Something moves, and what looks like a dime-sized pebble “wakes up” in a vast desert landscape. The pebble sends a signal to another small stone just 20 yards away. It too is awake. It detects carbon, nitrogen and sulfur dioxides. The chemical sensing stone sends its information, along with data from the seismic sensing pebble, to a node stone. The node collects data from hundreds of similarly disguised wireless microelectromechanical (MEMS) sensors and relays it to an unmanned aircraft that pieces together the information to identify a tank.
Advances in smart, low-cost integrated devices containing many different types of sensors, wireless transceivers and processors with significant computing capabilities could make the above scenario a reality in as few as five years, says Professor Parameswaran Ramanathan. Ramanathan recently received a $725,000 five-year grant to investigate issues in establishing and maintaining communication between sensor devices in wireless ad hoc surveillance networks. The grant is part of a multi-million dollar Multidisciplinary University Research Initiative (MURI) funded by the Army Research Laboratory. Co-investigators include Penn State, Duke, Cornell, UCLA, and Louisiana State Universities.
“Aircraft could sprinkle large numbers of these integrated devices over an area to construct a surveillance network capable of monitoring, detecting, and tracking threats from a variety of sources including vehicles, persons and biochemical agents,” says Ramanathan. “Of course there are dozens of non-military applications as well. Sensor networks could be deployed to help geologist, limnologists and others study the environment in ways that currently aren’t possible.”
But before such networks can be put to wide use, researchers have to solve some difficult problems. With thousands of sensors gathering and disseminating information, some are bound to deliver false readings. Not only do the devices need to be robust, but the network also must have strategies to sort out errors and pass on correct information. In addition, because the sensors will have limited power supplies and communication capabilities, the team must devise protocols and communications strategies that deliver the most information while using the least amount of power.
“If I use a lot of power, I can shout in some sense and reach maybe 20 devices,” says Ramanathan. “But if I whisper and use less power, I can only talk to two. Then those two devices can rebroadcast whatever they heard and get the whole network going. So the challenge there is of figuring out if is it best to shout and have everybody hear or talk a little and have everybody communicate. But then of course, everybody is communicating and everybody has to handle not only his or her own thing, but also relay something someone else said and that consumes more power. So there is a balance to strike.”
In some cases, the end user may not want the devices to talk at all. A lot depends on the question being asked of the network. Sayeed is working on signal processing strategies that most efficiently combine information from many kinds of sensors. He says it’s very much like bringing a blurry picture into focus. Speed is achieved and energy conserved by finding answers when the image is still soft.
“The key premise behind our approach is that not all things require the same level of accuracy,” he says. “If I want just to detect the presence of an object, that is a binary decision. It is either there or not. If after that I want to classify whether that object is a car or a person, I need more dimensions and accuracy. What we’re doing is multi-resolution signal processing.”
In mid-November, the team demonstrated the state of their technology at a military proving ground in the Southwest. Using its software with 90 deployed devices, the team successfully showcased the tracking of military vehicles.