Wireless Sensor Networks Are Taking Over

For military surveillance, a self-organizing network of minute, autonomous wireless sensors is very attractive. Once a group of nodes has been deployed, perhaps dropped by air, troop and vehicular movements can be remotely monitored. The initial research on this application helped define the mesh concept and software operating systems used in wireless sensor networks (WSN).

Although actual surveillance systems have been deployed and research continues, commercial WSN requirements are different. In many industrial applications, small size is not among the most important priorities. Instead, ease of use, the capability to provide a complete solution, and vendor support top the list.

For example, ZigBee is a WSN that eliminates the need for large amounts of office building control wiring. To continue reducing expenses, a typical lighting switch or temperature control must have a very long battery life. Having low battery maintenance obviously saves money, but to provide a service life of several years from a standard commercial battery requires exceptionally low power consumption.

In critical industrial applications, a wireless solution with a long battery life may be the only practical alternative. Consider a monitoring device mounted at the top of a smokestack. A conventional solution would entail expensive installation of hundreds of feet of environmentally protected wiring together with local power supplies and monitoring electronics. Even a wireless approach must be very reliable and require only infrequent battery replacement given the cost and difficulty of access.

Some types of WSNs retain a few line-powered nodes, typically high in the network hierarchy, closest to the computer actually receiving the sensor data. The rest of the nodes are battery powered or have the capability to be powered from an energy-scavenging device. To many companies in the WSN industry, the term wireless applies to all wiring, not just power and not just control.

Niek Van Dierdonck, vice president of strategy and product management at GreenPeak, expressed it this way, “We have developed an ultralow-power strategy because the strength of a truly wireless sensor network can only be fully utilized when the wiring for both the data communications and power cables can be eliminated. Eliminating the data cable solves only half the problem: The installer still needs to run a power wire. And, half a problem solved is as good as not solving it at all.”

Nodes operate with a very small duty cycle to conserve power. In other words, they are as inactive as possible for as long as possible. In WSNs with pre-assigned time slots, an on-board timer can alert the node to wake up at the proper time, transmit its information quickly, and shut down again. It also is possible for a node to monitor transmissions in the background, only turning on all its circuitry when the appropriate code has been received. Arch Rock calls this mode of operation passive vigilance.

The time synchronized mesh protocol (TSMP) used by Dust Networks reserves 47 of the 127-B maximum packet size specified by the 802.15.4 radio standard, allowing 80 B for the payload. Figure 1 shows the TSMP packet structure with its emphasis on data integrity. Direct sequence spread spectrum (DSSS) coding improves multipath performance, and frequency hopping spread spectrum (FHSS) operation across 16 predefined frequencies effectively increases channel bandwidth by a factor of 16.

The data rate supported by an ultralow-power radio limits the types of applications that can be addressed. The amount of on-board data reduction that can be performed is necessarily power limited, as is the amount of data that can be transmitted each time the node wakes up. Nevertheless, WSNs are well suited to applications such as process-control temperature and pressure monitoring; building heating, ventilation, air-conditioning (HVAC) control; and many types of remote environmental monitoring.

Very long battery life is not as important in many troubleshooting and monitoring test and measurement applications as it is in WSNs that are part of a building infrastructure. MicroStrain offers several wireless strain gauge-based data acquisition systems that are good examples of the power/speed trade-off. Data can be logged at up to a 2,048-Hz rate or streamed in real time at a rate up to 4 kHz by the V-Link® instrument, but its 600-mAh battery may have a life of only 55 days with four 1,000-? strain gauges (Figure 2).

WSN Architecture and Standards

A few specific terms are related to WSNs and the associated standardization efforts currently underway:

In contrast, Ember’s Zigbee technology includes a coordinator node that configures and controls the network, line- or battery-powered router nodes, and a number of battery-powered end devices that only communicate with the routers. Ember’s router nodes form the self-healing mesh and together with the end devices form a star-mesh hybrid topology.

