Sensing the Environment via Wi-Fi
May 11, 2007
Cambridge, Massachusetts is getting an open testbed of wireless sensors for use by all sorts of researchers, via the fledgling CitySense project.
When you think Wi-Fi and municipalities, you naturally think Internet access, but the City of Cambridge, Massachusetts (in partnership with its largest employer, Harvard University, and Cambridge-based government contractor and Internet pioneer BBN Technologies) has a different idea.
With funding of almost $1 million from the National Science Foundation and additional support from Microsoft, the group is building CitySense, a Wi-Fi-based sensor network that will eventually include 100 nodes around the city, many on light poles. Each node will initially include weather and/or pollution sensors. CitySense could support other types of sensors in the future and will be open for use by researchers and computer scientists from the Cambridge area and, theoretically, anywhere else in the world.
We want to make it an open testbed, says Matt Welsh, assistant professor of computer science at Harvard and one of the prime movers behind the project. Well allow researchers to get a log-in account and write software that runs on the nodes. Some may not even be interested in the sensor network they may want to test a new routing protocol, for example. Now they can do that with a simulation, using the City of Cambridge as a lab.
When Welsh and engineers at BBN first conceived the project, they saw it as a way to move sensor networking to a new and interesting level. Most sensor networks to date have involved placing battery-operated devices in remote locations such as forests and volcano sites. Because the batteries have to last a long time, computational and wireless networking power are constrained. By putting the sensor nodes on light poles and plugging them into the electrical grid a first, as far as the partners are aware those limitations disappear.
Each node will include a single-board PC (from Soekris Engineering) with RAM and flash memory storage, USB and Ethernet connections and two 802.11a/b/g radios (based on an Atheros chipset). One of the radios will function as a back channel, and one will be available for experiments. One will use a solid but not high-performance protocol to create the network. The other will be available for testing new protocols. Everything will be mounted in a weather-proof outdoor enclosure.
Four or five of the nodes will have wired connections to the Internet backbone. The rest will communicate via a mesh topology, likely using a protocol developed by BBN. Our intent is not to develop new wireless mesh technology, Welsh says. Well probably just pick up something already developed. None of the technology is written in stone, however, and may evolve as the project proceeds.
While scientists werent initially involved, Welsh had no trouble recruiting interested researchers, starting in his own backyard with the Harvard School of Public Health. The HSPS had been trying to track levels of airborne particles that can cause health problems. They wanted to know in what parts of the city the particles were most prevalent and what the factors were that determined that distribution. Proximity to a freeway, for example, with vehicles kicking up dust. Or perhaps wind velocity and direction, with prevailing winds blowing particles into certain parts of the city.
The HSPS researchers wanted to scientifically confirm such cause-and-effect hypotheses and then see how, if at all, that played into the incidence of related health problems in different parts of the city. They had tried working with backpack-mounted sensing devices, walking around Cambridge neighborhoods taking readings, but it wasnt terribly efficient or effective.
Its very, very time-consuming, Welsh notes. And because they have to do it manually, they only get readings from one point in time. They think it will be interesting to [put a sensor] on every light pole, that it will paint a very rich picture of whats going on with the pollution.
Other possibilities include mounting sensors to detect biochemical agents, or microphones or cameras to track road traffic. But thats challenging, Welsh notes. Because now youve got privacy concerns. You want to be very, very careful about how youre doing it.
In fact, one of the surprising concerns raised by the city was that, even without cameras, there would have to be a public education campaign so that citizens wouldnt wonder what the pole-mounted enclosures were for, or worry they were part of some surveillance activity. The city is also insisting the enclosures meet minimum aesthetic standards.
There are some practical limits on how many sensors can be mounted at each node, although they could conceivably be connected in a daisy-chain through the USB ports on the computers. But as Welsh points out, in a small city that covers an area about five miles by six miles, many types of sensors do not need to be at every node. Certainly, for weather tracking, its not necessary to have a sensor at all 100 nodes, though the airborne particle sensors need to be finer grained, he says.
The four-year project is midway through its first year. Welsh and BBN have set up about 10 experimental nodes at the university and at the companys facilities, some of them indoors. The partners expect to have between 30 and 40 nodes mounted outdoors by the end of the summer. They likely wont start mounting nodes on light poles until end of summer, because once the nodes are up on poles, it will be more difficult for the project partners to tinker with them.
And were sure to have to tinker with them because of the weather sensors, Welsh says. [If theyre on a pole], wed have to call the city and get an electrician to come out with a cherry picker to give us access. So we have to be careful.
All 100 nodes may not be in place until the fourth year. That will give the project participants plenty of time to refine the technology and ensure that they leave a good and long-lasting network in place at the end of the process. Meanwhile, there is still the possibility of expanding the network, to increase both node density and geographic coverage. The project team is talking to the city of Boston about placing nodes there as well.
The network could conceivably be used for public Internet access at some point, though that is not the main focus right now, Welsh says. Allowing the CitySense network to be used for public access may not be the best idea for a few reasons, he explains.
In a sensor network, its acceptable to have some delay, and that gives us a lot more flexibility in how we design the network, he says. We might want to send some of the data over multiple paths, for example, to get higher reliability but that could create greater delay times. And that wouldnt be so good for an Internet access network that has to send packets in real time. Cambridge has a separate Wi-Fi public access initiative, and there may be some sharing of infrastructure between the two projects.
Our first priority is we want to make sure the network is really useful for scientists who fundamentally will want to break it, Welsh says. If you give a researcher the opportunity to try out a new routing protocol, hes probably going to break something. We dont want to have to tell them, no, they cant do that [because we need to keep the network up for Internet access].
The project still has a long way to go before it can count itself a success, and Welsh admits it could face challenges. The wireless routing protocols may turn out to be not reliable enough, requiring the project team to redesign them. While the hardware will be in weather-resistant enclosures, it still may not be robust enough. It gets very cold here in winter, he notes. And because the nodes are not full-fledged PCs and have limited memory and storage, they may not have the power to accommodate all the research people want to do.
Still, thats not a failure, just a limitation. And attracting too many researchers to the project will be a nice problem to have, Welsh says. A good sign of success, in fact.