New frontier for wireless: Sensor networks

Wireless sensors finally are moving from starry-eyed predictions to hard-eyed but limited production.

Wireless sensors finally are moving from starry-eyed predictions to hard-eyed but limited production.

As a result, network executives eventually can expect to see whole new classes of objects on their networks: pipe heaters, air conditioners, electric pumps and even grapes.

Wired sensor networks have been around for decades, with an array of gauges measuring temperature, fluid levels, humidity and other attributes on pipelines, pumps, generators and manufacturing lines. Many of these run as separate wired networks, sometimes linked to a computer but often to a control panel that flashes lights or sounds an alarm when a temperature rises too high or a machine vibrates too much. Also wired in are actuators, which let the control panel slow down a pump or turn on a heater or a fan in response to the sensor data.

Now advances in silicon radio chips, coupled with cleverly crafted routing algorithms and network software are promising to eliminate those wires, and their installation and maintenance costs (see graphic). Mesh network topologies will let these wireless networks route around nodes that fail or whose radio signal is hammered by interference from heavy equipment. A gateway will create a two-way link with legacy control systems, hosts, wired LANs or the Internet.

This combination of sensors and low-power wireless networking "give inanimate things an identity," says Ian McPherson, president of Wireless Data Research. "You can ascribe things like attributes, location and a history to an object." Wireless sensors will slash the costs of collecting this data, analyzing it and acting on it, he says.

These networks can use several different wireless technologies, including IEEE 802.11 wireless LANs, Bluetooth and radio frequency identification (RFID). But right now most of the action is with low-power radios that have a range of about 30 to 200 feet and data rates of up to around 300K bit/sec. Most of these, with their accompanying network software and APIs, are proprietary products.

But the IEEE last year approved the 802.15.4 low-rate standard for a simple, short-range wireless network whose radio components could run several years on a single battery. The ZigBee Association, a group of vendors, anticipates finalizing by year-end an industry specification for the network software that will run on the 802.15.4 radio chips.

"Over the next 12 months, the wireless trials will move to limited deployments, and they'll be in areas like environmental condition monitoring and meter reading," McPherson says.

Several vendors confirm that deliberate pace of deployment.

Building-automation vendor Andover Controls in Andover, Mass., has created wireless sensors based on hardware and software from Ember. Andover is working with an unnamed manufacturer to add the sensors to air conditioners. With an optional motion detector, also wireless, a hotel chain could tell when guestrooms are empty and turn down heating or cooling. That could save hundreds of thousands of dollars across the chain because hotels could better manage electrical use during daylight hours when rates are highest, says David Craven, Andover's director of OEM sales.

Philips Lighting Electronics plans to launch this fall two or three customer beta tests for fluorescent light fixtures fitted with wireless sensors. A pilot network at the company's Rosemont, Ill., headquarters, collected sensor data that can be used to see whether lights are on or off, monitor energy usage and diagnose the lamp's performance, according to James Sekinger, the company's director of business development for digital systems. Product shipments are planned for early next year.

Tyco Thermal Controls is adding sensor nodes, up to about 300, to stress test a small pilot wireless network in one of its manufacturing plants, says Ken McCoy, general manager of the firm's electronics business unit in Menlo Park, Calif. The company makes pipe heating systems, which look like wide bands of tape wrapped around the pipe's length. The sensors measure temperature data in industrial pipes, and pass the data wirelessly to a control panel where the heaters can be turned on or off. Tyco plans to start several customer beta tests in the next two to three months.

"I'm replacing wire that [in industrial sites] may be $10 per foot," McCoy says. "For 50 feet of wire, that's $500. Wireless sensors won't make for a breakthrough in capabilities, but we'll be able to install these systems much less expensively."

The Tyco project has taken longer than expected, even though Tyco's goal was simply to replace the wired connection and preserve the same data inputs to its stand-alone control panel. Ember's hardware and software ran reliably with adequate performance at 915 MHz and 2.4 GHz out of the box, McCoy says.

But the Tyco engineers had to do some rethinking about how to structure the network and how to relate the sensor nodes to the Tyco control panels. "We thought [the sensor network] would map to our current practices," McCoy says. "We almost reversed our traditional way of thinking about the net."

Instead of the control panel requesting data from a given sensor, now the sensor is the initiator. In effect, the sensor searches for control panels that want to "consume" its data, and then subscribes to that controller's "consumption service."

Interfacing with networks complicates the picture even more, and little work has been done on that. Accenture Technology Labs, the Palo Alto research and development arm of the Accenture consulting and technology services firm, built a pilot wireless sensor network to study how such networks can be integrated with enterprise networks.

Early this year Accenture set up sensors to monitor soil conditions, temperatures and rainfall at Pickberry Vineyard, a small California grower of premium grapes for wine. Accenture used hardware and software from Millennial Net. "The core activity of setting up the wireless net was relatively easy: just take it out of the box, turn it on, and then start," says William Westerman, an associate partner at the labs.

But the sensors were often power-hungry, so data polling and transmission had to be fine-tuned not to exhaust their batteries. They also could be finicky, being jammed 3 feet into the ground. While the network proved reliable over the five months of testing, there were rare losses of a given sensor's signal.

The data was fed over the mesh network to the Millennial Net gateway, which was equipped with a cellular network card, for transmission to a server on the labs' Palo Alto network. There the data was stored in a database and then converted into Web services, Web pages or input into Excel databases.

Westerman says enterprise network executives need to think about creating multi-level sensor and RFID networks. Sometimes a vineyard node would send a false reading, for example, air temperature at 300 degrees. "We added a rule at the gateway to filter out readings that were obviously wrong, rather then transmit this data all the way back to our server," he says.

Some data aggregation will be needed at the level of the sensor network. Instead of 30 temperature sensors in a warehouse all sending back the same reading, group them into zones, and send back one data point, such as "Zone A, 72 degrees."

The best way to tackle these issues, Westerman says, is the old-fashioned way: create a pilot, bring in a development kit from one or more of the vendors and create your own sensor network.

Join the Network World communities on Facebook and LinkedIn to comment on topics that are top of mind.

Copyright © 2004 IDG Communications, Inc.

IT Salary Survey: The results are in