WSNs are not confined to industrial applications, but there are many industrial data acquisition and control requirements for which they are good solutions. This is the reason that there is so much activity toward standardization within organizations such as the ISA whose members are process control and industrial automation companies.

Many companies have developed proprietary WSNs, and as you would expect, they are not interoperable. The intention is for ISA-SP100 to be an open standard that builds on the experience of these companies much as the WiFi Forum has achieved interoperability among competing Wi-Fi products. The ISA committee goal is to adopt one standard by early 2008.

In the meantime, it’s useful to consider the unique attributes of the different approaches. Each manufacturer has achieved a degree of success in one or more markets, and some solutions are very specialized.

WSN Examples

Hierarchical Networks
A few of the larger application areas, HVAC, home area networks (HAN), and utility meter communications (advanced metering infrastructure—AMI), are addressed by the ZigBee specification. The IEEE 802.15.4 PHY and MAC specifications are used, but the higher network and application layers of the ZigBee protocol are defined by the ZigBee Alliance with more than 225 member companies.

ZigBee solves many of the problems concerning end users such as ease of installation and compatibility with co-located Wi-Fi and Bluetooth radios. It is used by hundreds of interoperable products. From the point of view of the companies involved in the SP100 effort, however, ZigBee does not guarantee the very high level of security and reliability required by critical control signaling encountered in process control and factory automation applications. This is a special concern that ISA-SP100 will address.

Sensicast’s SensiNet solutions are similar to ZigBee, with three layers with the mesh capabilities provided by the intermediate wireless routers, but a separate coordinating node is not required. However, unlike ZigBee, all parts of the WSN solution, from the Smart Sensors through to the application software, are made by Sensicast.

The company supports a number of networking protocols including the 802.15.4 physical radio specification and those parts of 802.11 covering Wi-Fi and wired Ethernet networks. In addition, Sensicast intends to support ISA-SP100 when the standard has been finalized.

Smart, battery-powered sensors transmit captured data to mesh routers that, in turn, communicate with a gateway. The network is self-configuring and, depending on the number of routers involved, self-healing. Hop-to-hop acknowledgement ensures that data has not been dropped. The networks have been applied in pharmaceutical, food and beverage, and industrial applications, especially where regulatory governmental environmental compliance is required as part of testing, manufacturing, and storage and shipping processes.

For industrial applications, Accutech, a division of Adaptive Instruments, provides a range of integrated sensing and signal-conditioning field nodes based on FHSS technology. The battery-powered nodes are FM/CSA/ATEX rated for Class 1, Division 1 hazardous environments with ambient temperatures from -40°C to +85°C, and up to 100 can form a LAN with a base radio node. Up to 255 base radios can be deployed in overlapping LANs. Figure 3 shows a field node attached to a range-extending antenna.

According to Wallace Lueders, the company’s vice president of sales and marketing, Accutech focuses on process measurement and industrial automation in continuous and batch manufacturing. A large number of industries as well as municipalities and hospitals are served by the more than 2,000 networks installed to date.

A fast and deterministic response time is ensured by the direct communications from field node to end PC or programmable logic controller (PLC) via a single base radio node. It also is possible to add base radio repeater nodes to create redundant paths for higher reliability. In addition, the repeater nodes can create an intermediate network layer by concentrating data from a number of base radio LANs. In these cases, the repeaters would connect directly to the network PC or PLC.

Millennial Net WSNs are based on the MeshScape network protocol that features persistent dynamic routing. This approach is advantageous in systems with rapidly changing topologies. In contrast to the static network representation that might be drawn on paper, actual WSN communications are dynamic, taking various routes through a mesh network as local conditions change.

The basic star, mesh, and star-mesh hybrid physical configurations are supported as are several data models. Periodic sampling is a good solution in temperature or pressure monitoring applications. Event-driven operation better suits alarm conditions that arise when preset limits are exceeded. Controller-based installations can use serial polling. And, when a portable gateway must be added to a network for ad hoc data access and analysis, an on-demand capability also is available.

