Feature Story

COGNITIVE RADIOS SOLVE A HOST OF PROBLEMS

You fly into an airport for the first time. Your cell phone comes to life, determines that you're in Buenos Aires, switches from CDMA to GSM, finds the right band, and logs you onto the cellular network, negotiating the new roaming arrangement for you. While you're in the airport, it routes your outgoing calls over the free local Wi-Fi network, onto which it logs you without being asked. After directing you to the rental car desk, it gives you directions to your first appointment, constantly updating the GPS map on its screen. When you tell it you'd like to take your guest to a mid-priced Italian restaurant, it finds the nearest one and books a reservation for you.

Is the foregoing a "Star Wars" fantasy? It's currently a daydream, but by no means a fantasy. The enabling technologies are here now, needing only to be stitched together-admittedly adding a lot of intelligence-into a cognitive radio.

What is a cognitive radio?

According to Bruce Fette, PhD, chief scientist for the Communications Networks Division at General Dynamics C4 Systems (www.gdc4s.com), "Cognitive radio is really additional functionality that is added to a software-defined radio (SDR). I personally consider it impractical to do a cognitive radio without starting with SDR. Generally speaking, cognitive features are defined in software, which is why you start with SDR."

Figure 1. SDR hardware architecture.

The IEEE defines a cognitive radio as a radio transmitter/receiver that is designed to intelligently detect whether a particular segment of the radio spectrum is currently in use and to rapidly jump into and out of that frequency range without interfering with the transmissions of others. There's a lot more to it-and certainly behind it-than that.

Joe Mitola of MITRE Corp. (www.mitre.com) is the acknowledged "father of cognitive radio," and he's its earliest and chief proponent. According to Mitola, "I coined the term 'cognitive radio (CR)' to represent the integration of substantial computational intelligence-particularly machine learning, vision, and natural language processing-into software-defined radio. CR embeds an RF-domain intelligent agent as a radio and information access proxy for the user, making a myriad of detailed radio use decisions on behalf of the user (not necessarily of the network) to use the radio spectrum more effectively." Cognitive radios are SDRs with cognitive capabilities. They are-if you will-really smart radios.

Cognitive radios are aware of their location, spectrum use, and even the preferences of their owners. Without relying on prior programming, cognitive radios can change frequency, power level, transmission mode, and modulation characteristics in response to changing conditions. By adapting to real-time spectrum conditions, cognitive radios provide their users with seamless, reliable communications while maximizing spectrum efficiency.

For example, your cell phone might contain a GPS that let it know you were about to enter the Holland Tunnel on your way to work. The cell phone knew from past experience that you'd lose the signal upon entering the tunnel, so it would switch bands and operating modes until you emerged at the other end. When you got to work in Manhattan, it would then detect the presence of your Wi-Fi network and make all outgoing calls through it instead of your cellular carrier. Switching to Bluetooth, it would automatically sync up with your PC's calendar and warn you of any conflicts. At lunch time, it would ask you-using speech recognition-what type of food you like and then show you the location of the nearest Chinese restaurant. It would also make reservations for you if you asked.

If that sounds fanciful, consider that all of the underlying technologies are available today in portable devices. Many high-end cell phones have Wi-Fi, Bluetooth, and GPS in addition to their cellular front ends. Just incorporating GPS into cell phones has already enabled location-based services to take off rapidly. And speech recognition for simple commands is a problem long since solved (text-to-speech is trivial). What's needed is the smarts-incorporated into software-to tie all this together in a way that enables creative new uses of a very sophisticated hardware platform.

Getting there in stages

Mitola described what he calls the "progression of awareness and adaptation toward cognitive radio (AACR) [that] leverages traditionally nonradio technologies: computer vision, navigation, speech recognition and synthesis, and the semantic web." Discrete steps on the path include radios that are at first aware of their environment; then can adapt to changes in it; and finally can anticipate changes in the environment and act based on what they have learned of their user's needs and preferences.

Figure 2. SDR (SCA) software architecture.

