Radios Get Smart – IEEE Spectrum


Illustration: John Hersey

Guglielmo Marconi’s first radio broadcast was a point-to-point affair, a transmission from England to one person in Nova Scotia. It was the radio equivalent of flying an airline’s passengers across the Atlantic one at a time. Today, we pack radio waves into the air as tightly as economy seats on a 747.

We need to. New radio technologies keep coming along—Wi-Fi, WiMax, Bluetooth, ZigBee, the growing panoply of cellular voice and digital services, broadcast satellite, and more. Each requires unique hardware appropriate to its special way of sending and receiving radio waves. To minimize interference, each is restricted to specific bands of the spectrum and is limited in the way it divides that spectrum into channels and in the encoding and modulation schemes it can use. If we tried to impose this sort of regime on road transportation, we might wind up with a system in which buses, cars, and trucks were each restricted to their own separate roadways.

Some of these rules can be justified by the fact that certain parts of the spectrum work best for certain radio applications. But others are merely the legacies of telecommunications history. Spectrum has long been allocated in a first-come, first-served way, with broadcast radio and television in particular getting the most desirable bands. Even though in much of the world cable and satellite television eclipsed broadcast television many years ago, broadcast’s control of prime spectral real estate will most likely continue for at least the next decade.

If radios could somehow use a portion of the broadcast TV spectrum without causing interference, cellular telephony and other important services would be able to exploit those bands, too. With more room to operate, cellphone calls would be a lot cheaper, as would mobile Internet access, which would get faster as well. Your handset would be able to pull in audio and visual entertainment from all over the globe, and videophoning would finally be a reality.

To manage such feats, cellphone handsets would have to be able to shift their frequency of operation on demand and without packing in lots of extra hardware. Telecommunications engineers have a name for that goal—software-defined radio. And the more visionary among them see it as a stepping-stone to an even more distant ideal. Their goal is a wireless device that is smart enough to analyze the radio environment and decide for itself the best spectral band and protocol to reach whatever base station it needs to communicate with, at the lowest level of power consumption.

The name for such remarkable systems is “cognitive radios,” and some are already emerging from the laboratory to be field-tested by the U.S. military, which has long sponsored research in the area.

Some analysts say it’s only a matter of time before cognitive radios get into the commercial arena, because the economics are compelling. Indeed, many experts say a typical cellphone a decade or two from now will have cognitive features.

Here’s why: widespread use of cognitive radios could make more efficient use of radio spectrum. Estimates of how much additional traffic the airwaves could hold vary, but by some accounts, less than 14 percent of radio spectrum is truly busy at any given time. That includes big chunks of spectrum that are assigned but that aren’t fully used. Prime among them are the upper ranges of the TV bands: channels 14 to 83, better known as ultrahigh frequency or UHF. In 2004, a study by the International Telecommunication Union, in Geneva, found that “many TV channels are unused over significant geographical areas” and concluded that “cognitive radio techniques appear to be a promising approach” for using spectrum more efficiently while avoiding interference with current operations.

It’s no surprise, then, that the possibility of supersmart radios has attracted the attention of governmental regulatory bodies on both sides of the Atlantic. The U.S. Federal Communications Commission (FCC) and its UK counterpart, the Office of Communications (Ofcom), have launched technical reviews of cognitive-radio technologies as a way of better managing the scarce resource that spectrum has become. As we’ll see, however, the two agencies couldn’t be more different in their approaches.

Deciding which portion of the spectrum to use at any given moment is only one aspect of what a cognitive radio could do. Adding audio and video sensors as well as speech- and vision-recognition capabilities, the radio handset could, for example, act as a health aid for the elderly. “It could detect heart palpitations; if necessary, it could ask the user ‘Do you feel okay?’ ” says Joseph Mitola III, a consulting scientist with Mitre Corp., in Bedford, Mass. “If you stuck a shock-vibration sensor on the cellphone and it detected a clonking as a person fell to the floor, it could dial emergency 911 and ask for help.” Mitola has the distinction of having coined the names for both software radio and cognitive radio.

To understand cognitive radio better, first consider how it improves on software-defined radio. A software-defined radio can easily switch among multiple wireless protocols or move to different frequencies, waveforms, protocols, or applications, but the user must command it to do so. Going back to the transportation analogy, software-defined radios are like ordinary cars, which need a driver to decide when to brake, speed up, and change lanes. Cognitive radio, on the other hand, is akin to having the car drive autonomously, making all the decisions itself. Cognitive radios not only sense the spectrum and determine how best to interface with one another, they also decide which radio network is best to use.

