DARPA wants engineers to rethink conventional approaches for breakthrough Spectrum Management
by noreply@blogger.com (brian wang) from NextBigFuture.com on (#2HXVG)
DARPA earlier this month hosted the Battle of the ModRecs-a low-key competitive opportunity for engineers with a penchant for antennas and algorithms to test their skills in modulation recognition.
"We're looking to push modulation recognition out of its comfort zone," said DARPA program manager Tom Rondeau. "We want scientists and engineers to rethink conventional approaches and advance the technology to new heights, so it will function dynamically and with precision-not just under laboratory conditions but in real-world scenarios."
Held for a few days in conjunction with the IEEE DySPAN conference in Baltimore, where Rondeau and staffers created a sort of pop-up spectrum testbed, the Battle put hand-coded, expert systems against newer, experimental systems designed to take advantage of recent advances in machine learning. "In this realistic scenario with complex waveforms, the hand-coded systems performed better than the machine-learning systems," Rondeau said. "But it was close, and we don't think that lead is going to last for long. We now have a better understanding of the state of the art and which directions to explore as we pursue our goal of more effectively managing the spectrum."
Read more
"We're looking to push modulation recognition out of its comfort zone," said DARPA program manager Tom Rondeau. "We want scientists and engineers to rethink conventional approaches and advance the technology to new heights, so it will function dynamically and with precision-not just under laboratory conditions but in real-world scenarios."
Held for a few days in conjunction with the IEEE DySPAN conference in Baltimore, where Rondeau and staffers created a sort of pop-up spectrum testbed, the Battle put hand-coded, expert systems against newer, experimental systems designed to take advantage of recent advances in machine learning. "In this realistic scenario with complex waveforms, the hand-coded systems performed better than the machine-learning systems," Rondeau said. "But it was close, and we don't think that lead is going to last for long. We now have a better understanding of the state of the art and which directions to explore as we pursue our goal of more effectively managing the spectrum."
Read more