Millennial Net recently introduced longer range options with output power up to 100 mW, the maximum level permitted in North America, compared to the previous 10-mW limit, still the globally permitted maximum. These new products allow network designers to trade range and power against number of nodes. Where previously many nodes were required to cover a large geographical distance in a hop-by-hop mode, 10x higher power can reduce node count and data latency at the expense of battery life for the high-power node.

GreenPeak features low-power routing, which means that the network does not depend on a backbone of line-powered wireless routers as do some ZigBee networks. Both end devices and routers can sample sensors and control actuators. However, the main distinction is the low power and associated low maintenance requirements of the entire network.

In addition to obvious design considerations that support low-power operation, GreenPeak has developed synchronization techniques. Nodes only wake up during their assigned time slot. Time slots overlap so that communications are possible between adjacent nodes, for example, but not simultaneously among a large group of nodes. This approach strictly enforces hop-by-hop message passing because only a small number of nearby nodes can be listening for messages at any time.

In this case, low power and long battery life have been emphasized at the expense of message latency. However, because the power required is so low, it is possible to use energy-scavenging devices as power sources, totally eliminating battery replacement maintenance.

The zSeries Wireless Sensor System from Omega Engineering features a number of sensors and battery-powered end devices for monitoring barometric pressure, relative humidity, and temperature. As many as 32 end devices can be located up to 300 ft from a line-powered coordinating node connected to an Ethernet network and the Internet.

Data is accessed from a Web browser via the IP address of the coordinating node. The system is low cost and easy to install and doesn’t need a PC for operation. With a multipoint-to-single-point architecture, the zSeries does not provide redundancy. The wireless end devices operate at 2.4 GHz and comply with the IEEE 802.15.4 radio specification.

Mesh Networks
Three companies support WSNs in which all the wireless nodes can send and receive information among themselves. The network implementations are proprietary but architecturally similar in having few restrictions on node capabilities.

Arch Rock has adopted IP-based connectivity throughout the WSN. The description of a new product, Primer Pack/IP, given by Brian Bohlig, vice president of marketing, indicates why this is relevant: “Primer Pack/IP extends the original Primer Pack by making it the first commercial implementation of the Internet Engineering Task Force (IETF) 6LoWPAN standard (RFC 4944) for IPv6 communications over low-power IEEE 802.15.4 wireless radio.

“Primer Pack/IP runs native IP end-to-end, taking the IP protocols beyond their current boundary at the WSN gateway and out to the individual sensor nodes,” he continued. “Because of IP’s pervasiveness as a global communications standard across industries, Primer Pack/IP sensor nodes will be able to communicate directly with other IP devices…. Network managers will gain direct, real-time access to sensor nodes and the ability to apply a broad range of Internet management and security tools.”

At the network layer, distinct protocols enable triple redundancy in the data-collection path, directed routing to task a specific node with an action, density-aware dissemination for reliable propagation of small objects throughout the network, and over-the-air programming and provisioning for disseminating large objects such as major system software updates.

Arch Rock technology has been incorporated in the Advanced Incident Response System (AIRS), an IPv6-enabled network that allows personnel from various local, municipal, and government agencies to communicate seamlessly during an event while minimizing dependencies on the fixed infrastructure of the disaster site.

Dust Networks was one of the first companies developing WSN products and continues to favor mesh networks with no special routing or infrastructure nodes. Data is encrypted for security, and all the nodes are both low power and routing-capable. Robert Shear, the company’s director of corporate marketing, focused on the bottom line: “The emphasis needs to be on the payback of the solution—period. Small size is not a primary goal, but ease of use, flexibility, and reliability are.

“Any viable WSN solution must be simple to install by instrument technicians, extremely reliable, and self powered.” He contended, “Networks that require extensive RF expertise to deploy or power supply cable runs will not be viable. The revolution with WSNs is not that instruments are now wireless. The revolution is that the networks just work. Now, you can put nodes places previously infeasible with wired solutions.”