Today’s cellphones are already marvels of miniaturization, but fundamentally, they are still just plain old radios. As you speak, the analog sounds are transformed into a stream of bits that will be conveyed through the air as variations in the frequency, phase, or amplitude of the transmitted carrier waveform. Nowadays, protocols for multiple access are used to divide up a portion of the radio spectrum and to use it more efficiently. Some share a slice of the spectrum temporally, giving, say, three phone calls a one-third time slot each, alternating between them rapidly. GSM uses a sophisticated version of this idea.

A code-division multiplexing transmitter breaks a band of spectrum into many spread spectrum channels and then employs correlation techniques to receive each of the channels simultaneously and separate them into the individual calls. Regardless of the protocol, the receiving radio’s antenna, analog-to-digital converter, and other circuitry expect waveforms to be transmitted in a particular band of frequencies, using a prescribed modulation type and a well-defined encoding scheme.

A software-defined radio uses an analog-to-digital converter capable of capturing waveforms from a much wider range of spectrum. The U.S. military’s first software radios operated between 4 megahertz and 400 MHz. Its newer ones stretch both end points, operating between 2 MHz and 2 gigahertz.

In practice, only a few bands will have strong signals. A spectrum analyzer would show various peaks, with valleys in between. In practice, only certain bands are of interest, such as the 1900 MHz typically used by GSM, or 1.6 GHz used by some satellite phones. After analog-to-digital conversion, the signals in the bands are then converted to a more convenient one, known as the intermediate frequency, and then converted again to a very low frequency called baseband for decoding and demodulation by general-purpose microprocessors or digital signal processors [see diagram, “Soft Sell”]. To transmit, the process is reversed.

Someday, 10 to 15 years from now, all cellphone handsets probably will be based on software-defined radios; perhaps soon after that, cognitive radios will start to dominate. Physics, as well as economics, will dictate the changes. “The energy consumed in a handset is growing by 30 percent a year; battery capacity is improving at 5 percent a year,” says Liesbet Van der Perre, scientific director at IMEC, the Interuniversity Microelectronics Center, in Leuven, Belgium. IMEC is exploring how software-defined radio can reduce a handset’s power consumption; worldwide, at least a couple of dozen companies are already hard at work pursuing that goal.

Those cellphones will be lighter and cheaper, and will last longer on a single battery charge, but they won’t solve the problem of overcrowded airwaves. To return to the highway analogy, what we need is to close all the gaps between cars and to use nearby secondary roads when the highways are jammed with traffic. That’s what cognitive radio can offer.

Such radios will figure out which bands are underused, rendezvous with one another there, and start communicating. If the primary occupant of that piece of the spectrum starts transmitting—maybe it’s one of those UHF television channels beginning its daily broadcasts—the cognitive radios will hop to another frequency, with neither the calling parties nor any of the TV viewers any the wiser. To do that, their circuitry has to be significantly more sensitive than the television set’s receiver, notes Bruce Fette, chief scientist for the communication networks division at General Dynamics Corp. and the editor of a 2006 reference book, Cognitive Radio Technology (Newnes).

To know whether a band is underutilized, the cognitive-radio base station must have a field marshal’s overview of the spectral combat zone from which to devise complicated plans of attack. And as more and more end-user devices are able to run up and down the spectrum, those plans will become ever more difficult to carry out. Ideally, the base station will also know that more TV stations are transmitting at 8 p.m. than at any other time, that more phones calls are made at 9 a.m. than at 8:30 a.m., and that more calls are made on Mondays than on Sundays, except for Mother’s Day.

If the cognitive-radio handset is to help out, it will have to determine exactly where in the world it is. Regulations differ from country to country, and not all use the same frequencies for the same technologies. A GSM cellular service in the United States operates on different bands from those used in Europe, for example. So mobile cognitive radios will also have to decipher satellite geolocation signals or query base stations to find out where they are.

As the power of microprocessors continues to increase, cognitive radios undoubtedly will start to learn the habits of their users and make educated guesses about how long a phone call is likely to last or which hours in a day are the busiest ones. (And if cognitive cellphones become that smart, maybe they’ll even stop themselves from ringing in a church or a movie theater.)

The challenges of true software-defined cellular systems, to say nothing of fully cognitive ones, are enormous. Doing more in software paradoxically requires much better hardware—specifically, the general-purpose processors in the radios will have to interpret many more signals coming from a wider swath of the spectrum, and they will have to do far more encoding and decoding, and signal analysis.

Processing signals at high frequencies makes such tasks even more difficult. For one thing, the higher you go, the tougher it is to push through walls, trees, hills, or other solid barriers, and—at the highest frequencies—the very air itself, making the signal harder to find among the noise. Only recently has Moore’s Law given us sufficiently powerful processors at reasonable prices.

