The concept of software-defined radios is not new. For nearly two decades, questions about the feature set, spectrum management, and overall radio management constantly alluded answers. Very smart designers and engineers were not able to find the right balance of knobs and switches that would enable the software-defined radio to operate effectively.
Until now.
Driven by research, collaboration, and a nice $2 million prize, several groups of engineers and researchers worked to provide an answer to the following question: could AI-powered software-defined radios operate in a flexible, ad-hoc spectrum sharing model without the guidance or limitations imposed by FCC?
The Defense Advanced Research Projects Agency’s (DARPA) posed this question through the Spectrum Collaboration Challenge. DARPA provided the platform and financial incentive to entice research into and development of radios that would be able to automatically adapt to changing conditions without human intervention while efficiently using a common spectrum.
Fifteen teams in the second year and ten teams in year three accepted the challenge to design, develop, and compete to create a software-defined radio that would meet these objectives. The competition between the teams was very close, as the winning team – Gatorwings, from the University of Florida, won by a single point.
The collaboration amongst radios was one primary requirement in the competition – the competing radios “told” each other how it was using the spectrum for their traffic. This communication enabled the other radios to use the available spectrum not currently being used by other radios. At the same time, each radio demonstrated that it could provide stable service and preserve connectivity to existing users. At the conclusion of the competition, DARPA Program Manager Paul Tilghman stated, “for the first time, we’ve seen AI-enabled radios…unlock the true potential of the RF spectrum.” The competition results indicate that these radios provided better performance in the 3.5GHz spectrum than LTE does. Through AI-powered dynamic management, these radios can stuff up to 3.5 times more traffic in the spectrum than the amount of traffic that dedicated spectrum channels enable for current LTE radios.
The results prove that software-defined radios are able to improve performance in the high-value spectrum that operators lease from the FCC. However, the interesting question becomes – with dynamically managed spectrum through these AI-powered radios, is there a future where network operators will not need an exclusive, licensed spectrum? Maybe, but not likely. A bigger question is that of costs – will software-defined radios have traditionally incurred high costs. Will they be able to meet price points and cost-per-bit values that today’s innovations are delivering on full system-on-chips? These current advancements in hardware are not only reducing the size of radios but are also increasing their capacity and reducing their cost-per-bit.
Congratulations to all the teams for their 3-year commitment and work that proves that AI-powered software-defined radios can more efficiently and effectively use spectrum, a resource that many in the wireless industry call scarce. The successes achieved and failures overcome by the participants highlight that new radio solutions can meet the challenges of dynamic real-time spectrum management. Now, it’s a matter of implementation and identifying beneficial markets and use cases.
Photo by DARPA