The Data Science Institute and the Electrical Engineering Department at Columbia University received a $650,000 National Science Foundation grant to develop energy-efficient sensors that allow mobile- and wireless-device users to tap into available unused channels in the radio-frequency spectrum. The sensors will enable future communication systems to flexibly share the spectrum. Wireless communications and mobile applications have placed an enormous strain on the electromagnetic spectrum, which is a finite and limited resource. "At some point in the future, as we keep using more and more mobile devices, the spectrum will run out of space," said John Wright, a DSI affiliate and electrical engineering professor who is the principal investigator on the project. "We'll use all the data-science tools we possess — machine learning, neural networks, algorithms and advanced computation techniques, in conjunction with new hardware devices — to sense pieces of the RF spectrum as they become available." Wright said that Peter Kinget, an electrical engineering professor at Columbia who specializes in analog and radio frequency(RF) integrated circuits, will design circuits that can create snapshots of a large portion of the spectrum. Wright will then use a few of the snapshots to design algorithms to reconstruct the spectrum and help design a more energy-efficient sensor. The project mixes the latest computational methods with novel hardware design. Wright will lead a team to develop algorithms and machine-learning methods to model and predict the available areas of the spectrum. Kinget's team will design circuits to sense the available channels in the spectrum. "This project builds upon our ongoing fruitful collaboration with John's team," Kinget said. "In the past couple of years, we have demonstrated several RF spectral sensors that generally used off-the-shelf signal-processing approaches with our custom hardware and have demonstrated significant speed and energy benefits. It will be exciting to see how much more progress we can make using new algorithms built on the latest insights in signal processing."