| The advantages of wireless sensors become more evident from a view of an ABB robot arm in a manufacturing cell. The wireless version (bottom) is more reliable and has fewer restrictions on movement caused by considerations for cable runs. |
| An ad hoc wireless network consists of wireless nodes that self-configure to form a network without any help from a central controller. Nodes in the network handle handshaking and other network tasks themselves through use of a distributed-control algorithm. |
Doug Molitor
Silicon Valley Wireless Communications Alliance
San Jose, Calif.
Get ready to say goodbye to a lot of the cabling that pervades the typical industrial plant. Wireless sensor networks are on the way. These innovative systems promise to bring unheard-of operating efficiencies to the average industrial concern. They'll make it practical to characterize aspects of machine operations that are little more than black-box processes today.
Though networks of wireless sensors are becoming commercially available, no question the topic is still a subject for research. Most work in this area concerns two hurdles: Energy constraints imposed by the use of batteries, and how to handle the loss of data packets in the low-speed networking schemes under consideration.
Among the institutions researching such issues is Stanford University and its Wireless Systems Lab (WSL) (http://systems.stanford.edu/) led by Dr. Andrea Goldsmith. Goldsmith is considered among the leading wireless-communications experts in the world.
Energy-constrained sensor networks are among the topics under scrutiny at WSL. One attraction is the elimination of communications cables in the design of machines, factories, or even for automated highways. Freedom from cabling means easier access and less wire-spaghetti in and around machines and equipment. Wireless also saves money throughout the life cycle of the installation because it reduces the cost of engineering and installing sensors for such purposes as statistical process control or quality assurance.
And wireless networks are easily deployed. Moreover, they have no single point of failure in many design topologies.
Some of the first applications for wireless sensor networks are in military and environmental settings. Large numbers of autonomous sensors will help detect and track battlefield activity, chemical spills, and even snowfall. Such applications could end up involving networks of perhaps thousands or even tens of thousands of wireless sensors. Networks set up in factories or plants are more likely to encompass a few hundred wireless sensors at most.
Clearly there is progress in both energy constraints and reliable networking. Recently ABB, the global power and automation technology firm, introduced a wireless sensor that successfully addressed both these issues. Its new wireless proximity switch boasts a wireless communication protocol that ensures reliable delivery of messages within the short timeframes required by modern, discrete automation systems. It operates in the 2.4-GHz band allocated to industrial and scientific users.
The ABB proximity sensor is an inductive sensor that also uses inductive coupling as a means of obtaining operating power.
The device needs no external power. It uses low-power sensor electronics and an energy source derived from magnetic fields. The self-contained energy supply completely eliminates the need for cables or periodic battery replacement.
The sensors typically provide feedback for a PLC that is controlling positioning equipment inside a manufacturing cell. The manufacturing cell is bathed in a magnetic field (at about 120 kHz) that the sensors pick up through inductive coils and convert into electrical power.
A different premise
Traditionally, most network research is based on two assumptions: That there's a need for a high data rate, and that the energy supply is unlimited or rechargeable. Neither assumption applies to networks of wireless sensors. Sensors generate only small amounts of data and don't need high speeds to transmit. And of course they generally work from batteries that may not be easily recharged or replaced.
Stanford's Goldsmith believes these realities will make wireless sensors in a factory setting quite different from traditional network devices. It's likely they'll "comprise ad hoc networks where multiple nodes communicate for one application, separate from other nodes communicating for another application, with no central coordination," she says.
An ad hoc network is a collection of nodes that self-configure to communicate without the aid of any established infrastructure. Nodes themselves handle the necessary handshaking and networking tasks, generally through use of a distributed-control algorithm.
This scheme contrasts with other common architectures such as wireless local-area networks (WLAN) where network devices usually communicate via a so-called backbone connection at a central access point. It also differs from cellular-phone systems. Here mobile units communicate by means of fixed base stations each serving an area of geography called a cell.
"One question we are investigating in the lab is how to build a network where energy is constrained," says Goldsmith. When energy is a constraint, everything the network does to consume energy must be analyzed. For example, data compression becomes a trade-off: What takes more energy, compressing data or transmitting uncompressed bits? The same question arises for optimization of routing protocols. It may be more energy efficient for network devices to use a simple relay scheme rather than spend resources finding the best path in the network.
Energy constraints demand that wireless hardware and software be designed together in a holistic approach that allows for their interdependence. Simply put, software protocols must be developed simultaneously with the hardware to implement strategies for conserving energy.
"Most of the interesting problems are interdisciplinary," says Stanford's Goldsmith. "But there is a huge resistance by many researchers to go outside their expertise. Research breakthroughs will necessarily come at the intersection of disciplines. This is the coming trend."