John Polson, an engineer with Bell Helicopter Textron, in Fort Worth, Texas, believes that all the pieces to make the leap from software radios to cognitive radios already exist. They include the ability of a handset to locate itself with GPS, to sense and analyze nearby spectrum, to know the date and time, to detect patterns and even biometric information from users, and to contain and manage a database of nations and regulations. Polson thinks the biggest challenge is designing clever algorithms that will take all that information and make decisions about where in the spectrum to operate at any given moment.

Before those decisions can be made, however, the two radios needing to communicate have to link up with each other on a single frequency among many possible ones. If both cognitive radios are mobile, the task is especially challenging. Polson compares it to two people with flashlights “searching for each other when they are a mile apart on a moonless, dark night out in the desert.” The beams may be visible, but only if the two people face each other at the same time do they have a chance to find each other. Polson says the cognitive radios will have to send out “probe” signals using a unique, easily seen waveform. But there’s still a problem if the two radios start out on different frequencies.

The situation is made worse by the fact that a frequency might seem available to one of the radios but not to the other. That happens when one radio is out of range of a primary user—say, a TV broadcaster—but the other is not. Then the frequency isn’t one that the two cognitive radios can use—but one radio doesn’t know that.

Neither radio can simply sit on a frequency and wait, because that frequency might be one the other won’t transmit on. So they each have to hop around, looking for the other. Polson says different strategies of sending out probe signals and waiting can be employed, depending on such considerations as the range of candidate frequencies. One involves “listening” for a length of time three times as long as it takes to send the probe signal.

Eventually, one node responds to the other’s overture. The two then exchange other signals to describe their perceptions of available frequencies. If one node knows that a frequency is in use by a transmitter the other can’t see, that slice of the spectrum is vetoed. When a frequency is finally agreed on, one of the nodes proposes a rendezvous, in terms of time, data rate, and waveform, allowing a communication session to finally begin.

All the software needed to make cognitive radio possible will go to waste if prevailing regulations don’t allow it to be implemented. Right now, governments give broadcasters and carriers exclusive use of the frequencies they are allocated.

Those regulations create scarcity, says Martin Cooper, a veteran of the wireless industry and inventor of the first cellphone, a 1-kilogram brick, in 1973. Cooper, now the executive chairman of Arraycom, in San Jose, is adamant in his view that spectrum is not inherently scarce. He cites the extraordinary progress made in spectrum usage (he calls it Cooper’s Law): the number of calls or data transactions conducted over useful spectrum has doubled every 30 months for the last 105 years, ever since Marconi set the ball rolling with his first radio transmission. “Almost all the improvement is due to the ability to reuse spectrum geographically,” says Cooper, whose company specializes in smart antennas that aid the sharing of spectrum by focusing signals in one direction.

A frequency used in one place can be recycled elsewhere as long as the signals fade sufficiently between the two locations. That is the basis of cellular telephony, where each base station covers a small area—the eponymous cell—and operates on a particular frequency, one chosen to avoid interfering with adjacent cells. That is, cells using the same frequency must be sufficiently far apart.

Telecom consultant Paul Kolodzy, a former FCC spectrum-policy advisor, notes that even if all the choice bands have been allocated, it doesn’t mean they are full. Probably less than 5 ­percent is being used at any given moment, he estimates.

None of this has escaped the attention of the FCC, which sees cognitive radio as an enabler of spectrum sharing. The agency issued two requests for industry comment in 2003 and 2004, the first regarding cognitive radio and the second the sharing of TV bands. The effect was to throw the spotlight on cognitive radio—until then confined to the laboratories—and create the first commercial opportunity for its deployment.

One person who responded to the FCC’s request regarding TV-band sharing was Carl R. Stevenson. “They were talking about using this unused TV spectrum, so we thought maybe we should be proactive and figure out more precisely what would be its best use,” says Stevenson, who chairs an IEEE working group, 802.22, that was formed to study the question. The group is developing standards for hardware and software based on cognitive-radio technologies that will allow wireless broadband services to share TV spectrum, without interference, in sparsely populated areas that do not have cable or digital subscriber line (DSL) service nor are likely to get it. “Protecting the licensed services is of paramount importance, so a cognitive approach was decided on,” Stevenson says. “There is a need to make sure TV signals and other licensed uses are not interfered with.”

The radio’s software must be able to sense the spectrum and characterize what it finds. It then “adapts”—deciding the best course of action and adjusting accordingly. Finally, it “learns,” by building a database of past actions—which it consults when the radio environment changes again.