SDRs are typically aware of their environment, at least to the extent of monitoring signal strength, signal-to-noise ratio, bit-error rate, adjacent channel interference, and possibly multipath reflections. They're typically programmed to respond in a fixed manner, adapting to the challenges by changing power level, transmission rate, or the direction of a smart antenna. 3G cell phones are already mode adaptive, switching from a high-data-rate, high-quality-of-service (QoS) mode to a low-data-rate, "stay connected" mode when signal levels drop. SDRs might be a bit more sophisticated, creating a software-defined notch or comb filter to knock down adjacent channel interference.

To illustrate the progression, if you add a GPS to a cell phone but don't connect it to the rest of the system, the phone may be more convenient-you know where you are, but the phone doesn't. Once the phone becomes location aware, you can program it to change bands or transmission modes when you arrive at certain locations; the phone then becomes RF-location adaptive. When the phone learns to associate its location (you're near the entrance to the Holland Tunnel), speed (fast = you're in your car), and time of day (7:00AM = you're on the way to the office), it can proactively change bands, transmission modes, power level, and whatever else to enable seamless communication while you're in transit. Now you've got a cognitive radio.

All of this leads to the ideal cognitive radio (iCR), which "transparently accesses useful information via whatever wireless means might be available," explains Mitola. The focus shifts from staying connected to staying informed; from QoS to QoI (quality of information). To that end, Mitola moves XML-which itself extended HTML-into the radio domain in a new meta-language, Radio XML (RXML), which understands and processes RF parameters the same way your browser handles active web pages.

Awareness and learning

A cognitive radio needs to be aware of not just the RF environment but that of its user, too. For example, if a cognitive army field radio were lost, it could use optical sensors to verify its user, backed up by voice recognition. It might also recognize a set of procedures that its owner usually practiced and challenge a potential new user for a password or other form of identification before proceeding. In a medical context, a cognitive radio might query its owner if it sensed a fall or a sudden change in heart rate, calling 911 or the doctor if it didn't receive an appropriate response. User-perception technologies-including computer vision, location information, and speech recognition, in Mitola's words, transform a radio "from a bit pipe to [a] perceptive RF portal."

Preprogramming an SDR to handle a narrow range of scenarios is possible, if tedious; doing so for a wide range of environmental changes is impossible. So cognitive radios need to learn from experience. They can do so in one of two ways: learning by being told and learning by observing.

Learning by being told means learning from other cognitive radios. If a patient from San Francisco who carried a medical cognitive radio moved to Chicago, her radio might learn that the calling frequencies for emergency services were different in the new location, as were the protocols. The San Francisco radio would learn the new protocols from other cognitive radios in Chicago. They'd pass this information along in an RXML format that all could understand. They'd also register the newcomer onto the local ad hoc network of emergency medical cognitive radios so that instant communication would be possible in the future if necessary.

Cognitive radios can also learn by observing use patterns from their operators. They might then either repeat the operations when in an identical situation-about to enter the Holland Tunnel-or they might generalize from one instance to similar ones, a process called instance-based reasoning (IBR). Instead of being trained with a large number of examples, cognitive radios using IBR would generalize from one instance and learn exceptions as they went along.

This is one area where cognitive radios start getting into uncharted waters and raise some significant regulatory issues. Who is going to license a radio that can transmit on any frequency and in any mode it wants, trusting it to always make the right decisions?

The FCC and CR

The U.S. Federal Communications Commission (FCC) and European regulators have given cognitive radio a surprisingly warm welcome, recognizing its potential to help increase spectrum efficiency-such as using the analog TV bands in the US that will open up in the next few years as TV goes digital. FCC 03-322 (Facilitating Opportunities for Flexible, Efficient, and Reliable Spectrum Use Employing Cognitive Radio Technologies) allows TV-aware radios to establish low-power (Part 15) ad hoc wireless networks. These are aware/adaptive radios by the definitions we've been using and not fully cognitive radios, but the FCC ruling is a major step in that direction.

According to Bill Krenik, senior director of wireless research at Texas Instruments (www.ti.com), "If we presume that the radios will be used in an appropriate way, and that they're going to be designed to fall under some standards requirements and will be sanctioned or licensed by the FCC, then we can build a radio that fundamentally has the capability to do a lot of different things on a lot of different bands. But we'll only do those things in those bands when it's appropriate." Just as with today's radios, it will be essential to have a suitable certification procedure to ensure that all deployed handsets meet criteria to avoid harmful interference.

