A Wireless Sensor City

Continued from page 1

For now, says Welsh, his team has no plans to integrate video cameras into the network. However, he believes that the project is a great opportunity to explore the social ramifications of collecting data on an entire city. "CitySense is a good way to come face-to-face with the questions of what it means to outfit a city like this," he says.

In recent years, wireless-sensor networks have gained more prominence as a commercially viable technology thanks to companies such as Dust Networks and Arch Rock, both University of California, Berkeley, research spinoffs. The most common types of sensor networks are often found in industrial settings, where they monitor manufacturing equipment in hard-to-reach places. For the most part, these sensor networks employ tiny, battery-powered sensor "motes" designed to use little power and collect specific types of data.

Building a wireless backbone across an entire city that can accommodate numerous, simultaneous research problems will be tough, says Joshua Bers, a researcher who leads BBN's effort on the CitySense project. One challenge will be to make the network reliable enough by keeping the hardware from failing. "Clearly, if researchers are going to be using it, we don't want this thing to crash and have to have someone go up a light pole 30 feet off the ground to fix it," Bers says. There's one way to safeguard against this, he says: there will be software that will monitor the health of the node, and if something goes wrong, it will automatically reboot.

In addition, CitySense will use "mesh" networking to send data from one point in the network to another point far away. In such a networking scheme, information is transferred to its final destination by hopping from node to node instead of being transmitted directly. BBN has developed a number of mesh-networking protocols that adjust when nodes fail.

Another issue that will need to be addressed, says Welsh, is hardware resource allocations. Since numerous projects and various software applications will be running on the same hardware simultaneously, the researchers will need to find a way to make sure that the processors and memory are divvied up appropriately.