Designs based on the standard layered approach to network protocols don't work well in energy-constrained networks. This eliminates from consideration the standard ISO communications model where each layer of the protocol stack is designed and operated independently. The ISO model defines interfaces between layers as static and independent. It works well for high-speed network design where energy is no problem. The difficulty with using it for energy-constrained wireless networks is that a lot of power gets dissipated running software that basically just hands off information from one layer to the next.
Another hurdle getting scrutiny is loss of communication packets. Lost packets are a given, Goldsmith notes. "Wireless networks are notorious for communications packet loss. In a chemical process or paper manufacturing, packet loss is unacceptable, but the typical off-the-shelf communications controller is not robust enough and never accounted for packet loss," she says.
So far the wireless market has focused on high rather than low data rates. For example, cellular network providers are promoting high-bandwidth tasks such as sending photos by phone. Packets get lost in such applications, but current microcontroller designs reconstruct them via error-correction algorithms that use data from received packets.
The reconstruction problem is more complicated when there are fewer packets with which to rebuild missing data. Ditto for small packets each carrying minimal data. That is one reason ABB's recent development of a wireless industrial-grade proximity switch was such an achievement.
It incorporates a wireless communication module for the power supply, signal transmission, and man-machine communication. ABB researchers devised a special version of the Bluetooth protocol that ensures reliable delivery of messages within the short timeframes that discrete automation systems demand.
Interference problems
Wireless setups save money throughout the life of the installation because, among other things, they eliminate unplanned production stops from cable problems. Wireless systems are also more flexible than wired equivalents in mobility and simple reconfiguration.
But the reliability such systems gain through cable elimination they potentially give back in RF interference concerns. Each node in a wireless sensor system broadcasts to others within range over a decentralized multihop routing system. Messages meant for a specific sensor may be relayed via several others. The RF traffic such a scheme entails can be significant when networks contain hundreds or even thousands of wireless sensors. Complicating the situation further is the presence of numerous noise sources in the same frequency range planned for sensor networks. Microwave ovens and portable phones, for example, emit RF in the same general frequency band used by Wi-Fi equipment.
Experts say that RF interference is not a showstopper for wireless nets, even those in factories characterized by a lot of impulse noise from welders and other sources of arcing. But wireless sensors must allow for these realities and include filtering or other countermeasures that let them function in the midst of potentially significant noise.
Indications are that wireless sensors will also be wired into other kinds of wireless communication devices. Similar challenges occur among noncommunication networks that are already part of the wireless revolution. According to Philip Macafee, president of PowerTouch Technologies Inc., "One of the future keys to success will be the ability to integrate multiple wireless networks together. You may have process or plant control in one or two well-managed sensor networks. But plant security is on radio, executives have their cell phones, and maintenance people probably use push-to-talk. These island networks are a growing operational problem."
San Francisco-based PowerTouch is developing software that brings together multiple forms of wireless communications including RF-sensors, push-to-talk radio, 800-MHz radio, cell phones, and text messaging.Unmanned vehicles
Another research area for sensor networks is in the control of multiple unmanned vehicles, operating in an automated highway system. This application is currently in the prototype stage. It uses data from sensors on-board the vehicle, from other vehicles, and from a highway infrastructure. Cars would use the sensors as feedback about driving conditions and relationships to other vehicles on the road.
One simulation discovered that communications from the front of a fleet didn't reach the back in a manner timely enough to pass along information about such hazards as curves, potholes, or lane changes. The result: Vehicles at the back of the simulated fleet crashed.
But "The fix was simple," according to Stanford's Goldsmith, "because we understood the needs of the control system. The analysis was that platoons of vehicles were affected by what's called string stability. The problem was solved by going from multihop serial communications where vehicles relayed data down line, to broadcast where all vehicles received information simultaneously."
For more information on the ABB system mentioned, here is a PDF article.Make contact:
Look ma, no wires
Every modern production line is characterized by numerous sensors and actuators. Each one of them demands cabling to handle power and data. This cabling is not only costly to install but also can be a source of failure.
In 2002, automation supplier ABB received the gold European Innovation Award for its wireless sensor technology meant to address these problems. ABB wireless proximity switches make restrictions from cabling on movable machine parts in manufacturing cells a thing of the past.
The wireless sensors do not require power cables but instead draw their power from an electromagnetic field set up in the manufacturing cell. Wireless proximity switches in the cell pick up energy from the ac field inductively using small coils. The sensors also have small radio transceivers for communicating with a special input module. This module behaves like a traditional input module, handling up to 60 wireless proximity switches simultaneously. Five input modules can coexist in the same area to handle up to 300 sensors in one manufacturing cell.