BBN’s WSN experience has a large military element, which necessarily puts higher value on small size and low energy requirements than most commercial applications. Principal Scientist Jason Redi said, “We support full mesh networks with equal peers. Years ago, we used a dynamic hierarchy for routing but found that the complexity in maintaining the hierarchy was a problem. Hierarchy can be used to distinguish different kinds of nodes, but most of our customers have homogeneous systems so peer-to-peer meshes work better for them.

“Many of our sensor technologies are used in military applications,” he explained. “Examples include soldiers dropping sensors while clearing a building to determine if someone else enters the structure or watching for patterns of vehicles at certain intersections. These applications usually require multihopping of information back to places where humans can observe it.”

For military sensors, small size and low energy requirements are more important than ease of integration. Wireless interoperability in the commercial world wasn’t possible until recently, but as people become used to it, size and energy may again become primary considerations because they affect weight and logistics.

Testing WSNs

Test and measurement applications typically involve higher data rates than those compatible with ultralow-power WSNs. The 4-kHz rate of MicroStrain’s V-Link instrument is adequate for many mechanical or electromechanical tests, but battery life is relatively short. There no doubt are similar types of test or troubleshooting applications for which higher data rates and shorter battery life would be a good fit.

From a test and measurement point of view, ultralow-power WSNs are important as products that need to be tested rather than a means of performing a test. These networks are proliferating, and being wireless, their performance is subject to degradation from many sources. For example, adding or relocating large equipment in an industrial plant could affect signal propagation. Similarly, extending a WSN or relocating its routers may affect latency.

It’s logical then to question how extensively a WSN has been tested before deployment and by what means this was accomplished. As an example, Sun Labs, the applied research and advanced development arm of Sun Microsystems, has developed a software application that supports interactions among real and virtual nodes on a simulated network. The real Java-based nodes, Sun small programmable object technology, or Sun SPOTs, together with development software, can be purchased by anyone who wishes to experiment with this WSN technology.

GreenPeak’s Mr. Van Dierdonck included simulation as the first step in a four-stage procedure. New ideas are simulated and then tested in wireless nodes within a well-controlled network environment. To separate the network under test from the test-logging system, a high-speed wired backbone is used to provide performance feedback. Endurance testing follows and finally field test to guarantee correct functional working.

BBN’s Jason Redi described a three-step process. “We use shared-code software, where code that runs in the nodes is exactly the same code that runs in simulation. We test against traffic load, density, network size, mobility, and node failure. We also have a node emulation platform that supports real-time testing on the actual processors but with simulated radio hardware. Finally, we run our networks 24/7 throughout our building and continuously stress test new software.”

In a form of self-test, Arch Rock sensor nodes send heartbeat packets periodically to check network reliability and connectivity. The packets contain key health statistics that the management server can analyze to detect any problems. Because Arch Rock WSNs use native IP, standard network management tools such as SNMP, Ping, and packet sniffers can be used to diagnose problems.

Ember’s ZigBee platforms include interface ports that provide a nonintrusive trace of all packets sent or received at a device. Using an adapter that communicates with the debugging application via TCP/IP over Ethernet, PC-based tools can take advantage of the on-chip debugging I/O. This wired channel provides significant bandwidth for debugging data without swamping the wireless channel. The on-board interface port also avoids the ambiguity that can result when a sniffer receives transmissions that the node being investigated did not.

Summary

In many ways, testing a WSN is much like testing any packet-based communications network. However, network characteristics such as ultralow-power consumption, a hierarchy of node types, and a need for very high security can create challenges in addition to those usually addressed by protocol analyzers and traffic generators. Perhaps the most familiar WSN from the test point of view is the Arch Rock system. Although it actually is wireless, it behaves like a standard IPv6 network.

At this stage in WSN development, a large number of protocols are in use. As the market matures, this number should decrease. It also will be affected by standardization initiatives such as ISA-SP100.

WSNs have progressed well beyond being interesting curiosities. In many cases, they are the preferred solution. In addition to HVAC, process control, and industrial automation, WSNs are finding their way into applications as diverse as parking garage CO₂ monitoring (GreenPeak) and evaluating the temperature control of the penguin habitat at the Pittsburgh Zoo (Sensicast Systems).

January 2008