Up for the challenge? Army wants help in applying AI, machine learning to signal detection
By Ms. Nancy Jones-Bonbrest, Army Rapid Capabilities Office
ABERDEEN PROVING GROUND, Md. (April 24, 2018) -- Digitally speaking, it's loud out there. Cell phones, satellite signals, radars and more are making the electromagnetic spectrum a crowded and congested space.
To cut through the noise and increase the speed at which Electronic Warfare Officers can identify and react to these signals on the battlefield, the Army's Rapid Capabilities Office is seeking new technologies that apply artificial intelligence and machine learning to paint a picture of the electromagnetic spectrum.
The RCO is inviting anyone with a possible capability to participate in the Army Signal Classification Challenge, scheduled to go live on Challenge.gov and FBO.gov on April 30. At this time, registration will be open and upon approval, will allow competitors access to the training dataset consisting of over 4.3 million instances across 24 different modulations, which includes a noise class. The effort is seeking solutions that can perform blind signal classification quickly and accurately.
The RCO is hoping to reach highly skilled data scientists and innovation leaders who can bring machine learning algorithms and the supporting processes, methods and tools needed to improve the speed and agility of identification and classification of various signals within the electromagnetic spectrum.
"This is a competition to find the 'best of the best' in artificial intelligence and machine learning that can do blind signal classification," said Rob Monto, the director of the RCO's Emerging Technologies Office. "We are hoping to attract everybody and anybody that has potential solutions in this space and we are making the process as open as possible."
Specifically, blind signal classification requires little-to-no prior knowledge about the signal being detected in that specific instance. The solution would automatically classify the modulation, or change of a radio frequency (RF) waveform, as a first step towards signal classification.
"The challenge allows participants to submit their scores daily during the challenge to see how well they are performing," Monto said. "There will be a status leader board, so everyone will be updated automatically on how they are placing against others. This will help drive competition."
A total of $150,000 will be awarded, with the winner receiving $100,000, second place receiving $30,000, and third place receiving $20,000. The intent is to also include the potential for a follow-on opportunity for possible contract awards.
To participate in the first phase of the challenge, competitors can visit: http://www.challenge.gov/challenge/army-signal-classification-challenge starting April 30, for an overview of the challenge. From there, they will be directed to a third-party site to accept the terms and conditions prior to establishing an account and accessing the competition platform and test data sets. There will also be a link on Federal Business Opportunities starting April 30.
The Army Signal Classification Challenge competition will be open for approximately 90 days from the initial launch date. Participants will have at least 60 days to develop their models and work with the training data, with shorter timeframes allowed for each test dataset submission. There will be two test datasets provided that will be the basis for judging submissions. The first will be released approximately 67 days after the challenge launch, with a solution submission window of 15 days. A second, more complex test dataset will be released 84 days after the challenge launch, with a shorter submission window of only seven days. Participants' overall challenge score will be based on a combined weighted score for both test datasets. Participants will be able to see how they are performing in relation to others, in real time via the participant leaderboard. At the end of the competition the top three competitors who meet all the terms and conditions of the challenge will receive a monetary award.
The RCO is utilizing both Challenge.gov and FBO.gov to attract a broad range of potential solutions in applying artificial intelligence and machine learning to signal detection to combat the growing complexity of the electromagnetic spectrum. Currently, the classic signal detection process is no longer efficient in understanding the vast amount of information presented. This additional complexity involves identifying the source and location of a potentially infinite number of electronic signals through various sensing means, which require new and innovative approaches to modern era signal problems.
While Electronic Warfare Officers (EWOs) will remain in the driver's seat for pinpointing various signals, analyzing their impact, and making recommendations to their commanders, the use of artificial intelligence and machine learning can help Soldiers to detect patterns and zero in on areas of significance while filtering out clutter.
The RCO recently delivered new electronic warfare prototype systems in response to an Operational Needs Statement from U.S. Army Europe. Soldiers can use the equipment to implement electronic protection for their own formations, as well as to detect and understand enemy activity in the electromagnetic spectrum and disrupt adversaries through electronic attack effects. The Army Signal Classification Challenge and other RCO initiatives could eventually enhance the initial prototypes while informing the Army's electronic warfare programs of record.
"We want to apply emerging technologies like artificial intelligence to the domain space the Army RCO is working in. We can help reduce the cognitive burden of EWOs," Monto said.