ECE: Electrical & Computer Engineering
ECE News

Radios that Think and Learn

Print Version

image of article

(388KB .pdf) Print Version

Like 9/11, Katrina showed how emergency response is seriously hindered when responders cannot talk to each other because of incompatible communications equipment. When police, fire, medical, and other responders arrive from different jurisdictions, they bring their own equipment. The different radios are purchased to meet the budget, geography, and political constraints of their own regions and typically use different frequencies and competing waveform standards.

Photograph of Charles Bostian

Charles Bostian, holding a GNU software radio board, believes that the Virginia Tech cognitive radio project will change the very way that radio engineering is done. It was the glow of vacuum tubes, like those superimposed in the background, that first attracted Bostian to radio engineering.

An ECE-led team believes it can solve these interoperability problems — without buying new radios for every public safety organization in the country. Their goal: to build radio software that works with off-the-shelf equipment to automatically find an open, legal frequency, and establish communication with another radio and/or a base station. With grants from the National Institute of Justice (NIJ) and the National Science Foundation (NSF), the team is working to produce functional prototypes by next year.

Radios that think for themselves

“The radio must think for itself,” says Alumni Distinguished Professor Charles Bostian, who serves as the faculty lead on the effort. “Emergency responders emphatically do not want a radio that requires hands-on adjustment by an expert. They need a radio that is smart enough to find the best path of opportunity, configure itself, and communicate — all with minimal human intervention.”

Bostian’s team believes the answer is a platform-independent cognitive radio system. A cognitive radio, according to Bostian, combines artificial intelligence with software defined radio (SDR) technology to create a transceiver that is aware of the RF environment, its own capabilities, policies that define legal operation, and its user’s needs and operating privileges. Unlike adaptive radios that adapt to anticipated events, cognitive radios learn from their experience and can function in unanticipated situations.

“A cognitive radio consists of a cognitive engine — a hardware-independent software package — controlling an SDR,” he explains. “The cognitive engine sets the SDR’s operating parameters, or turns the knobs, then observes the results, or reads the meters, and optimizes its operation within the governing rules.”

The proprietary Virginia Tech cognitive engine is based on genetic algorithms in a network that was developed for an earlier NSF disaster communications project. The algorithms are modeled on human learning and incorporate logic, randomness, and adaptive memory. Tech’s Wireless System Genetic Algorithm (WSGA) optimizes and adapts the communications system, while the Cognitive System Monitor (CSM) handles the cognitive functions, short- and long-term memory, and control.

Overcoming AI prejudice

“Our first important breakthrough was this development of a computationally efficient way of implementing rapid machine learning in a trial-and-error process based on using genetic algorithms,” Bostian says. “We used a proof-of-concept prototype of our cognitive engine to control a ‘dumb’ legacy ‘hardware’ radio. The resulting cognitive radio could identify the presence of a jammer and change its modulation index, transmitter power, and FEC coding in a way that minimized the effects of the jammer. This prototype demonstrated learning, and we were off.”

When he explains the cognitive radio engine and its genetic algorithms, Bostian describes his previous prejudice that artificial intelligence was all hype and little results. His students, however, were correctly convinced it was the solution and “pulled me into an entirely new area of research — at a time when I could have coasted into retirement.”

Instead of retiring, he says he is having the greatest fun of his career. “We are changing the way radio engineering is done,” he says. Moreover, although he has had many excellent students over the years, “this particular group of students stands out. Every single person on the team is exceptional,” he says. In addition to setting much of the technical direction, the students also actively develop proposals and are pushing to commercialize the technology.

The team’s two current projects will provide a large step toward commercialization. The NIJ effort involves building a radio that can recognize and interoperate with three commonly used and mutually incompatible public safety waveform standards. The NSF effort extends the technology to investigate spectrum access and to study networks that contain both legacy and cognitive radios.

Public safety prototypes

Photograph of a man holding a walkie talkieThe team has been working since August to develop a cognitive radio for public service use that can operate with other radios, or serve as a bridge for older radios that do not have the software capability. Funded by a $420,000 grant from the NIJ, the prototype will run on a software-defined radio that is being built by Innovative Wireless Technologies.