Howard Pace is deputy director of the Joint Program Executive Office (JPEO) for the Joint Tactical Radio System (JTRS) (http://jtrs.spawar.navy.mil/sca/), which is defining and developing the software communications architecture (SCA)-the de facto standard for SDR-for the U.S. military. While actively pursuing the benefits of cognitive radio, he also acknowledges the regulatory challenges. "We currently statically lock up tons of spectrum, and some of it is not always used and lays dormant," says Pace. "If we had the ability to sample the available spectrum, and then put communications where it's not being used, and switch when it was being used without interfering with the primary user of that spectrum, it would have enormous benefits. That type of technology has been experimented with at DARPA and has been talked about widely. But even if I were to bring you that ability in a box today, without policy changes at the ITU and the FCC you couldn't use the technology."

Pace outlines one example: "We designed a wideband networking waveform (WBNW). It can go anywhere from 2 MHz to 2 GHz. But getting certification of that waveform becomes almost an impossible matter, not only because they don't want to certify a waveform that can span such a range, but they want you to specifically limit the waveform to only being able to transmit in those ranges for which you're authorized. So for this technology to be realized, we need to visit not only the technology but also the policy end of allowing ourselves to take advantage of it if that technology proves itself mature."

There may be a workaround, though not a simple one. General Dynamics' Fette is actively involved with the SDR Forum in trying to get cognitive radio over the regulatory hurdles. "As soon as you touch spectrum, the regulators will have something to say. The cognitive radio committee is working very hard to structure the radio architectures so that the regulatory community can provide the radio with a machine-understandable policy that reflects what the regulators feel the radio must do to be compliant," Fette explains. In short, cognitive radio suppliers would have to encapsulate all the FCC spectrum use regulations into a machine-readable database that would govern the operations of a cognitive radio. The cognitive radio could do whatever it deemed appropriate while still staying within regulatory bounds. "The notion of providing [regulators] with policy actions in which they insert their regulatory requirements is one efficient way for the regulators to interact with radios that exhibit cognitive behavior."

The FCC may not be the only ones wanting a say in cognitive radio, according to Fette. "The network operators may also have policies that they want the radios to observe. I can envision there being policies associated with the network operators, the regulators, and possibly others; so each of the people in the value chain may have a piece of the specified policies that regulate cognitive radios."

Fette's suggestion is probably the way cognitive radio will deal with spectrum rules. In Europe, regulators seem not to be too worried. Mitola told Portable Design about attending an ITU meeting in Europe. According to Mitola, "The regulators held a workshop in Europe recently, and they were pretty blasé about the whole thing. They said basically, 'All we need to know is which manufacturer to fault.'"

Cognitive radio in public safety

The military aside, the biggest need for SDR is among public safety organizations. Police, fire, and ambulance units in the same county can rarely communicate directly with each other, needing to go through a central dispatcher. At the recent SDR Forum Technical Conference, public safety officials from Florida and Louisiana described the communications chaos following in the wake of recent hurricanes. They were adamant that they needed cognitive radio capabilities ASAP.

Cognitive radios have enough built-in intelligence to communicate on a machine-to-machine level; identify themselves and their communications capabilities; form ad hoc networks, all adopting common frequency bands and transmission techniques; and log onto these networks as their status allows. Public safety radios would all have encrypted security codes that enabled them to log onto public safety networks; ham radio and CB operators would only be allowed onto those networks with special permission, and then they'd still be restricted to transmitting on their own bands while listening to the public safety channels. Those communicating with them from the public safety channels would switch their receiver frequencies automatically at the direction of the cognitive radios on the network. The radios would automatically set up and manage the network for the operators.

According to JTRS's Pace, "APCO 25 has always been the first responders' answer to what they should use for interoperability between themselves. My suggestion is that as more data becomes available, it's more and more important that APCO 25 should be replaced by something that has a data capability based on an SDR." Not to mention the intelligence and flexibility of full cognitive radios.

Return to: 2007 Feature Stories