect.