Self-organizing sensor networks already are addressing environmental monitoring problems over large areas
Courtesy of Lockheed Martin
The Gotham United Cable Company today admitted that it had destroyed five factories belonging to rival networking giant Amalgamated Wireless. As he was being led away, Gothamï¿½s chairman commented, ï¿½Thereï¿½s almost nothing connected by wires anymore; everythingï¿½s gone wireless.ï¿½
Although the scenario is fictitious, the chairmanï¿½s comments soon may be close to the truth. Today, many highway bridges and buildings are being monitored by wireless networks of accelerometers and strain gauges. Wireless sensors were instrumental in the accident investigation of the Space Shuttle Columbia, and in the battlefield of the future, networked sensors will provide tactical support.
Some things are obvious about wireless networks, such as the absence of wires. In building-automation applications, running the control wiring for lighting is very expensive. Replacing conventional controls with wireless ones eliminates this cost while providing a degree of flexibility that previously was impractical. For example, changing the position of a light switch is as easy as sticking it on another wall.
Your preferred settings can be stored as part of the identification information encoded on your company ID badge. When you enter a room, the lighting and temperature automatically adjust to the data read by the entry scannerï¿½unless perhaps someone of higher rank already is in the room. Tracking individuals throughout a building as well as other security-related activities are some of the benefits of a wireless network.
The use of wireless sensor and actuator networks in heating, ventilation, and cooling (HVAC) building-automation applications is a commercial reality. For example, Advance Transformer, a division of Philips Lighting, has based a series of wireless lighting products on Emberï¿½s EmberNet hardware and software.
Millennial Net also markets wireless HVAC products and in addition is active in the mechanical inspection and measurement area. The Outrunner project, undertaken with Alstom Power, provides a simplified method of inspecting large turbine generator shafts for radial runout, a common cause of excessive vibration. Custom software analyzes data received from wireless dial indicators positioned at the two support trunnions and at the axial position being measured.
The largest initiative involving low-speed wireless sensors is the ZigBee Alliance, which includes major companies such as Motorola, Philips, Honeywell, Eaton, and ABB. ZigBee is aimed at low-power, low data-rate home automation, HVAC, and industrial networking applications. Ember and Millennial Net also are members of the ZigBee Alliance.
Much closer to the research stage are the efforts to develop networks of smart battlefield sensors. Ultimately, sensors scattered from an airplane would automatically form an autonomous network capable of detecting target position, speed, and size. Ideally, each sensor would be very small, giving rise to the term smart dust.
The original research in this area was largely done at the University of California (UC) at Berkeley, where the TinyOS operating system was developed as a framework for the network software, and the term mote was applied to the individual radio nodes. Each node was associated with one or more sensors. Dr. Kris Pister, who led much of the research, has formed a company called Dust Networks to commercialize some of the technology.
In addition to UC Berkeley, UC Los Angeles, Intel Research Labs, Robert Bosch, the U.S. Air Force Research Labs, Crossbow Technology, and the MIT Media Lab have been researching distributed sensor networks. The work at MIT extends ideas developed in W. Buteraï¿½s 2002 doctoral dissertation, Programming a Paintable Computer. Designing an operating system that scales well for hundreds of thousands of nodes is not trivial and is at least half of the sensor network problem: the actual sensors and node hardware are the other half.
Of these organizations, Crossbow Technology has produced the MICA range of commercial products that includes motes the size of two AA batteries side by side and a smaller 1″ dia version as large as a stack of two or three quarters (Figure 1). The company has developed a range of MEMS-based sensors that complements the small-size/low-power theme.
Dust Networks has successfully addressed applications such as power monitoring in supermarkets. Typically, a store has many freezers but little control or knowledge of the power being drawn by each one. Rather than disruptively attach traditional test and measurement instruments, Dustï¿½s very small nodes sense the current drawn and wirelessly transmit the measurements to a central collection node. The solution is quick to install and remove, needs no external power source, doesnï¿½t disrupt the existing wiring, and simultaneously monitors hundreds of loads.
Less Obvious Considerations
For applications that require long battery life, such as building automation or surveillance operations in areas that may be inaccessible, the separate nodes are designed to operate on very low power. To conserve power, the hardware sleeps as much as possible, waking up, for example, only when a sample must be taken or data received from another node. With low RF output power, reliable communications are limited to a range of 10 to 100 m.
Several network topologies extend this range. A mesh structure in which all the nodes are identical offers many benefits although it is not the lowest power configuration. The node density in this type of mesh must be high enough that each node has more than one neighbor within its communications range.
Redundancy contributes to robust networks that can overcome signal propagation degradation or the loss of a node. This topology also is the only one suited to forming tactical networks, for example by scattering nodes from an airplane. If different types of nodes were required periodically within the network, straightforward deployment might not be possible.
In those applications where you can determine the physical location of each node, a three-tier system of endpoint, router, and gateway has a power advantage. In such a star-mesh network, the many endpoints that directly connect to sensors donï¿½t have to constantly listen for transmissions from other endpoints. The endpoints communicate with the routers to transmit their data.
The routers form a network with other routers, passing data on until it eventually reaches a gateway, the interface to a conventional network or device such as a LAN, the Internet, or a PC. Commands from the routers to their associated endpoints can be arranged to avoid the need for constant monitoring by the endpoints.
Networks such as ZigBee, Millennial Net, and EmberNet use both battery-powered and AC-powered nodes. For example, endpoints attached to sensors typically would be battery-powered, but routers and gateways, where data streams are aggregated and the data rate is higher, could be AC-powered. In a lighting application, freely located photo sensors and switches would be battery-powered, but the actuator nodes associated with electronic fluorescent ballasts already have AC power available in the ballast.