In addition to developing and validating the software and interfacing it to the radio platforms, the challenges include obtaining the first FCC certification for a cognitive radio and developing a user interface. “If this radio is not easy-to-use for the responder in the field, it will not be used and we will have failed,” Bostian says, adding that it is his first engineering project where ease of use by non-technical people is a critical issue.

This past fall found team members tagging along with police at Hokie football games, observing the officers and their communications, and gathering data. “Interoperability is a big problem even for the Virginia Tech police,” Bostian explains. “At every football game, up to 135 officers are brought in from other jurisdictions. Most of the radios from different agencies cannot talk to each other.”

With human factors playing a role in the project, industrial and systems engineering professors Tonya Smith-Jackson and Woodrow Winchester have joined the team. With the further addition of economist Sheryl Ball, the team is starting work with economic theory to help create a system that can learn the user’s needs based on experimental cost-modeling analysis. “This work is preliminary right now,” Bostian says, “but we feel we can learn user needs as an input to the case-based optimization system of the cognitive radio.”

Cognitive radio network behavior

While the NIJ effort is aimed at establishing interoperability and user needs, the team is also studying cognitive radio networks and dynamic spectrum allocation, with a $750,000 grant from the NSF NetS Programmable Wireless Information Networks program. The NSF project uses the GNU radio software, a lower-cost system than the public safety radios. Another prototype is being developed to allow WiFi-like unlicensed operation, using unoccupied TV channels as a test case.

As television broadcast completes its transition to digital TV, the issue with unoccupied channels is growing. One suggestion includes allowing cognitive radios to use the channels. A cognitive radio would know its location and know which channels were potentially available. “It would then listen in those channels and identify licensed users who might be operating there and other access points like itself,” Bostian explains.

Photograph of the cognitive radio team

The cognitive radio team draws its members from many backgrounds. From left: Lariss Marple, ECE undergraduate; Woodrow Winchester, assistant professor, ISE; Bin Le, ECE Ph.D. student; Younhseng (Sam) Shi, ECE Ph.D. student; Allen MacKenzie, assistant professor, ECE and co-investigator; Michael Hsiao, ECE associate professor and co-investigator; Akilah Hugine, ECE master's student; Adam Ferguson, economics Ph.D. student; Tom Rondeau, ECE Ph.D. student; and Daniel Friend, ECE Ph.D. student.

“It would configure itself to avoid causing any interference to the licensed users and negotiate with its unlicensed peers to find a way to share the channel with minimal interference.”

The NSF effort is also aimed at understanding the behavior of networks that include both cognitive radios and legacy systems. “Each cognitive radio will have developed its own information. When placed in a network with other CRs, how will it share its knowledge to minimize interference and power use while maximizing network quality of service?”

1,000 points

After proving interoperability and studying network behavior, Bostian’s team wants to deploy the technology on a testbed with more than 1,000 nodes. “We see an opportunity to deploy our engine on more than 1,000 nodes and experiment with the largest and most intelligent wireless network testbed ever built,” he says.

“Imagine a hand-held radio that offers advanced multimedia services,” he continues. “Now picture more than 1,000 of these in the hands of Virginia Tech faculty, staff, and students. Think of what we can learn about network behavior, spectrum access, and human-computer interaction!”

What the team ultimately imagines, however, is five to 10 years from now, emergency response teams from all over the country being able to communicate during any situation and to minimize damage from disasters like Katrina.

When disaster hits the communications infrastructure

Photograph of hurricane ravaged New OrleansMuch of today’s communications depends on small, low-powered wireless devices, like cell phones. Small, however, limits the distance and power of communications links. Small devices have short antennas, which mean high frequencies, which mean line-of-sight propagation. Small devices also have long battery life, which means low power, which leads to short range capabilities — even less than the line of sight distance. This works as long as the infrastructure exists to support it.

All of these things happened in New Orleans. And on 9/11.

What happens after infrastructure failure?

Virginia Tech solution:

Build and deploy drop-in WiFi-like systems with base stations on balloons or slowly orbiting aircraft. This concept was the key to developing the Virginia Tech cognitive radio - a radio that can automatically find and establish open channels with other radios or base stations. The cognitive radio is under development with rugged, commercial units available on the street in five to 10 years.