As the 802.22 group envisions it, a base station will take the data that the surrounding user terminals collect and decide which channels in the area can be used without causing interference. In turn, during idle times, the terminals will survey frequencies adjacent to the one that the base station selects to build up a picture of free channels, details of which will be sent out to all units. If an incumbent signal then appears, the terminals will move to another frequency or shut themselves down in an orderly way.

Moreover, “a user terminal is not allowed to transmit without permission from the base station,” Stevenson says. Because the band being used is licensed to television, much is known a priori about channel locations and their signal characteristics.

There’s one other complication the group has to worry about, though: TV stations also use secondary low-power auxiliary transmitters—for example, for a wireless link to a referee at a ball game. Because the secondary low-power broadcasts are local and intermittent, they present a challenging problem.

The 802.22 group expects to complete its standard by next year, likely making it the first commercial scheme available for cognitive radio. Stevenson describes the 802.22 standard being developed as “cognitive radio–lite” because it is targeted at a specific band, with a limited set of incumbents, rather than trying to “be everything to everyone, everywhere.”

Not everyone is ready to open up the airwaves. UK regulator Ofcom, in particular, seems to be heading in the opposite direction. Whereas cognitive radio breaks down the barriers between different uses—TV, AM/FM, cellular, Wi-Fi, and the rest—or at least makes those barriers less important, Ofcom is sharpening them. The regulator is slicing up the spectral pie in ever-finer wedges and giving companies exclusive control of the ones they come to “own.”

Instead of filling underused bands with smarter and smarter radios, Ofcom plans to deregulate the airwaves in such a way that the licensee can relicense some of its rights to other parties, which would then have exclusive use of certain frequencies.

Ofcom expects to convert more than 70 percent of the UK’s spectrum to the new regime by 2010; in 2004 no such arrangements were in place. Ofcom, in short, is setting up a major roadblock for cognitive radio.

“We are suspicious of cognitive radio because of the hidden-terminal problem,” says William Webb, R&D chief at Ofcom. The terminal—a cellphone, for example—can be hidden because a building stands between it and the cognitive radio that is vying to use the same frequency. No amount of sensitivity by the cognitive radio’s receiver will uncover it. “You can use databases and network information between radios, but you can’t be 100 percent sure,” Webb says. Ofcom says it’s therefore inappropriate to introduce cognitive radios in spectrum owned by others—that is, unless the incumbents themselves decide they want to allow them.

The U.S. military, on the other hand, is less concerned with hidden terminals than with the fact that it currently uses more than 40 different types of radios. Many of them are highly specialized, such as one employed exclusively to communicate with Army rotary-wing aircraft. Another is used only for voice communications between infantry soldiers.

The military’s Joint Tactical Radio System initiative is an R&D project designed to create a software-defined radio that can unify many, if not all, of these applications. The Navy is already using a ­software-defined radio on some submarines and ­surface ships. The Digital Modular Radio, as it is called, replaces many different sets of gear the Navy previously deployed.

As for cognitive radios, prototypes are now ready for field-testing. The Defense Advanced Research Projects Agency is evaluating the fruits of its cognitive-radio next-generation project. Its goal is to demonstrate that radios can access spectrum dynamically without causing interference.

The military’s interest in cognitive radio stems from its need to set up communications in any environment—not all of them battlefields. For example, it conducted humanitarian operations in the Philippines last February after a disastrous mudslide. Technicians set up communications links with local agencies while complying with the country’s regulatory policies.

Of course, the military is also interested in the uses of cognitive radio in battle, during which it wouldn’t have to adhere to any spectrum policies. Nonetheless, the radios must communicate in hostile electromagnetic environments, avoid jamming, and change encryption methods on the fly—all of which are better done through software.

Whether it takes five years or 15, cognitive radio is coming to both military and civilian life, and it promises to alter the telecommunications landscape.

Walter Tuttlebee, executive director of Mobile VCE, an industry-sponsored research organization in Basingstoke, England, expects the sophistication of radio cognition to advance in stages, just as human cognition blossoms in early life. “A fully developed mature human being is very different from a 1-week-old child,” he says. “Yes, if cognitive radio could mature into a 70-year-old gentleman tomorrow, it would revolutionize the wireless industry. But what we have is a very young baby.”

About the Author

Roy Rubenstein is a technology journalist based in Israel.

To Probe Further

Details of DARPA’s Wireless Network After Next project are at

Joseph Mitola III’s newest book, Cognitive Radio Architecture: The Engineering Foundations of Radio XML, was published in September by Wiley. Cognitive Radio Technology, edited by Bruce Fette, was also published last year, by Newnes.

To participate in cognitive radio developments, see