As a result, there are subtle differences among mesh networks. Depending on the spacing of the routers in a star-mesh network, an endpoint may have only one communications path. Additionally, router spacing determines the degree of redundancy in the mesh formed by the routers.
In practice, either a star-mesh or a pure mesh approach initially can work. However, Mike Horton, CEO of Crossbow Technology, argues that a mesh network of identical nodes offers practical advantages. ï¿½Because the RF world is very dynamic, the redundancy provided by a mesh network in which each node can communicate with several others avoids having a single point of failure. Even in relatively static situations where site surveys have been done, if somebody moves one piece of equipment, the RF environment may significantly change.ï¿½
If a mesh network is compared to a single-hop radio link, the mesh network can be shown to require much less power. This is because the power needed for reliable communications goes as the square of the distance in open areas but as distance to the fourth power in built-up urban areas. So, making multiple short hops can consume much less power than a single hop of the total distance although the application must tolerate some data latency.
According to the www.zigbee.org website, the goal of the ZigBee Alliance is to provide ï¿½wireless control that simply works.ï¿½ When the ZigBee standard is approved, it should support interoperability among a wide range of devices from different manufacturers, much as the Bluetooth specification and the 802.11 Wi-Fi standards have done.
The ZigBee physical (PHY) and medium access control (MAC) network layers are specified by IEEE 802.15.4, which supports a communications distance of up to 100 meters in three bands with separate maximum transmission rates: 868 MHz at 250 kb/s, 915 MHz at 40 kb/s, and 2.4 GHz at 20 kb/s.
The physical layer uses direct sequence spread spectrum (DSSS) encoding, which allows low-cost analog implementation. Channel access is via carrier sense multiple access (CSMA) with collision avoidance (CA). The data rates possible under 802.15.4 generally are low, especially if the 2.4-GHz band must be used to avoid the crowded 900-MHz area, for example, which accommodates global system for mobile communications (GSM) phones in Europe. Nevertheless, having a standard that governs the physical radio reduces the burden of compliance testing.
For HVAC applications, bandwidth is not an issue. The variables in a controlled space donï¿½t change that quickly. However, for test and measurement applications, low-speed, low-power constraints of wireless sensor networks have been overcome by adding local storage to the relevant nodes.
Examples of products taking this approach are the Crossbow MDA420CA, which provides a 25-kS/s sampling rate with a 16-b ADC and a 128k local buffer, and the Invocon WB MicroTAU tri-axial accelerometer unit, which samples at 20 kS/s. After the test is complete, data can be downloaded from the local memory at a rate the network can sustain.
The Invocon wireless nodes provide intersensor synchronization to within 6 ï¿½s. This means that when the user is ready to download the acquired data from all the sensors, it already is time-tagged and synchronized to greater accuracy than required for the typical modal analysis that will follow (Figure 2).
Applications that require continuous monitoring, relatively high bandwidth, and thousands of sensors, such as an airframe structural test, may be incompatible with todayï¿½s wireless sensors and networks. Development of wired sensor networks will allow a pair of wires to support perhaps a thousand sensors, eliminating most of the cost associated with the present practice of point-to-point wiring for each sensor in a large test. Making the step to a wireless sensor network for this type of testing is at least a few years in the future.
Differentiating Among Offerings
Once ZigBee becomes established, the smaller players in wireless low-speed distributed sensor networks must offer distinctive benefits to remain viable. Millennial Net and Ember have successfully identified early adopters that are actively using their proprietary technologies. Both companies have joined the ZigBee Alliance and chosen 802.15.4 for their PHY and MAC layers.
Crossbow Technology is easily distinguished as one of the few companies with a strong sensor program. Both tilt sensors and accelerometers are based on the companyï¿½s proprietary MEMS technology. Sensors available for use with the MICA motes include light, temperature, relative humidity, barometric pressure, acceleration for seismic applications, acoustic, and magnetic.
Invocon has been delivering custom, self-organizing wireless sensor networks for aerospace and industrial applications for at least 10 years. The company provided 130 nodes to monitor a highway bridge in Texas and has developed many special-purpose products for NASA.
For example, the Shuttle Wireless Instrumentation System network monitored the temperature of equipment as it was being transferred from the Shuttle Cargo Bay to the International Space Station. This network flew on Shuttle flight STS-92 Oct.11, 2000, and on STS-97 Nov. 30, 2000.
If you have a large-scale data acquisition application, you should consider using a wireless sensor network. Hundreds of thousands of nodes have been shipped, major companies are relying on the technology, and more types of applications are being addressed daily.
Among the benefits of such a network are the accuracy and timesavings that follow from self-organization. Because these networks are scalable to very large sizes, they are self-organizing. Manual organization, which might only be tedious at 50 nodes, soon becomes a major part of the overall job at 1,000. And for nodes with global positioning system capabilities, you donï¿½t even need to record where each node has been placed.
All manufacturers we contacted for this article agreed that integration will play a large part in future node size and cost reduction. The remaining size limitation is the battery that can become smaller as a result of new battery technologies or with lower power node electronics. Research is continuing into alternate power sources to drive very low-power nodes. Two possible approaches involve harnessing energy from ambient radio waves or seismic vibrations.
So if one day your cell phone bombards you with an unusually high number of ads for local services, it could be because of the cigarette butt you kicked while walking down the street. The kick both triggered and powered the innocent-appearing wireless node that you thought was just harmless litter. Armed with a camera and biosensors, it measured your relevant parameters and transmitted them in milliseconds to a nearby crushed Coke can, the closest wireless router in its address list.
FOR MORE INFORMATION
on Millennial Net
on the ZigBee Alliance