Turning the world into a sensor network

Try to imagine a "world littered with trillions" of wireless sensors. Now try to imagine the problems getting even a few thousand of them to work together in any kind of intelligible way so you can know if that interstate bridge is near collapse or the natural gas pipe behind a housing development has a crack in it or how dropping your AC temperature by 3 degrees during peak demand will clobber your electric bill.

Wireless sensor net helps UC-Berkeley boost data center efficiencies by 20%

Those are the problems that a new research project at Carnegie Mellon University (CMU) is going to explore. It has, as most such government-industry-academia joint efforts do, the cumbersome name of Pennsylvania Smart Infrastructure Incubator (PSII). The basic idea: Bring together some smart people, give them state of the art facilities and communications, and ask them to wrestle with how to build and run really big sensor networks that can deliver useable information.

CMU already has a lot of practical experience in sensors. It's launched an internal project called Sensor Andrew, which is gradually adding a wireless sensor infrastructure burrowed into every campus building. So far, Sensor Andrew reaches five buildings on the Pittsburgh campus, each using the networks for different purposes, from tracking locations of people to warning that a printer is still using maximum power, due to a low-toner alert, instead of shutting down.

The campus sensor network makes use of homegrown technology: a low-cost wireless mesh node called FireFly, and a real-time operating system specifically designed for such networks. Like other similar products, FireFly uses an IEEE 802.15.4 transceiver, good for 150 to 300 feet. It has a maximum raw data rate of 250Kbps and an 8-bit microcontroller, and SD Flash card slot, to process data from four optional on-board sensors: light, audio, temperature, humidity, acceleration.

What's different is that the FireFly node also has a low-power AM/FM radio receiver. That radio can pick up a periodic time synchronization pulse, from an AM carrier current transmitter that can flood an eight-story building with the signal. This kind of synchronization makes possible very energy-efficient operation, and extends the battery life of each node by a factor of four or five, according to CMU. The pulse enables precise scheduling of data transmits and receives, leaving the nodes "sleeping" the rest of the time.

The use of a real-time OS, called Nano-RK, for FireFly reflects the embedded systems background of Ragunathan Rajkumar, a professor with CMU's Department of Electrical and Computer Engineering, who oversaw FireFly's development. The RTOS creates a "bounded" system, with a high degree of predictability. What's more, tasks can specify their differing resource demands, and Nano-RK creates guaranteed, controlled access to resources like CPU cycles and network packets. Temperature changes slowly; audio requires a much higher sampling rate. "We can deal with multiple sensors at once," Rajkumar says. "Each sensors operates at its own ‘natural’ frequency."

One key result is that Nano-RK can enforce energy budgets at both the task level and the system level, minimizing power.

The FireFly infrastructure has been deployed in one of the many coalmines honeycombing the Pittsburgh area. During regular operation, the sensors provide a range of standard sensor data, and most importantly track the location of each miner. In an emergency, the FireFly network can switch over to high-rate operation for voice communications.

It's these kinds of uses that the Commonwealth of Pennsylvania hopes to encourage via the PSII project, turning the Pittsburgh area into a showplace for "smart infrastructure" technology: the hardware and software needed to deploy and manage lots of wireless sensors, often in places "where the sun don't shine." The state awarded a multi-million dollar development grant to launch PSII on the CMU campus, as part of the Department of Civil and Environmental Engineering. Construction workers are creating two research centers, one funded in part by Canadian transportation giant Bombardier, the other by IBM.

The Bombardier Collaboration Center will draw researchers from business and academia to collaborate on ways to combine civil infrastructure systems, including transit operations, with "cyber infrastructure" systems – computers, networks, sensors and software. The IBM Smarter Infrastructure Lab will focus on the challenges of collecting and processing real-time sensor network data, and of creating analytic tools to understand what it means.

"A lot of the research work will be less on the sensors themselves and more on the fact that many kinds of sensor systems are being deployed," says Matt Sanfilippo, PSII executive director. "The challenge is how to make them talk to each other, to network [together] multiple networks, and in a relatively simple way to manage them and the systems they're installed in.

In a typical modern building, both heating/air conditioning systems and fire and smoke detectors could make use of real-time data from temperature sensors, Sanfilippo says. But today, these are separate and usually proprietary systems. "We want a way of doing sensing that can make the data available to any application that needs that specific data," he says. "It's analogous to the evolution of the Internet, and we want to leverage that."

Another research focus will be combining data from point sensors, like temperature or pressure, with streaming data from video or infrared cameras, which can treated just like conventional sensors.

Both the Bombardier and IBM facilities will be outfitted with an array of sensors and sensor networks.

CMU's past sensor projects, via its Center for Sensed Critical Infrastructure Research, show both the challenges and the potential for intelligent wireless sensor networks. One recent project with the federal Department of Energy focused on attaching ultrasonic sensors to natural gas pipelines to detect metal degradation and cracks. Not itself an original idea, the CMU work focused on how to greatly improve the accuracy of such systems, to distinguish between a real crack and the vibration caused by an 18-wheeler rumbling over a crossing.

In some cases, projects try to avoid installing new sensors and look for new ways to exploit existing systems. One building energy management project, funded by the National Science Foundation and Bosch Corp., set up a system at the central breaker panel to monitor and correlate energy use patterns. The system watched electricity flows and associated distinct flows with specific devices such as air conditioners or clothes dryers. Software analyzed the use and began recommending changes to energy consumption levels, and time of day use, to minimize energy use and lower the electric bill.

"You can do a lot with what's already there and working, sometimes taking advantage of new analytics, without having to deploy new sensor networks," Sanfilippo says.

John Cox covers wireless networking and mobile computing for Network World.

Twitter: http://twitter.com/johnwcoxnww

Blog RSS feed: http://www.networkworld.com/community/blog/2989/feed

Read more about anti-malware in Network World's Anti-malware section.

Join the newsletter!

Error: Please check your email address.

Tags wireless sensorsNetworkingwirelessWireless Management

Show Comments