Machine Learning Makes Sense of RF Signals

DARPA funded project sees machine learning being used for making sense of radio frequency signals in crowded electromagnetic environments.


adio frequency signals for controlling and communicating with drones, vehicles and a variety of equipment, both civilian and military, are competing for the airwaves with

other signals in a radio frequency environment that is becoming increasingly crowded. Whilst this environment has always seen an increase in density, the Internet of Things (IoT) has accelerated this increase to levels that require new techniques to manage. In response to this challenge, defence contractor

BAE Systems has been awarded funding from the Defence Advanced Research Projects Agency (DARPA) in the USA to integrate machine-learning (ML) technology into platforms that decipher radio frequency signals. Its “Controllable Hardware Integration for Machine-learning Enabled Real- time Adaptivity” (CHIMERA) system provides a reconfigurable hardware platform for ML algorithm developers to make sense of radio frequency (RF) signals in increasingly crowded electromagnetic spectrum environments. The contract, which is worth up to $4.7 million, is dependent on the successful completion of milestones, which include hardware delivery along with integration and demonstration support. DARPA is highly active in encouraging software based innovation in communications for military systems and has active projects in the design of cyber- physical systems, swarm autonomy testing and the development of methods for unlocking the potential of the RF spectrum using artificial intelligence.

RF MACHINE LEARNING The contract follows on from a previous similar award from DARPA for the development of data- driven ML algorithms under the same Radio Frequency Machine Learning Systems (RFMLS) programme.

❱ ❱ Dense RF environments will be able to be analysed in both civilian and military applications using machine learning technology

CHIMERA’s hardware platform will enable

algorithm developers to decipher the ever-growing number of RF signals, providing both commercial and military users with greater automated situational awareness of their operating environment. RFMLS requires robust, adaptable hardware with a multitude of control surfaces to enable improved discrimination of signals in the kind of dense spectrum environments that are likely to exist in the future. According to Dave Logan, vice president

and general manager of Command, Control, Communications, Computers, Intelligence, Surveillance and Reconnaissance (C4ISR) Systems at BAE Systems, CHIMERA brings the flexibility of a software based system to hardware. “Machine- learning is on the verge of revolutionising signals intelligence technology, just as it has in other industries,” he says.

EVOLVING THREATS In an evolving threat environment, CHIMERA will enable ML software development to adapt the hardware’s RF configuration in real time to optimise mission performance. This capability has never before been available in a system based on hardware alone. The system provides multiple control surfaces for the user, enabling on-the-fly performance trade-offs that can maximise its sensitivity, selectivity and scalability depending on mission requirements. The system’s open architecture interfaces allow for third party algorithm development, making it future-proof and easily upgradable upon deployment. Other RF functions, including communications,

radar and electronic warfare, can also benefit from this agile hardware platform, which has a reconfigurable array, front-end, full transceiver and digital pre-processing stage. Work on these phases of the programme will take place at BAE Systems’ sites in Hudson and Merrimack, New Hampshire, and Dallas, Texas in the USA.

Electronics Testing 2019 /// 5

Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32