Communications of the ACM
AI Contest Aims to Build the Communications Network of the Future
The U.S. Defense Advanced Research Project Agency's Spectrum Allocation Challenge calls on participants to reimagine spectrum access strategies and develop a new wireless paradigm for how radio networks share the RF spectrum.
Credit: U.S. Department of Commerce
The tech and military communities are abuzz with the promise and peril of artificial intelligence (AI). Some believe machine learning—a subset of AI that allows machines to improve themselves—will usher in a new era of technological progress greater than the dawn of the Internet. Others warn of the danger that this type of technology may improve itself enough to become extremely powerful—and extremely dangerous.
Leveler heads are using machine learning to solve problems human minds cannot tackle alone. One such problem seems mundane on the surface, but has profound implications for how the world works and wars: the limitations of the radio frequency (RF) spectrum.
The RF spectrum spans three kilohertz (kHz) to 300 gigahertz (GHz), and includes all the wavelengths we use for wireless communications, yet it operates by a decidedly unmodern methodology. The frequency is divided into “exclusively licensed bands” allocated over “large, geographically defined regions,” explains the Defense Advanced Research Projects Agency, better known as DARPA, a research wing of the U.S. Department of Defense.
At any moment, many frequency bands may be completely unused (such as the frequencies allocated to firefighters, if they are not battling a blaze), while others are being overused (like consumer cellular wavelengths on a Saturday night), creating what DARPA calls “unnecessary conditions of scarcity.”
Dividing the RF spectrum into discrete bands makes sure different types of communications do not interfere with one another; on the other hand, it is woefully inefficient. Because all traffic is routed through a relatively small number of licensed bands, much of the RF spectrum is not used at any given time.
That is why DARPA is sponsoring the Spectrum Collaboration Challenge (SC2), a competition with prize money in the millions of dollars to improve how the RF spectrum functions.
Products ranging from commercial health trackers to unmanned military aircraft to consumer smartphones to emergency services radios communicate wirelessly via portions of the RF spectrum. Billions of devices are connected to the Internet, and more come online every second (Cisco Internet Business Solutions Group has forecast the growth of the Internet of Things will result in 50 billion devices connected to the Internet by 2020), so the Challenge could not be more timely.
The purpose of the Challenge, says DARPA, is “to ensure that the exponentially growing number of military and civilian wireless devices will have full access to the increasingly crowded electromagnetic spectrum.”
To truly modernize this system, second-by-second regulation of RF usage would be needed, and DARPA believes machine learning provides the means to accomplish that. In fact, the terms of the Spectrum Collaboration Challenge specifically call for a machine learning solution to the problem:
Competitors will reimagine spectrum access strategies and develop a new wireless paradigm in which radio networks will autonomously collaborate and reason about how to share the RF spectrum, avoiding interference and jointly exploiting opportunities to achieve the most efficient use of the available spectrum.
DARPA anticipates contestants will achieve these goals by leveraging breakthroughs in AI and machine learning, as well as by using the increasingly better capabilities of software-defined radio. The goal is spectrum abundance: a communications paradigm that will work for everyone all the time and will dramatically increase the bandwidth available to devices and machines. The resulting system, hopes DARPA, will be a hyper-efficient spectrum management system that regulates how billions of communications are allocated across frequency bands in real time.
“We’ve seen AI excel at playing games like chess and Go,” says Spectrum Collaboration Challenge program manager Paul Tilghman. “Games are good, but can AI help us take action to solve a problem that impacts all of us every second of every day, and help us get to the future we are ultimately headed toward—one in which almost everything communicates wirelessly by way of the electromagnetic spectrum?
“Ensuring that this future of ubiquitous wireless connectivity is possible is what the Spectrum Collaboration Challenge is all about.”
The organization is putting big money up to make that dream a reality. DARPA hopes $2 million in first-prize money will compel participation. In all, $3.75 million in total is up for grabs, says DARPA public affairs representative Ivan Amato, with second place netting $1 million, and third place $750,000.
DARPA plans on funding some challenge participants; Amato says contest organizers have not yet announced which proposals they will fund, or to what extent. Regardless, DARPA will get some help from the National Science Foundation (NSF), which in July announced it will fund select challenge participants who do not receive DARPA funding for the Spectrum Collaboration Challenge. This includes students at universities who may be drafted by contest participants to work on the challenge. The play is just another step in the NSF’s spearheading of the Advanced Wireless Research Initiative launched by President Obama in July 2015, says NSF public affairs specialist Aaron Dubrow.
“Part of these investments will support researcher and student participation in DARPA’s SC2,” he says. “Like NSF’s other targeted awards to further student research efforts, participants in the DARPA SC2 will engage in spectrum research for discrete periods of time – from a few weeks to a year, as determined by the awardee—allowing them to gain exposure to the challenges in the field and the ways scientists address these challenges through research and experimentation.”
This kind of broad academic exposure to spectrum research, the NSF hopes, will pay dividends beyond solving the DARPA challenge.
“We view enabling the participation of students in the DARPA SC2 as an opportunity to support the education of a next-generation workforce of researchers and professionals in wireless technologies, including spectrum management,” says Dubrow.
The Federal Communications Commission (FCC) did not respond to repeated requests for comment on how advancements from the contest might integrate into the RF spectrum, which it regulates. According to a press release that includes announcement of the Spectrum Collaboration Challenge, the NSF says “important spectrum policy work carried out by the Federal Communications Commission (FCC), will drive the information technology- and innovation-based economic growth at the heart of the American economy,” though it does not mention that body’s role in the contest.
Logan Kugler is a freelance technology writer based in Tampa, FL. He has written for over 60 major publications.
No entries found