Veni, vidi, vici: I came, I saw, I conquered. Which of the 10 teams participating in the Spectrum Collaboration Challenge (SC2) finals uttered Julius Caesar’s famous words as they emerged victorious from DARPA’s Colosseum?
This colosseum is not found in imperial ancient Rome, but in the modern surrounds of the Johns Hopkins University Applied Physics Laboratory in Laurel, Maryland. October 2019 saw the culmination of the three-year-long competition, which pitted teams of researchers against each other in a battle to dominate the electromagnetic spectrum by maximizing spectral efficiency.
As the wireless spectrum becomes more congested—being accessed by more devices, with a greater demand for wide bandwidths and high data rates—it’s not sufficient for EM (electromagnetic) systems solely to expand into untapped portions of higher-frequency spectrum. To achieve effective communication, it’s key that systems collaboratively share spectrum, and continue to operate reliably in contested spectral environments.
With Colosseum, DARPA aimed to create a virtual proving ground on which AI-based algorithms could be put through their paces. Teams of researchers are tasked with generating a radio paradigm that provides stable service, reuses spectrum, prioritizes traffic, handles surges, and operates in congested environments, so Colosseum needed to dynamically emulate a complete EM environment with four key characteristics:
Large scale: To study radio interoperability in real time, more than 100 software defined radios (SDRs) should mimic a realistic RF environment.
Full mesh: Every radio must be able to hear every other radio, as they would over the air in a set geographic area.
Wideband: Colosseum emulates wireless interactions across a wide, yet constrained, bandwidth of 80 MHz.
Neighborhood-sized: It required a ~1 km2 congested spectrum environment.
To achieve this, Colosseum comprised two overarching components: A pool of SDR resources, which SC2 competitors use as a common platform to build their intelligent radios; and a massive channel emulator (MCHEM) that emulates radio wave interactions in the physical world. From any one radio’s perspective, it appears to be operating in an open-air environment.
The radios are off-the-shelf Ettus Research USRP X310 SDRs connected to commodity rack servers. Developing the channel emulator required technologies capable of huge data throughput and processing. Commercial off-the-shelf processing and data movement technologies kept costs from escalating.
In total, 256 radio inputs and outputs combine to 20.48 GHz of RF input and output bandwidth, summing to a gargantuan 102.4 GB/s of data throughput. It is no mean feat handling this volume of data, so 64 overclocked FPGAs, across 16 ATCA-3671 units, perform the required digital signal processing.
DARPA’s Colosseum MCHEM, with 128 USRP SDRs, was transported coast-to-coast to LA for the SC2 grand finale.
Almost three years ago, more than 30 teams set out to take on the SC2, beginning in a preliminary round. 15 teams progressed to the second round, before being whittled down to the final 10 that participated in the finale. Teams flew in from as far as Belgium to wrestle for the $2 million top prize.'
The entire Colosseum system was dismantled, transported across the US, and reassembled so that the final could take place live on the keynote stage of MWC Los Angeles 2019, on October 23. View it on-demand at https://www.spectrumcollaborationchallenge.com/.
GNU Radio Foundation president Ben Hilburn commentates on a tense match in the “Alleyways of Austin.”
On stage, DARPA program manager Paul Tilghman asked, “Can we make radios that are smart enough to manage spectrum themselves?” The challenge’s main success criterion was to create autonomous systems that can beat the status quo of manual, human-controlled spectrum allocation. This was tested by running rounds of “matches” in simulated environments, such as a military operation in an urban environment; a natural disaster relief scenario; controlling a wildfire; and a day at the mall. Teams could not simply flood the spectrum and interfere with their rivals. The only way to win was to collaborate, ensuring that the entire ensemble could establish critical communications, while maximizing spectral reuse, stability, and data rates for their own data links.
The first round began in ominous fashion for University of Florida-based Team GatorWings, which barely avoided elimination at the first hurdle in the urban scenario, “Alleyways of Austin.” However, they quickly found their stride, impressively navigating the subsequent challenges, and eventually being crowned champions ahead of Vanderbilt University’s Team MarmotE, and the independent trio Zylinium.
Congratulations, Team GatorWings. You came, you saw, you conquered!