AI Systems Collaborate to Master Shrinking Spectrum Availability

Military AI

The Pentagon’s plan to use machine intelligence to better manage the finite and increasingly crowded electromagnetic spectrum took a small leap forward last month, as the Defense Advanced Research Projects Agency (DARPA) hosted Phase 2 of its Spectrum Collaboration Challenge (SC2).

As 15 teams demonstrated at Johns Hopkins University Applied Physics Laboratory (APL) in Laurel, Md., artificial intelligence can manage a shifting menu of radio frequencies. The technology reached an inflection point, or at least displayed “a technological shift,” as DARPA Program Manager Paul Tilghman called it. “For the first time, we saw autonomous collaboration outperform the status quo for spectrum management,” Tilghman said. That is, the machines managed the spectrum better than humans do, which the Department of Defense sees as an essential step to maintaining a full range of communications in the increasingly crowded, complex, and contested spectrum arena.

The electromagnetic spectrum is involved in virtually every phase of military operations, including wireless communications with bases and deployed units, satellite operations, cybersecurity, aircraft and naval operations, weapons systems, surveillance drones, and internet of things (IoT) devices. The need to keep all those lines of communication open among the military services and allies is compounded further by increased demands for commercial use– the military is in the process of turning over up to 500 MHz of spectrum to the private sector by 2020–as well as the realities of electronic warfare (EW).

The DoD has increased its emphasis on EW, making it a key part in the plans of the military services. The Army, for instance, is planning to add EW platoons to each brigade combat team’s military intelligence company as part of the expansion of its cyber forces. During exercises last summer, the Army conducted its first electronic attack in Europe since the Cold War.

Other countries–most notably Russia during its incursion into Ukraine in 2014–are making greater use of EW, which has prompted the DoD to step up its own pace.

The scarcity of available spectrum and the interference caused by adversaries means that demands on spectrum are fast-moving, changing from day to day or even moment to moment. Managing access requires tactics for adroitly sharing frequencies, preventing disruptions in service, and finding ways to make use of every available hertz, in a fluid rather than static environment. The Defense Information Systems Agency, DARPA, and other organizations are looking to machine learning to facilitate management.

During DARPA’s Spectrum Collaboration Challenge, which it said was the first of its kind, the 15 teams competed in more than 400 matches overall at APL’s Colosseum, a large radio-frequency testbed developed specifically for the SC2 competition. The teams competed in six scenarios, such as Wildfire, which simulated an emergency response situation, and Alleys of Austin, which replicated the communications challenges of soldiers working their way through an urban environment. “This real-world relevance was critical for us as we want to ensure these technologies can continue to develop after the event and can transition to commercial and/or military applications,” Tilghman said.

Six top-scoring teams came away with prizes of $750,000 each, and all 15 are eligible for the next and final stage, the 2019 Spectrum Collaboration Challenge grand finale Oct. 23 during the three-day MWC19 telecommunications show in Los Angeles. Three winning teams in that showdown will win prizes of $2 million, $1 million, and $750,000 respectively. DARPA, meanwhile, hopes to come away from the competition with machine-learning technologies that will help the DoD maximize use of the spectrum.

“We’re very encouraged by the results we saw at PE2 [the second preliminary event of SC2],” Tilghman said. “The teams’ radios faced new and unexpected scenarios but were still able to demonstrate smart, collaborative decision making. PE2 showed us that AI and machine learning’s application to wireless spectrum management creates a very real opportunity to rethink our current century-old approach.”

Recent