Initial testing of a cognitive radio that military officials plan to be sold for less than $500 per unit met expectations, and the project remains on track for realizing its goal of completing development of the devices in late 2010, according to an official for the Defense Advanced Research Projects Agency (DARPA).

In November, DARPA conducted a “very, very successful” demonstration using the first 12 nodes—not packaged in the ultimate radio form factor, although the cards have been reduced to the proper size—built in the project, said project manager Preston Marshall.

“The tests went very well--normally, something goes wrong, but nothing went wrong,” Marshall said. “It satisfied me that we don’t have any insurmountable problem that we didn’t see and that there’s nothing but engineering issues between here and there. I’m sure we’ll have a few [challenges], but there’s no question that what we needed to have clearly is there.”

Among other features, the test demonstrated dynamic spectrum access, a fundamental characteristic of cognitive radios that allow all communicating radios to locate available spectrum and switch to open frequencies if they encounter interference. In addition to operating in traditional peer-to-peer mode for voice communications, each radio served as a node in an ad-hoc data network, Marshall said.

This ability to access IP networks enables the radios to extend their effective range if even one radio in the ad hoc network is connected to the Internet, he said.

“If I’m carrying a push-to-talk radio in the field and I come across several of my own radios, I find one that has access to the Internet, it runs across the Internet and gets delivered at the far end to another set of radios that then relay it to an ultimate radio,” Marshall said. “Instead of having discreet trunking systems like we think of them now, any IP network becomes a trunking system.

“Today, we think of push-to-talk radios as having the range of the radio. The first step was that it actually has the range of any connected node to any connected node to any connected node through intermediate relay. With IP, we now have the ability to say that its range is essentially worldwide—any one of us who are contacted to IP can get to these radios, and they’re completely interoperable.”

For voice communications, this system is effective because tests demonstrate that push-to-talk voice can be transmitted over three intermediate “hops” before reaching an IP network while adding less than 100 milliseconds of latency to the transmission, Marshall said.

On the data side of the equation, a new technology called disruptive-tolerant networking (DTN) ensures that no packets in a transmission are lost, even if a connection is interrupted for a period of time because a user moved into an area where the radio is outside of the signal range. Each radio is expected to include several gigabytes of memory so it can store all information it may need to relay, Marshall said.

“It’s one of the features we wanted, because people walk behind buildings, or they walk behind metal vehicles—there’s always some disruption on the battlefields or, in [public safety] rescue scenarios,” he said. “You want the system to keep working despite those.”

In addition, the radios demonstrated the ability to form an infrastructureless network automatically and without preplanned spectrum coordination, Marshall said.

“One of the things we wanted to do was never have a network fail because of human planning error,” he said. “This network had no planning before it was deployed; all of its planning is based on what it perceives.”

In April or May, DARPA plans to conclude the first phase of the project by conducting a larger demonstration of the technology with the devices in their handheld form factor and with users that haven’t been trained to use the radio, Marshall said.

“After we get through May, then we start to move to the next phase, where we start to create something that’s never really existed, which is this total peer-to-peer management of content,” he said. “We first have to build networks, and then we start building things that look more like web services.”

Marshall said he believes DARPA is on track to meet the project goal of developing a cognitive radio by late 2010 that vendors can profitably sell for less than $500, if the vendor is confident the market will buy at least 100,000 units. DARPA is hopeful public-safety entities will adopt the technology, which would increase the size of the market and lead to a decrease in the cost of the equipment, he said.

“We benefit if there’s a broader market for this equipment,” Marshall said.